<%BANNER%>

How Does Migration Affect Agricultural Labor Productivity? The Case of Mexican Rural Households

Permanent Link: http://ufdc.ufl.edu/UFE0022736/00001

Material Information

Title: How Does Migration Affect Agricultural Labor Productivity? The Case of Mexican Rural Households
Physical Description: 1 online resource (112 p.)
Language: english
Creator: Ramirez Rodrigues, Melissa
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2008

Subjects

Subjects / Keywords: Food and Resource Economics -- Dissertations, Academic -- UF
Genre: Food and Resource Economics thesis, M.S.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Using micro-level agricultural data from the ENHRUM survey, I examine the impact of international labor-out migration on the agricultural production of Mexican rural households. The study evaluates how households reallocate labor and capital resources as consequence of labor out-migration and incorporates a productivity variable to measure the efficiency of this reallocation. Estimating a Heckman Two-Stage model we capture the labor productivity of the household accounting for the selectivity of landholding. The results suggest that international labor-out migration and the formation of social networks have a negative impact on the household labor productivity. Migrant households are less labor productive than households with no migratory experience by 28,655.31 Mexican pesos. It seems that migrant households are not investing enough in capital-intensive resources to compensate for the reduction in labor supply. Changes in the intra-household allocation of labor are not observed. The education level of the household head and spouse, the tenancy status of the land and the location of the household are other factors affecting the labor productivity of the household.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Melissa Ramirez Rodrigues.
Thesis: Thesis (M.S.)--University of Florida, 2008.
Local: Adviser: Carrion-Flores, Carmen.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2010-08-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2008
System ID: UFE0022736:00001

Permanent Link: http://ufdc.ufl.edu/UFE0022736/00001

Material Information

Title: How Does Migration Affect Agricultural Labor Productivity? The Case of Mexican Rural Households
Physical Description: 1 online resource (112 p.)
Language: english
Creator: Ramirez Rodrigues, Melissa
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2008

Subjects

Subjects / Keywords: Food and Resource Economics -- Dissertations, Academic -- UF
Genre: Food and Resource Economics thesis, M.S.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Using micro-level agricultural data from the ENHRUM survey, I examine the impact of international labor-out migration on the agricultural production of Mexican rural households. The study evaluates how households reallocate labor and capital resources as consequence of labor out-migration and incorporates a productivity variable to measure the efficiency of this reallocation. Estimating a Heckman Two-Stage model we capture the labor productivity of the household accounting for the selectivity of landholding. The results suggest that international labor-out migration and the formation of social networks have a negative impact on the household labor productivity. Migrant households are less labor productive than households with no migratory experience by 28,655.31 Mexican pesos. It seems that migrant households are not investing enough in capital-intensive resources to compensate for the reduction in labor supply. Changes in the intra-household allocation of labor are not observed. The education level of the household head and spouse, the tenancy status of the land and the location of the household are other factors affecting the labor productivity of the household.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Melissa Ramirez Rodrigues.
Thesis: Thesis (M.S.)--University of Florida, 2008.
Local: Adviser: Carrion-Flores, Carmen.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2010-08-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2008
System ID: UFE0022736:00001


This item has the following downloads:


Full Text





HOW DOES MIGRATION AFFECT AGRICULTURAL LABOR PRODUCTIVITY?
THE CASE OF MEXICAN RURAL HOUSEHOLDS





















By

MELISSA A. RAMIREZ RODRIGUES


A THESIS PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
MASTER OF SCIENCE

UNIVERSITY OF FLORIDA

2008


































2008 Melissa A. Ramirez Rodrigues


































To all those who follow their dreams









ACKNOWLEDGMENTS

My deepest gratitude and love go to my mentors in life: my beloved parents, Agustin

Ramirez and Aidil Rodrigues; and my dearest sister, Milena. Their wisdom, dedication and

unconditional love helped me achieve everything I have in life and helped me fulfill my dreams.

I would also like to thank my chair, Dr. Carmen Carrion-Flores; and my supervisory committee,

Dr. Carmen Deere for the support and guidance that made all this work possible.











TABLE OF CONTENTS
page

COPYRIGHT 2....___ .... ... ...............

DEDICATION ................................. ..................... .........3

ACKNOWLEDGMENTS ........... ......... ......... .........................4

LIST OF TABLES ......................... ........ ..... ...................... ........ .7

LIST OF FIGURES ................................... ............ .. .............................8

A B STR A C T ................... ......... ..........................................................................9

CHAPTER

1 INTRODUCTION ............... ...................... .............................. .............. 10

2 OVERVIEW OF THE AGRICULTURAL SECTOR IN MEXICO.............................. 13

Introdu action ............................ ............... 13
Review of A agricultural Policies in M exico .......................................................................... 13
Land Reform s in M exico............................................................... ... ......... 14
D om estic M market Intervention .................................................................................... 16
International Trade Regulation.................................... ......... 19
Im plications for the A agricultural Sector ...................................................................................21
Conclusion ................... ................... ........................................... ........ 28

3 MIGRATION AND AGRICULTURAL PRODUCTIVITY ...............................................29

Introduction.................2...........9
Mexican Migration Literature.........................................30
Immigration Reform s in the United States............................................................. 30
The Bracero program ........................................................30
Immigration Reform and Control Act (IRCA)...................................................31
Border enforce ent ................... ............................. ......... .... ...... ............... 32
The Evolution of M exican M igration.................................................................... 32
M igratory patterns .............. ........................ ............. .... ...... ...... 32
M igratory flow s ................... .............................. ...... ...... .. ......... .35
Causes of Migration ................... ... ... ........................37
Impact of Migration in the Sending Communities............... ..................38
Remittances ................................................39
Hometown associations........................................41
Labor Out-Migration and the Agricultural Productivity Literature.............. ...42
Changes in Farming Practices and Decisions.............. .....................44
Gender Productivity .......... ......... ..................49










C conclusion ............. ................... ........................................ 51

4 DATA ANALYSIS AND METHODOLOGY........................................ 52

Introduction.......................................................... .........52
Data and Descriptive Statistics ............................................ ........ .....53
The Mexican National Rural Household Survey (ENHRUM)...................................... 53
Sam ple D description ...................... ...... ................55
Landholder and non landholder mean differences ......................................57
The landholder sam ple ........................ ...... ......................59
M igrant and non migrant mean differences ....................... ....................... 61
Description of landholding households accounting for migratory status..............63
H eckm an Tw o-Stage Procedure .................................................. ............... 65
Variables D description ........................................ ........ .... .. ..............67
Dependent Variables ................................................. ..............67
Independent V ariables .................. ............. .. ........... ................... ..68
Endogeneity Issues ................................................71
Conclusion......................................................... ................ ........ 73

5 RESULTS ...........................................................................84

Introduction.............................................. .........84
Estimation of OLS ................................................84
Heckman Two-Stage Estim ation................................................................ .. ............. 86
Who holds Land?............................... .................................86
What affects Labor Productivity? .............. ... ................... ................... ...........88
Model 1 household head labor productivity (sample A)......................................88
Model 2 household labor productivity (sample B).......................................90
Addressing M igration Endogeneity ............................................................ ...... ......... 91
Conclusion......................................................... ................ ........ 93

6 CONCLUSION.........................................................101

LIST OF REFERENCHES ......................................................... ..............106

BIOGRAPHICAL SKETCH ................................................................... ........... 112
















6











LIST OF TABLES


Table page

4-1 The ENHRUM community codes............................... ...............75

4-2 Landholder and non landholder mean differences......................................................77

4-3 Migrant and non migrant mean differences.......................................................... 78

4-4 Migrant landholder and non migrant landholder mean differences............... ...............79

4-5 Selection equation variables statistics............................................................................... 80

4-6 Regression equation variables statistics.................................................................. 81

5-1 Model 1 household head labor productivity. .......................................94

5-2 Model 1 Heckman selection model (first-stage) ................................... .....95

5-3 M odel 2 household labor productivity ................................... ..................... 96

5-4 Model 2 Heckman selection model (first-stage) ................................... .....97

5-5 L ogit m odel for m migration .............................................................98

5-6 Model 2 household labor productivity solving for the endogeneity problem ................99

5-7 Model 2 Heckman selection model solving for the endogeneity problem (first-stage) ..100











LIST OF FIGURES

Figure page

4-1 A ge of the household head........................................... .... .......................... ............82

4-2 Education level of the household head ...... ............ ............................82

4-3 Number of members in the household.........................................................................82

4-4 Comparison of education level by landholding status........................................... 83

4-5 Landholder and non landholder spatial distribution ................. ................. ..........83

4-6 Migrant and non migrant spatial distribution......... ....... ................ 83









Abstract of Thesis Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Master of Science

HOW DOES MIGRATION AFFECT AGRICULTURAL LABOR PRODUCTIVITY?
THE CASE OF MEXICAN RURAL HOUSEHOLDS

By

Melissa A. Ramirez Rodrigues

August 2008

Chair: Carmen Carrion-Flores
Major: Food and Resource Economics

Using micro-level agricultural data from the ENHRUM survey, I examine the impact of

international labor-out migration on the agricultural production of Mexican rural households.

The study evaluates how households reallocate labor and capital resources as consequence of

labor out-migration and incorporates a productivity variable to measure the efficiency of this

reallocation. Estimating a Heckman Two-Stage model we capture the labor productivity of the

household accounting for the selectivity of landholding. The results suggest that international

labor-out migration and the formation of social networks have a negative impact on the

household labor productivity. Migrant households are less labor productive than households with

no migratory experience by 28,655.31 Mexican pesos. It seems that migrant households are not

investing enough in capital-intensive resources to compensate for the reduction in labor supply.

Changes in the intra-household allocation of labor are not observed. The education level of the

household head and spouse, the tenancy status of the land and the location of the household are

other factors affecting the labor productivity of the household.









CHAPTER 1
INTRODUCTION

Migration is increasingly being considered in the literature on agricultural productivity in

developing countries, it has become a common practice for rural households worldwide.

Migration has been found to affect a household's decisions in three important ways. First,

migration reduces the labor availability of the household; second, it generates an increase in the

household's income through remittances sent by the migrant; and third, it strengthens the

formation of social networks, which can be used to promote the migration of other household

members.

In the case of Mexico, migration is one of the off-farm activities that rural households

rely upon heavily. According to the Consejo Nacional de Poblacion (CONAPO), the number of

Mexicans engaging in a migratory experience to U.S. reached 3.3 million between 1990 and

2000. Furthermore the composition of the annual net flow to this country has increased by a

factor of three in the last three decades, leading to the formation of a Mexican migrant

community in the U.S. that reached 26.7 million in 2003.1

Thus far the studies on migration have not reached a consensus on the way migration

influences the farming practices and decisions of the household. However, it has been noticed

that the household's initial endowments as well as the type of migration lead to different effects

of labor out-migration on the sending community. Furthermore, it has been found that, in

general, remittances are used to relax credit constraints and improve the farm management

practices of the household.

Analysis of the way in which Mexican international migration affects the labor

productivity in rural households is scarce. Existing studies suggest farming differences between

1 Approximately 9.9 million people represent the migrants that were born in Mexico and the remaining 16.8 millions
represent the population already born in the United States but with Mexican heritage.









migrant and non migrant households depend not only on the household's initial endowments but

also on how labor-intensive the farming practices of the households are, and the household's

ability to substitute the family labor with reciprocal or wage labor i.e.; it depends upon how rural

labor markets work.

This study central research question is the impact of international migration on the

household labor productivity. The study aims to evaluate the reallocation of labor and capital

resources as a consequence of labor out-migration and measure the efficiency of this

reallocation. My primary hypothesis is that, as a household strategy to manage uncertainty and

market imperfections, migrant households maintain their agricultural production levels by

investing more in capital-intensive inputs to compensate for the reduced labor force availability

due to the migration of at least one of the household members. My corollary hypothesis is that

labor productivity, measured as the agricultural output generated per day of work, will be greater

in migrant households compared to non migrant households. To test these hypotheses, my study

employs econometric techniques using the Mexican National Rural Household Survey

(ENHRUM), a nationally representative sample of 1765 household co-directed by the Colegio de

Mexico (COLMEX) and the University of California at Davis in 2002/2003.

The contribution our study to the existing literature focuses on three main points. First,

we rely on the New Economics of Labor Migration (NELM) approach, using the household as

the unit of analysis to study the way labor out-migration influences the labor productivity of rural

households. Second, we estimate labor productivity accounting for the selectivity of whether or

not a household has access to land. The idea behind this is that agricultural productivity can only

be measured for those households holding land, and until now no study recognized this









selectivity when studying labor productivity. Finally, we introduce into the labor productivity

analysis the study of social networks.









CHAPTER 2
OVERVIEW OF THE AGRICULTURAL SECTOR IN MEXICO

Introduction

The purpose of this chapter is to review the status of the agricultural sector on the past

three decades. A review of past and current agricultural policies in Mexico is crucial in

understanding the way these changes have currently influenced the agricultural sector. These

agricultural policies have directly or indirectly changed farmers' agricultural practices affecting

how the agricultural sector operates as well as its productivity. This chapter is structured as

follows. The first section summarizes the domestic and international agricultural policies

institutionalized in Mexico during the twentieth century. The second section reviews the

implications of the policies for the agricultural sector. Finally, some conclusions are presented in

the third and last section.

Review of Agricultural Policies in Mexico

At the beginning of the twentieth century agriculture employed an important share of the

labor force but most of the agricultural workers were landless (Fernandez-Comejo and

Shumway, 1997; Villa-Issa, 1990). The concentration of land in the hands of a few and the

inequalities among social classes were two factors leading to the Mexican Revolution (1910-17).

After the Revolution and during the 1920's the agricultural sector received little or no

investment. It was not until the 1930's that there was a substantial increase in public investment

in the agricultural sector, such as the construction of roads, irrigation systems and the

intensification of the land reform. From the 1930's until the 1980's the government played a key

role in the development of the sector, for instance, with the creation of the ejido and the

institutionalization of CONASUPO (Yunez-Naude, 2003; Yunez-Naude and Barceinas, 2000).

The next sub-sections describe these in detail.









Land Reforms in Mexico

Following the Mexican Revolution, the first land reform2 was particularly important

because it not only reallocated the possession of the land, but also set the foundation for the

contemporary agrarian system. In essence, this reform encompassed a new system of tenancy

called the ejido system, which consisted in communal land possession but generally individual

farming. The ejido was made up of the ejidatarios, who are the farmers who have rights to the

ejido land (called agrarian reform rights). This type of ownership has remained effective until the

present.

During the government of Lazaro Cardenas (1934-1940) there was a large-scale

redistribution of land. By 1940, the ejido sector possessed 22.5 per cent of the agricultural land

and 47.4 per cent of the arable land of the country (Assies, 2008). Two additional presidential

periods characterized by important redistribution of land were those of Gustavo Diaz Ordaz

(1964-1970) and Luis Echeverria (1970-1976). During these two periods, however, no

significant amount of irrigated land was redistributed.

There were three major restrictions imposed on the way the ejido operated. First, there

was a labor restriction, where the ejidatarios were not allowed to hire labor. Second, if the

ejidatarios resided away from their allocated land for more than two years, they ran the risk of

losing their ejidal rights. Moreover, within this system, long-term production contracts with

farmers outside the ejido were not allowed (Johnson, 2001). In practice, however these

restrictions were not always followed. For instance, illegal renting of ejido land to commercial

farms was a common practice among farmers as was migration (Assies, 2008).

With the creation of ejidos the government aimed to promote productivity and satisfy the

internal market for agricultural products. In its early stage, the land agrarian reform was backed

2 The first land reform consisted in the 1917 amendment of Article 27 of the Constitution.









up by technical assistance, credit and supply of seeds. However, with the instauration of an

import-substitution industrialization model in the early 1930s, the policy gradually shifted away

from the agrarian sector toward the industrial sector favoring the provision of cheap food for an

increasingly industrializing country (Assies, 2008).

The "social sector" consisting of ejidos and agrarian communities was thus confined to

the production of staples under price regulation and subsidies. At the same time, however,

policies promoting the investment in irrigation systems and capital-intensive production favored

the development of the private sector and the production of high value exports, giving rise to the

formation of a dual agrarian structure and a deepening in the existing regional differences that

persists until nowadays (Assies, 2008).

By the early 1990s the ejido system accounted for approximately 100 million hectares

entailing half of the national farmland (Femandez-Cornejo and Shumway, 1997). Furthermore,

the land was distributed to nearly 3 million peasants that represented about three quarters of total

producers, grouped in 26,796 ejidos and 2,366 agrarian communities (Quintana, Borquez and

Aviles, 1998).A typical ejido would consist of approximately 74 ejidatarios and possess some

2,000 hectares The average ejidatario would hold 9.2 hectares in two parcels and have access to

28 hectares of common land (Assies, 2008).

One of the most radical policies institutionalized in Mexico during the 1990s was the

second land reform or counter reform. The 1992 amendment of Article 27 of the Constitution put

an end to the land redistribution process existent in the country since the 1930s. This reform

aimed to transform the collective possession of land into an individual possession, setting the

conditions necessary to start the privatization process of land. It also laid the foundations for

trade liberalization of the agricultural sector (Fernandez Cornejo and Shumway, 1997). It was









believed the reform was going to help to overcome the crisis in the sector through the expansion

consolidation of rental markets, increased productivity and the promotion of foreign investment

(Assies, 2008). It has been noticed however, that "in a context of globalization and asymmetric

free-trade relations" (Assies, 2008, pg 33) the agrarian crisis has only intensified.

Under the 1992 reform, the ejidatarios were granted the opportunity to certify their land

rights if the ejido consented to participate in Procede.3 They were also allowed to hire labor and

grow any crop and market it wherever they wanted. Long-term production contracts with

outsiders were made also feasible. In a general way, this reform reintroduced a market oriented

scheme into the agricultural sector, allowing farmers to respond directly to market incentives and

disincentives.

One important feature of Procede is that the decision to participate in this certification

program not necessarily resulted in the privatization of the ejido. A governmental report

capturing information for 1992-2005, for example, showed that in total only 1% of the social

property entering the certification process achieved full private property status in this time

period. Moreover, 60% of this privatization was done for urbanization purposes (Assies, 2008).

This number clearly suggests that one of the main goals of the reform, to start the privatization

process of the land, in order to capitalize the ejidos has not been achieved.

Domestic Market Intervention

The Mexican government has always played an active role in regulating the agricultural

domestic market. For instance, in 1965 the government creased a state trading enterprise (STE)

called The National Company of Popular Subsistence (CONASUPO). The organization's main



3 Ejidatarios participating in PROCEDE had the right to legally sell, rent, sharecrop or mortgage their land. The
decision to sell ejido land to outsiders, however, required the approval of two-thirds of the ejido general assembly,
witnessed by a government representative.









objective consisted in promoting the domestic market by subsidizing both producers and

consumers and regulating international trade through direct imports (Yunez-Naude 2003).

In its first stages, CONASUPO was designed as an economic development tool to protect

small staple-farmers as well as low-income consumers. To protect producers, CONASUPO

absorbed the transaction costs farmers faced by reducing the number of intermediaries involved

in purchase-sale transactions; it also guaranteed crop support prices4. It also promoted

production subsidies, including input subsidies for water, electricity and fertilizers (Fernandez-

Cornejo and Shumway, 1997). CONASUPO also managed subsidiary programs for processing,

storing, distributing and selling the crops. At some point, CONASUPO exerted control over

30% of the total gross domestic agricultural production (Yunez-Naude, 2003). The most

important crop, however, was corn, representing 56% of the total value of crops managed by

CONASUPO (Yunez-Naude and Barceinas, 2000).

By the end of the 1980's some of the tasks performed by CONASUPO began to decline;

and by mid 1990's most of CONASUPO's programs were already dismantled6, privatized or

transferred to farmers. For instance, the processing of corn was privatized and the processing of

wheat to make bread was ended. In addition, the warehouses for basic crop storage belonging to

CONASUPO were transferred to farmers and local authorities.7 Finally, price intervention was

reduced to just corn and beans.

Corn and beans were the last two staple crops administered by CONASUPO since these

two crops, representing Mexico's two major staple crops produced by the larger number of

4 The agricultural crops involved in the CONASUPO's programs were barley, beans, copra, corn, cotton, rice,
sesame, sorghum, soybeans, sunflower and wheat.
5 To help low income consumers, CONASUPO sold basic foods to rural and urban costumers at very low prices.
Some of these goods included: corn, flour, wheat pasta, edible oils and fluid milk (Yunez-Naude, 2003).
6 The only two entities that survived this dismantling process were the LICONSA entity in charge of processing
milk powder to produce fluid milk for access to the poor at subsidized prices; and the retail store DICONSA
responsible for distributing basic food to low-income consumers at low prices.
7 Also, one of the extension programs called CECONCA, used for technical supports to farmers was also abolished.









peasant households, required a longer transformation period. By the end of 1995, CONASUPO

was still a "last resort" buyer of corn and beans at minimum prices. It was also in charge of

regulating the external trade of both crops. In 1998 however, CONASUPO's involvement in

social programs to assist the poor was ending, undermining the main reason for the existence of

the company. CONASUPO was subsequently liquidated during the Zedillo administration (1995-

2000) (Yunez-Naude, 2003).

In 1991 a new agency was created by the Agricultural Ministry called Support Services

for Agricultural Marketing (ASERCA) 8. This agency emerged as a substitute for CONASUPO

although ASERCA has no mandate with respect to price fixing commodity imports. Some of the

tasks this agency carries out include marketing and the coordination of direct income transfer

programs.

Between 1992 and 1996 and under the supervision of ASERCA, two programs were

developed, the Program of Direct Payments to the Countryside (Procampo) 9 and Alliance for

the Countryside. The goal of these two programs was to support agricultural producers without

interfering with the new rural market economy. Although the goal of these two programs is the

same, they differ in the way they are managed and funded. Alliance for the Countryside is state-

managed while Procampo is managed by the federal government. Moreover, a portion of the

Alliance for the Countryside program is funded using farmers' resources.

The Program of direct Payment to the Countryside (Procampo) replaced the traditional

price support system by an income direct payment for farmers based on the number of acres

devoted to the production of maize, beans, wheat, rice, cotton, soybeans, safflower, barley, and

sorghum. The Alliance for the Countryside provided farmers with financial aid, technical and


8 ASERCA stands in Spanish for Apoyo y Servicio a la Comercializaci6n Agropecuaria.
9 This program was expected to last 15 years ending in 2008.









marketing assistance and training. In essence, its objectives were to: (1) increase the investment

in capital intensity technology; (2) support the transformation of agriculture toward areas with

comparative advantage; (3) promote the creation of distribution channels for the products

commercialization (Cord and Wodon, 2001).

With the elimination of CONASUPO along with the creation of Procampo and Alliance

for the Countryside, Mexico was laying the foundations for the trade liberalization of the

agricultural sector. With these changes as well as other market oriented policies, Mexico was

preparing itself to enter into the General Agreement on Tariffs and Trade (GATT) as well as the

North American Trade Agreement (NAFTA)10.

International Trade Regulation

International trade for agricultural products was also regulated heavily by the Mexican

government. For instance, CONASUPO also had an active role regulating international trade

through direct imports in the early-mid 1980s"1; but just as with the domestic market,

CONASUPO's participation in trade regulation through the direct import of basic crops began to

decrease considerably in the following years. 12 The government also controlled trade volumes,

imposing tariffs, quotes and licensing requirements.

Other policies (outside the agricultural sector) that influenced the sector's performance

were exchange rates policies, investment policy in the rural sector, as well as state investment in

infrastructure, transportation and communication (Villa-Issa, 1998).13



10 Femandez-Comejo, 1997
11 CONASUPO accounted for 95% of total rice imports, 83% of corn imports and 68% of wheat import in 1983-88
period; for 99% of beans import from 1989 until 1993; and more than 95% of total sorghum and soybean imports at
the beginning of the 1970s (Yunez-Naude and Barceinas, 2000).
12 For example, the rice imported by CONASUPO reduced from 25% in 1989-1993 to zero in 1994-1996. Also, its
corn imports declined from 38% in 1989-93 to 16% 1994-96. And CONASUPO direct imports of beans and wheat
reached cero by the period 1994-96 (Yunez-Naude, 2003)
13 For instance, from 1955 to 1972 the scarce amount of private investments in the rural sector, due to a lag of 19%
in the farm prices, was offset by public investment. During that time, the exchange rate was also overvalued









In the mid 1980s Mexico started a series of adjustments to the existing economic model.

In 1986 Mexico became a member of the General Agreement on Tariffs and Trade (GATT) and

in 1994 Mexico was admitted into NAFTA. The admittance of Mexico represented an important

step toward a market-oriented strategy and consolidated many of the structural changes that

began in the early 1980s. The most important structural changes included the substitution of the

import substitution model by a market oriented model with a diminished participation of the

government. An important trade policy change was the shift from import licenses to tariff rate

quotas (TRQs).14

The North American Free Trade Agreement (NAFTA) included two separate agreements,

one between Mexico and the United States and the other one between Mexico and Canada. It

was agreed that import levels below the consented quota would not be subject to tariffs. A 15

year period (1994-2008) was set to eliminate the over quota tariffs for corn, dry beans and milk

powder (milk powder was not negotiated between Mexico and Canada).

Mexico has also signed other FTAs with Latin American and European countries. After

its incorporation as a full member of the WTO in 1995, Mexico agreed during the Uruguay

Round to set a tariff base of 25% for almost all agricultural products, with the promise of

reducing it an additional 1% by 2000. The basic crops subject to TRQs in the NAFTA

negotiation were also kept valid in this negotiation adding wheat to this type of trade regulation.

Canada and United States however have larger quota access and lower above quota tariffs.

In summary, this overview of the Mexican policy reforms is a starting point to understand

the agricultural sector. The next section focuses in explaining the implications for the agricultural

sector.


14 The crops that were changed from import licenses to TRQs during the NAFTA negotiation were: barley, beans,
corn and milk powder.









Implications for the Agricultural Sector

All the changes taking place in the agricultural sector -trade liberalization, reform of the

ejido and retreat of State- have led to a new incentive structure affecting farmers' behavior and

consequently the way they operate. This section analyzes the adjustment strategies farmers have

adopted as a consequence of these changes in the agricultural sector.

In general terms, the reforms were expected to have a positive impact on agricultural

sector productivity. Some of the expected results were: a decline of small, less productive

farmers, who under this new scenario would be willing to sell their land and move out of

agriculture; an increase in crop diversification toward more marketable crops; and finally, an

increase of capital intensification in the agricultural sector. For a number of reasons, however,

the reforms have not produced the expected results (Johnson, 2001; Cord and Wodon, 2001;

Davis, 2000; Assies, 2008).

There are different and inconclusive answers to this puzzling situation. According to

Davis (2000) for instance, farmers are assuming a risk-averse strategy, in which they abstain

themselves from incurring big changes and they diversify their sources of income in order to

reduce uncertainty. From this point of view, farmers can assure their subsistence by remaining in

the same crop production; not investing in technological inputs such as fertilizers, machinery and

improved seeds; and keeping a secure source of money through off-farm activities or migrant

remittances.

Assies (2008), on the other hand, suggests the changes taking place in the agricultural

sector -trade liberalization, reform of the ejido and retreat of State- have only deepened the crisis

in the rural sector because of the inaccessibility to credit15, insurance, market, modern inputs and


15 The total amount of credit the rural sector had access to decrease from 30 percent to 20 percent in 1997 (Assies,
2008).









technical assistance in the rural sector. For a better understanding of farmers' behavior and its

impact on productivity, I turn to how ownership of major agricultural assets has changed since

the policies reforms were implemented.

Landholdings experienced some changes after the ejido reforms. In an analysis done by

Davis (2000) for example, based on panel data for 1,287 ejido households, it was found that from

1994 to 1997, the amount of land owned by an individual increased on average by 25%, from 8

NRE16 hectares to 10 NRE hectares. The increase of owned land can be attributed to the fact that

common land owned by ejidos was divided after the reform was implemented, as well as to an

increase in land converted into agriculture.

In spite of the expansion in average land ownership, the changes in land tenure appear to

have had no impact on productivity. Johnson (2001) tested the hypothesis that farmers faced

asset-based credit rationing, meaning the amount of credit offered to individuals was constrained

by the lack of assets. This hypothesis suggested that farmers did not invest in productive assets

because of their inability to access the credit market. She found however no evidence to support

this hypothesis, implying that the lack of collateral and credit is not the cause of low-capital use

and low productivity observed in the agriculture. This finding is very important in the sense that

it shows that the reform of tenancy in Mexico by itself will not have a positive impact in

agricultural productivity.

Crop prices have been changing through the years. There is strong evidence suggesting

that after the trade reform, the Mexican domestic agricultural prices are indeed converging

toward international prices. OECD estimates reflect how the reduction in the nominal protection

of basic staples has proceeded over time. For example, maize protection decreased from 109% to

51% from 1993 to 1994, to 24.13% in 1995. In the case of yellow maize, the protection

16 National Rain fed Equivalents









decreased from 77%, to 28% and then 5%. Some other crops following the same declining

pattern are sorghum, soybean, wheat and barley. In the specific case of rice, the protection

estimate decreased and became negative around 1991 (Yunez-Naude and Barceinas, 2000).

There is also evidence suggesting a decrease in the production of rain-fed agriculture,

which is the realm of small and medium size producers. Between 1985 and 1990, for example,

the principal products of rain-fed agriculture decreased by 0.60 per cent per year and between

1990 and 1994, they fell by 4.35 percent per year. Maize production fell by 4.64 per cent

annually and beans by 2.63 per cent (Assies, 2008).

Little research has been done on the crop diversification topic. However, in general terms

it appears farmers are not undergoing crop diversification after guaranteed prices were

eliminated. Although the amount of land has increased, maize, bean and fodder crops remain the

staples of most producers in most regions of the country. Between 1994 and 1997 for instance,

Davis (2000) found that 75% of the surveyed ejido households planted only maize, while 19%

intercropped maize with other crops, leaving the growing of fruit and vegetables as well as

fodder unchanged. These results support the hypothesis that farmers are behaving as risk-averse

agents investing in low price, riskless production instead of undertaking the risk of producing

high value crops. On the other hand, it has also been argued that crop diversification has not

materialized because a considerable number of small farmers produce for own consumption

purposes, remaining indifferent to reductions in the relative price gap.

As mentioned above, the government has played an important role in providing technical

support to farmers through two of its programs, CONASUPO in the past and now ASERCA. The

government has focused on the diffusion of improved crop varieties, which has proven to be an

important source of agricultural productivity (Wood, You and Zhang, 2004). Through









Procampo, with a coverage rate equal to 80% of all ejidatarios, the government encourages the

use of high yield variety (HYV) seeds among basic grains producers.17 And with less success,

Alliance, which has reached only 12% of the ejidatarios, the government promotes investment in

capital intensity technology and the transformation of agriculture toward areas with comparative

advantage (Cord and Wodon, 2001).

Usage of technology, however, does not depend on governmental support alone.

Evidence suggests that the characteristics of the household as well as of the community, such as

farm size, community infrastructure and household member's education and income also play an

important role in determining the use of technology in agriculture. For instance, Davis (2000)

found that after the reforms, larger farmers made more use of HYV seeds, chemicals, technical

assistance and credit, while small and poor households were the less likely to invest in

technology use.

Wood, You and Zhang, (2004) found that over time, most of the agricultural R&D has

favored irrigated production systems, where the potential for technology spillovers is greater

than for heterogeneous rain fed areas. Many agricultural technologies are often location specific.

This means that a large part of the agricultural research is directed to overcome site specific

constraints in crop production such as, increasing plants tolerance to frost or drought, or

increasing plant resistance against a specific pest or disease. Homogeneous areas have more

potential for agricultural R&D spillovers, thus research is more abundant on irrigated production

systems.




1 Studies measuring the impact of Procampo on agricultural productivity have found that the program has increased
the agricultural income of the households keeping farmers growing their crops. However, the program has had little
impact on the productivity itself. Assies (2008) for instance, suggests the program has been insufficient to help the
rural sector make the transition to other commercial crops.









In general, the increased use of technology in the agricultural sector remains a long run

goal. The ejido reforms have not had a substantial effect on increasing capital intensification as

expected. Government continues supporting the diffusion of technology among farmers, but

because of its potential spillovers, the farmers taking advantage of this technology have mostly

been large, modernized farms. New programs oriented toward small farmers are still needed in

order to bring technology to less productive farmers and have a major impact in the rural sector.

Returning to farmers' risk-averse strategies, livestock accumulation seems to be an

important strategy for farmers because it keeps their savings relatively liquid but also protects the

household from macroeconomic shocks such as inflation or devaluation. In addition, livestock

and livestock derivatives consumed at home represent a fundamental part of the household diet.

Finally, animal by-products such as the sheep's wool also represent an important source of

income.

Evidence suggests a clear increase in livestock accumulation after the ejido reforms were

implemented. Davis (2000) found that on average the number of heads of cattle owned per

household increased by almost 20% from 1994 to 1997. About half of the households surveyed

had poultry, followed by pigs. Milk was produced by 25% of the households and eggs by 38%.

Off-farm activities have always represented an important source of income for many rural

households. Evidence from El Salvador, Mexico and Ecuador suggest that nonagricultural

employment generates 40 to 50% of a rural household's income (Araujo, 2004), representing

from 38% on the largest farms to 77% on the smallest (Araujo, de Janvry and Sadoulet, 2002).

Furthermore, about 60% of rural Mexican households have some family member working off-

farm (Davis, 2000).









Studies reveal that demographic characteristics of the household members such as

gender, age, ethnicity (de Janvry and Sadoulet, 2001) and secondary education (Araujo, de

Janvry and Sadoulet,.2002) play an important role stimulating off-farm activities. The location of

the community including the proximity of the community to an economic center, and the

availability of roads connecting the community also affect the propensity to be involved in off-

farm activities (Araujo, de Janvry and Sadoulet, 2002).

In Mexico, the main sources of off-farm income come from nonagricultural employment,

followed by other income, which includes governmental direct income transfers and welfare

programs, and remittances. The ejido reforms have had an important effect on this specific asset.

Evidence suggests that, after the reform, the diversification toward off-farm activities has

considerably increased. Davis (2000) found that the proportion of families participating in off-

farm activities rose by 33% between 1994 and 1997, encompassing up to 60% of the ejido

households. The success of Procampo has also increased the dependency of many farmers on

this source of off-farm income (direct income transfers).

The impact of off-farm activities on agricultural productivity has not been widely studied.

However, it seems agricultural production and off-farm activities are negative correlated. This

means the share of total household income derived from off-farm activities is inversely

correlated to a farm's size. The exception to this pattern is remittance, which is frequently found

among medium size farms (de Janvry and Sadoulet, 2001).

Furthermore, evidence suggests road availability as well as proximity to an economic

center influence the effect agricultural output has on off-farm activities. For instance, if there is

road nearby and the distance between the rural community and the economic center is not so

large, high value agricultural output would promote off-farm activities through service and









manufacturing employment growth. On the other hand, if the community is isolated, high value

agricultural output would shrink off-farm employment (Arajuo, de Janvry and Sadoulet, 2002).

One of the major strategies of rural households is international migration. Through the

years this phenomenon has been expanding in the rural sector. In his survey of ejido households,

Davis (2000) found that around 45% of the households had either a family member with

migratory experience to the U.S. or children or siblings living there. Moreover, around 50% of

the households with more than 5 NRE hectares reported some connection with the U.S.

Researchers have noticed that the formation of social networks over the years has promoted

migration through the reduction of the risk and transaction costs embedded in the migratory

experience.

Researchers studying the impact of the reforms anticipated an increase in out-migration

in the agricultural sector. Studies measuring the impact ofNAFTA, for example, predicted a

decrease in rural employment and wages, generating an emigration of as high as 800,000 people

from the rural sector, migrating mostly to the United States (Cornelius and Martin, 1993).

However, to date this prediction has not materialized. As a matter of fact, the agricultural sector

continues employing around 20% of the population (Taylor, Yunez-Naude and Dyer,. 1999;

Davis, 2000).

Lastly, an important consequence of international migration in the sending communities

is remittances, which represents an important source of income in rural areas. In 2003, for

instance, Mexican immigrants living in the US sent $14 billion in remittances to their relatives in

Mexico (Orozco and Lapointe, 2004). Different studies have been carried out in the impact of

remittances on rural Mexico, reaching no consensus on its impact. Chapter 3 discusses in detail

the findings of these studies.









Conclusion

This chapter presented an overview of the main changes in the Mexican agricultural

sector until the present time. As opposed to past policies, current policies attempt to set the

necessary conditions for a market oriented strategy. However, it seems trade liberalization and

the retreat of the State, have not made producers more responsive to market signals as expected.

In addition, the ejido reforms alone have not created enough incentives to increase the

productivity in the agricultural sector.

The way the agricultural sector will achieve competitiveness within the international

markets remains an enigma. Some believe that big private entrepreneurs will bring

competitiveness to the sector; others believe small farmers who are now land owners and with

government assistance will be able to gain competitiveness in the international market, bringing

new forms of self-employment and poverty alleviation (Quintana, Borquez and Aviles, 1998).

Research in this area is still limited.

As mentioned before, the impact off-farm activities on agricultural productivity has not

been widely studied. The next chapter will be devoted to migration, one of the off-farm activities

rural households rely upon heavily, and its linkage to the agricultural sector in rural Mexico.









CHAPTER 3
MIGRATION AND AGRICULTURAL PRODUCTIVITY

Introduction

Migration is being increasingly considered in the literature on agricultural productivity in

developing countries. This is because migration has become a common practice for many rural

households world-wide. Migration has been found to affect a household's decisions in three

important ways. First, migration reduces the labor availability of the household; second, it

generates an increase in the household's income through remittances sent by the migrant; and

third, it strengthens the formation of social networks, which can be used to promote the

migration of other household members. The aim of this chapter is to review the existing

literature on this topic.

In order to understand the impact of migration on the agricultural sector of the home

country, specifically the way in which the household's structure changes due to the migration of

one of its members and its impact on productivity, it is fundamental to study not only the

demographic characteristics of the migrant and the household, but also to understand the

composition of the migratory flows, the macroeconomic factors inducing the migration (Orrenius

and Zavodny, 2005;Comelius, 2001;Jones, 1995;Donato, 1999; 1994), and the inherent dynamics

of migration (Davis, Stecklov and Winters, 2001; Massey and Espinosa, 1997;Massey, Goldring

and Durand, 1994).

In the specific case of Mexico, migration has become a common practice in rural Mexico.

According to the CONAPO the number of Mexicans engaging in a migratory experience to the

U.S. reached 3.3 million between 1990 and 2000. Furthermore, remittances have become an

important part of the Mexican economy reaching the second place in source of foreign currency

after oil exports. The structure of this chapter is the following. The literature on Mexican









migration is reviewed first, followed by the literature on labor out-migration and agricultural

productivity.

Mexican Migration Literature

Immigration Reforms in the United States

The Mexican agricultural sector has always been very close linked to the agricultural

sector of the United States. Some factors explaining this relationship include the geographical

closeness between the countries, the similarities in climate, the bonds among relatives and

economic factors. In the specific case of the labor market, the immigration laws have also played

an important role in connecting the agricultural labor markets of the two countries. The major

U.S. immigration reforms during the twentieth century are described below:

The Bracero program

Migration of Mexican farmers to work in US fields has been a common practice since the

1940's. The first major Mexican migratory flow took place during the Bracero Accord, which

was implemented between 1942 and 1964 to face the shortages of agricultural labor in the United

States as consequence of World War II. The Bracero program allowed Mexicans to migrate

temporarily for agricultural employment in the United States encouraging seasonality in

migration flows, with cyclical movements across countries (Donato, 1994). This program

comprised approximately 4.5 million Mexican agricultural workers in total (Massey and

Espinosa, 1997).

During the same period, the U.S. Congress also passed the Immigration and Nationality

Act (INA) of 1952 promoting the allocation of visas to relatives of US citizens and bracero

workers believed not to have an adverse impact on the US labor market. Consequently, many

relatives of Mexican farmers enrolled in the Bracero program were able to apply and get visas.









After the Bracero program ended, there was a decline in the number of visas issued to

Mexicans. For instance, prior to 1965, there were no numerical limits to the legal entry of

Mexicans; in 1965 Mexico was placed under a hemispheric quota of 120,000, meaning Mexico

had to compete with other Latin American and Caribbean countries for visas. In 1976, it was

placed under a country quota of 20,000; in 1978 it was included under a global ceiling of

290,000; and in 1980 the global ceiling was reduced to 270,000 (Massey and Espinosa, 1997).

Immigration Reform and Control Act (IRCA)

Decrease in opportunities to enter the country legally, led to an increase in illegal

migration to the United Sates. Indeed, the percentage of migrants leaving Mexico illegally

increased from 37 percent during the Bracero program to 53 percent in 1965-68. Taking into

account this situation and in an attempt to reduce undocumented migration to the United States,

in 1986 the U.S. Congress passed the Immigration Reform and Control Act (IRCA).

This Act generated several measures to stop the illegal migratory flow. These included

increased border enforcement, employer sanctions against those who knowingly hired

undocumented migrants, a supplemental guest worker program, a modification of the H-2

program, and amnesty to migrants already resident in the United States (Donato, 1994; Iwai,

Emerson and Walters, 2006).

There were two groups of immigrants that were eligible for legalization under IRCA: the

first group was formed by those who had resided in the United States since before January 1,

1982; the second group were seasonal agricultural workers enrolled in the Special Agricultural

Worker (SAW) Program and employed for a minimum of 90 days in the year prior to May, 1986.

Three million Mexicans applied for legalization, and nearly 2.7 million were granted permanent

residence (Rytina, 2002; Orrenius and Zavodny, 2005). Of this total, approximately, 1.3 million

belonged to the second group (Iwai, Emerson and Walters, 2006).









Border enforcement

In spite of the endless efforts to stop illegal immigration, the number of illegal migrants

entering the United States continues to increase. In 1992, Donato (1994) found that 73 percent of

the migrants undergoing a first trip entered the United States without documents. Moreover,

studies suggest that the Mexican illegal population in the United States has grown from 1.1

million in 1980 to 2 million in 1990 and 4.8 million in 2000, with an average annual growth of

90,000 in the 1980s and 280,000 in the 1990s. From the total population of unauthorized

residents in the United States, Mexicans account for 69 percent of the undocumented residents

(Angelucci, 2005).

In the search to stop illegal entry, the government has turned to border enforcement to

decrease illegal immigration, especially since the 1986 IRCA. In the last two decades, the U.S.

government has raised the enforcement budget of the U.S. Border Patrol from $290 million in

1980 to $1.7 billion in 1998 and more than $2 billion in 2006. Two additional pieces of

immigration legislation passed after IRCA related to border enforcement were the Immigration

Act (IA) of 1990 and the Illegal Immigration and Responsibility Act (IIRA) of 1996 (Carrion-

Flores, 2007). The current debate regarding building a wall on the southern border and the huge

expenses inherent in this project, questions the effectiveness of this measure to stop people from

entering the country illegally.

The Evolution of Mexican Migration

Migratory patterns

Researchers working on Mexican migration have observed three migratory patterns:

permanent migration, characterized by a long interval migration (more than five years), where

migrants normally achieve legal status; temporary migration, characterized by shorter trips done

mostly illegally; and return migration, characterized by the return of the migrant to the home









country for either a period of time or forever. Temporary migration if done continuously is

known as cyclical or repeating migration and is also associated with illegal migration.

In the literature, permanent migrants are commonly more skilled, with better jobs and

opportunities than temporary migrants. When an individual first migrates to the foreign country,

he usually has little or no skills valuable in the foreign labor market; only after some years when

the migrant has learned certain skills such as the foreign language and has acquired some

experience, can he aspire to a better job and look for a legal status (Borj as, 1984).

When the phenomenon of migration occurs both the receiving and the sending country

undergo a change. Most of the research studying the impact of migration on the United States

has focused on analyzing legal migrants, based on permanent migration. This is due to data

constraints: data availability still plays an important role in defining the unit of study and most

data on illegal immigration is limited. In the case of Mexico, data availability is a constraint

because most of the databases on migration are not nationally representative. Also they only

keep track of temporary migrants who return to the country of origin at the time of the survey. So

much of the analysis on migration in Mexico focuses on temporary migration.

Cyclical migration has been the migratory pattern dominating the Mexican migration

literature. Since the implementation of the Bracero program and until the 1990's, studies have

found that Mexican migrants migrate to the United States repeatedly. This cyclical pattern

becomes evident in a study carried out by Angelucci (2005), where Mexican migratory inflows

and out flows were calculated. In the study it was found that the migratory inflow of Mexicans

between 1972 and 1993 rose to 1,265,000 people, while the outflow was around 95% of the

annual inflows (Angelucci, 2005).









Many researchers view Mexican migration to the United States as a self-perpetuating

process. As the amount of prior U.S. experience grows, and the number of trips to the U.S.

increases, so does the likelihood of repeat migration. Apparently, the nature of cyclical migration

is associated with the changes in the family life cycle: increasing for young, unmarried men,

falling with marriage, and then increasing again as children grow and the household's

consumption needs rise (Massey and Espinosa, 1997).

Kinship ties also play an important role in the migratory decision. Once the migrant has

achieved the reunification of the family in the host country, the probability of returning to

Mexico is reduced significantly. Better-educated migrants are more likely to shorten their trips

compared with less-educated migrants. The migrant's geographic location of origin also affects

the duration of the trip. It seems distance is positively correlated with the duration of the trip.

Furthermore, migrants coming from rural areas also spend more time in the United States

compared to those coming from urban areas (Carrion-Flores, 2007).

Finally, the cyclical migratory pattern of Mexican migrants has begun to change in the

past two decades, as border control has become more rigorous. Apparently, Mexican migrants

are very sensitive to changes in border enforcement because they perceive it as an increase in the

cost of migration, reducing migratory inflows to the United States (Hanson and Spilimbergo,

1999; Orrenius, 1999). At the same time, it discourages recurrent returns to Mexico, and

consequently lengthens the time spent in the United States. For instance, a one unit increase in

border controls has proved to decrease the individual likelihood of returning to Mexico by 31.8

percent. This means that if normally 46% of the illegal residents in the United States return to

Mexico each year, an increase in one unit of border control reduce the number of returns by 31%

percent (Angelucci, 2005).









Migratory flows

In conjunction with the migratory pattern, the composition of the migratory flows has

also experienced significant changes during the twentieth century. As mentioned above,

economic conditions, social ties, and political issues such as border enforcement play a

determining role in inducing or deterring the migration of Mexicans, primarily undocumented,

into the United States.

Economic conditions in both Mexico and the United States have proven to influence the

individual decision to migrate to the United States. For instance, an increase in the U.S. expected

wage is commonly associated with an increase in the length of the trip (Carrion-Flores, 2007).

Furthermore, a 10% decrease in the real Mexican manufacturing wages is associated with at least

a 6% increase in attempted illegal border crossings (Hanson and Spilimbergo, 1999). And, older

migrants are more responsive to increases in U.S. farm wages, while nonagricultural wages and

the minimum wage in the United States have greater influence on sons' migration decision

(Orrenius and Zavodny, 2005).

When measuring self-selectivity among undocumented immigrants from Mexico,

research suggests that higher average U.S. wages and higher minimum wages are associated with

more and less-skilled immigration that lead to a negative selection process. Improved conditions

in the Mexican economy lead to less migration but also to relatively lower education levels

among those who do migrate. In general, skilled workers seem to be more responsive to changes

in the Mexican economy and the unskilled, more responsive to changes in the economy of the

U.S. It seems skilled workers in Mexico are more tied to the Mexican economy through physical

or human capital, making it more difficult to react to temporary changes. In addition, skilled

workers, unlike unskilled workers, are able to use their savings as a measure to smooth

consumption for a longer period of time (Orrenius and Zavodny, 2005).









Demographics of the migratory flows have also experienced some changes throughout

the years. In a study carried out on first-time migrants' occupational decision, a shift toward non-

agricultural jobs was found in the recent years. During the Bracero program around 76% of the

migrants worked in agriculture on their first U.S. trip. After 1964 however, this percentage

dropped by 30% percent. Since then, many migrants have shifted toward unskilled jobs such as

manufacturing, service and construction. Furthermore the number of migrants employed in

skilled jobs increased from 3 percent to 14 percent of total migrants between 1942 and 1992

(Donato, 1994).

Immigration reforms in the United States have been an important factor defining the

demographics of the migratory flows. During the 1942-1964 period migration was comprised

primarily of men over 15 years of age, with almost half of them being bracero workers. After

1964, when many bracero workers achieved legal status and were able to sponsor their families,

the composition of cohorts changed. Women and children were increasingly likely to leave on a

first trip.

As more restrictions were implemented and the likelihood of entering the United States

legally reduced, the flow of women continued to increase while the flow of children was reduced

dramatically. For example, the percentage of migrants less than 15 years old dropped from 20

percent in 1977-1981 to 14 percent in 1987-92. On the contrary, women migrating to the United

States increased from 28 to 32 percent between 1969-76 and 1977-81 respectively.

In addition, studies suggest border enforcement is not only affecting the age and gender

of the composition of the migratory flows, but also their level of education. The idea behind is

that border enforcement represents an increase in the cost of migration, making it more difficult

for unskilled workers to raise that money and consequently limiting the migration to only those









who can. Higher-skilled workers are more likely to migrate, increasing positive selection among

illegal immigrants (Orrenius and Zavodny, 2005).

On the other hand, it has been found social ties affect migratory flows by reducing of

migration costs. Using social network as a proxy for migration cost and dividing the sample into

communities with low and high-migration costs, a study found that about 38% of households

head living in low-cost communities have ever migrated to the United States; while 30% have

done so in average-cost communities; and only 17% in high cost communities (Orrenius and

Zavodny, 2005).

Causes of Migration

There is no consensus in the literature on migration about what causes individuals to

migrate. Different approaches have been developed and employed in different contexts. The

most widely used until recently, however, is an economic decision-making framework in which

the individual migration decision is based on comparing the expected net present value of

income in the destination country and in the place of origin. Todaro (1980) formalizes this

framework and predicts that migration occurs only when the expected net present value of the

earnings (net of transportation cost), weighted by the probability of employment at the

destination country is positive (Chiswick and Hatton, 2003; Moretti, 1999).

Other approaches have been developed as alternatives to understand the causes of

migration. Most of these models re-introduce the importance of the social context as an

explanatory tool of migration. For instance, a sociological approach of migration relies on

components such as cultural18 and social capital to understand migration decisions (Castles,



18 Cultural capital is defined as the knowledge acquired about other societies and the work opportunities they offer,
including information about the labor market and the living conditions. Social capital is more commonly used
specially in terms of social networks and refers to the connections established among relatives, friends or people in a
community to reduce the transaction costs and risks of migrating (Castles,2002).









2002). Complementary reasons inducing migration include a demographic approach (Massey,

Goldring and Durand, 1994). However, the model that has gained acceptance and become an

important conceptual framework for migration in the recent years is the New Economics of

Labor Migration approach (NELM) (Stark and Bloom, 1985).

This approach studies the prospective migrant as a social agent involved in a family's and

community decision-making. The migration decision is linked to the family's strategy to manage

uncertainty, diversify the income portfolio and alleviate liquidity constraints through remittances

(Castles, 2002; Stark, 1991). In consequence, the model suggests migrants, although separated

physically, maintain relationships with their families during the migratory process.

In the specific case of Mexico, Mexican migration to the United States has proven to be

determined by factors other than just the economic condition of the two countries. Furthermore,

as noted previously, a common practice of rural households is the diversification of their sources

of income, which suggests households entail a risk-averse complementary income generation

strategy to confront incomplete or non-existent markets (Davis, Stecklov and Winters, 2001).

For this reason, this research will use the NELM approach as a framework to model Mexican

migration.

Impact of Migration in the Sending Communities

Sociologists have extensively studied the relationship between migration and community

development, paying special attention to the links migrant establish with people and

communities located in nations other than those to which they migrate (Vertovec, 2004). From

this point of view, physical barriers as well as the physical location of the migrant lack of

importance and instead efforts are made to quantify the participation of immigrants in the

economic, political and cultural life of their country of origin through the constant flow of ideas,









money, and information. (Portes, Escobar and Radford, 2007). This stateless way of studying

migration is known as transnationalism in the sociological literature.

In his work, Castles reinforces the idea of transnationalism by reformulating the role of

the immigrant under the new conditions of globalization. He argues that a global world would

also affect the way migrants are conceived. Migrants will be each time more diverse in social

and cultural attributes, and the types of migration will not be limited to the three types mentioned

in the last section. New types of migration will include the repeated, circulatory and retirement

migration. Also the worldwide use of internet and the improvements in transportation are

expected to strengthen informal networks improving communication and organization among its

members. According to the transnationalism literature remittances and hometown associations

represent two important aspects of migrant transnationalism.

Remittances

Studies on the effects of migration in the sending country began receiving special

attention, as the amount of remittances sent by the migrants to their families and home

communities increased and became significant in volume. For instance, in 2003, Mexican

immigrants living in the US sent $14 billion dollars in remittances to their relatives in Mexico.

Also, the estimated amount of annual remittances in the world-wide is over $100 billion dollars

(Orozco and Lapointe, 2004).

Given this inflow, that promises to increase in the coming years, researchers are

motivated to study the use and impact the remittances have on the household's economic

activities, taking into account that in most cases, migrants came from rural areas. Until now,

however, the research on remittances has led to contradictory results and no consensus has been

reached on the impact of remittances in the sending country.









Positive evidence on development suggests remittances are directly invested in the

development of small businesses such as manufacturing and craft companies, as well as in the

purchase of productive inputs such as land, seeds, fertilizers and livestock. On the other hand,

the negative findings indicate that remittances are not invested in productive activities but on the

contrary, are spent on consumption goods such as food, cars, radio and television. Also, these are

found to create big inequalities among community members and create a "culture of economic

dependency" (Vertovec, 2004). Furthermore, there is the concern that remittances reduce the

supply of labor by recipients in the labor market, affecting the economic activity adversely

(Chami, Fullenkamp and Jahjah, 2005).

In another study, it has been found that the productive use of remittances is positively

associated with education. In general terms, better educated migrants are more likely to have

their recipient families invest their remittances in housing or productive capital instead of

spending it on consumption or nondurable goods (Durant et. al., 1996). When introducing

remittances into a family context model, where the relationship between migrant and family is

characterized by altruism, it has been found remittances serve as compensatory transfers to help

families overcome financial constraints created by poor economic performances (Chami,

Fullenkamp and Jahjah, 2005).

Finally, in research using a disaggregated, rural economy wide modeling (DREM)

approach, an increase in migrant's remittances by 10% was simulated. This increase in direct

transfers translated into a rise in international migration, which in turn drove up the cost of

agricultural labor by 1% negatively affecting cash crop and commercial maize production by

between 0.5% and 2%. However, the income of household groups accessing remittances

increased between 2% and 5%. An interesting fact is that remittances stimulate subsistence









household's consumption demand for maize, driving up the shadow price of maize and

stimulating subsistence production (Taylor, Dyer and Yunez-Naude, 2005).

Hometown associations

There is a long history of migrants collecting money and sending it to home communities

for collective benefits. However, it was not until the 1990s that the study of these associations

increased. A hometown association can be defined as an organization of immigrants from the

same town in a host country who meet for social and multi-aid purposes (Caglar, 2006).

Activities performed by these associations vary greatly. Some are involved in charitable

work, such as enhancement of the church or the graveyard, while others focus on infrastructure

improvement, such as building sewage treatment plants, providing electricity, paving roads, and

improving health care and school facilities. They can also serve as means of fundraising when

natural disasters occur, or for the celebration of the town patron.

The characteristics of members are as diverse as the activities performed by these

associations. In a study of associations from three Latin American-origin immigrant groups in

the East Cost of the United States, Portes, Escobar and Radford, (2007) found that the personal

characteristics of the immigrants play a determining role influencing the activities undertaken by

an organization. Some of these characteristics include the education, age and legal status of the

immigrants, as well as their duration in the host country and their origins (rural or urban) For

instance, migrants coming from rural areas tend to create associations not linked to politics while

immigrants from urban origins tend to become more involved in the politics of their countries.

Yet, it is difficult to generalize from these results.

In combination with the immigrant's background, the policies developed by the sending

government can also determine the investment decisions of the associations. Some of the

schemes and financial incentives that have been used to channel the hometown associations









(HTAs) investments consist of reduction in tariffs on the importation of machinery and

equipment, preferential access to capital goods as well as joint-investments between local

government and migrant organizations (Caglar, 2006).

The Mexican Government has played a leading role directing the course of the HTAs

activities in the country. Its major effort culminated in the creation of the Institute of Mexican

Abroad (IME) to promote the participation of these associations (Portes, Escobar and Radford,

2007). Another initiative of the government was The Citizen Initiative Program 2x1 created in

the 1990s, in which for each dollar raised by the hometown associations, the federal and state

government each contributed a dollar to a community project. In 2002, the program was changed

into 3x1 to incorporate the municipal governments into this program (Orozco and Lapointe,

2004; Vertovec, 2004).

In summary, the Mexican migration literature suggests that the demographics of the

individual, the creation of social networks, and the economic and political condition of both

countries affect the way in which the migratory process evolves. The literature also highlights

the enormous heterogeneity that exists among the migrants working in the United States. These

factors need to be taken into account when modeling. Now we turn our attention to the

agricultural productivity literature to study the way in which the decision of migration affects the

farming practices of the migrant households compared to non-migrant households.

Labor Out-Migration and the Agricultural Productivity Literature

There is an extensive body of literature that relies on agricultural productivity to measure

the efficiency of the allocation of resources. Agricultural productivity is commonly defined as

the ratio between total output and total input measured in a given time period (Christensen,

1975). There is no consensus on the way agricultural productivity should be measured.









Researchers use partial or total productivity, gross or net productivity depending on the aim of

the research and the availability of data (Dovring, 1979).

Land productivity is a topic that has focused the attention of many researchers. The

impact of tenure security on agricultural productivity and the findings on farm size-productivity

relationship are summarized by (Kimhi, 2003; Johnson, 2001; Hayes, Roth and Zepeda,1997).

Land management decisions such as crop choice, planting dates, fertilizer use rates and capital

use have also proven to impact crop yields. 19

In the specific case of labor productivity, there is an extensive body of literature studying

the impact of migration in the host country (Napasintuwong and Emerson, (2005); Iwai,

Napasintuwong, and Emerson, (2005); Hashida and Perloff, (1996)) but the study of labor out-

migration and its impact on the agricultural productivity in the sending community is a small but

increasing research area that has attracted the interest of researchers in the last decade. The idea

of introducing migration into the agricultural productivity analysis of the local of origin is that

national as well as international migration reduces the labor supply in the community, affecting

the farming practices decisions of the household, which in turn can lead to a change in

productivity.

This topic has recently gained the attention of agricultural economists due to several

reasons: first, the growth in volume of migration; second, most migrants come from rural areas

where agriculture remains one of the primary economic activities; and third, migration, through

the flow of remittances, is expected to alleviate liquidity constraints, caused by credit and other

markets imperfections in the rural economy, enabling farmers to invest more heavily in

enhancing productive assets.


19 For a list of papers exploring the relationship between crop management and productivity see Pender et.al. (2003),
Jansen et.al. (2006) and Tittonell et.al. (2007).









The literature on migration and agricultural productivity in the home country can be

divided into two main areas. The first basically focuses on the study of migration and the way it

affects the farming practices of the household through changes in input use, the second is

relatively new and introduces the concept of gender into the labor out-migration analysis. We

summarize this literature in below.

Changes in Farming Practices and Decisions

No consensus has been reached on the way migration influences the farming practices

and decisions of the household. Migration is used in different circumstances either as a strategy

to enhance the productive use of inputs or as a mechanism for moving out of agriculture. Most of

the empirical work; however, coincides in finding differences in the farming practices of

migrants compared with non-migrants households.

How migration is incorporated into the model also varies greatly from one study to

another. Some studies use a Total Factor Productivity approach (TFP) in their analysis to

estimate a production function and evaluate the effects of migration on the crop output

(Nonthakot and Villano, (2008); Ortega-Sanchez and Findeis, (2001)). Other studies adopt a

partial productivity methodology and focus their analysis on the effect of migration on a specific

farming practice (Mendola, 2008; Miluka, et.al. 2007).

Using a cross-sectional household survey, Mendola (2008) tests whether migration

stimulates the use of high-yielding seed technology in rural Bangladesh where labor migration

has been an enduring phenomenon. The empirical analysis of this paper, based on the NELM

insights, addresses the fact that farm households typically face income uncertainty, and measures

the effect of migration on risk-taking behavior in agricultural production. An important

contribution of the paper is the differentiation the author makes between temporary-domestic,









permanent-domestic, and international migration to account for heterogeneous household

migration strategies. .

The author finds that households engaging in international migration, which normally

have higher initial asset holdings to support migration expenses, are more likely to employ

modern farming technology, such as HYVs of rice, thereby achieving higher productivity after

migration of a household member. Asset-poor farm households are more likely to engage on

domestic migration, and rely more on conservative strategies, which do not drive production

increases.

Mendola (2008) argues that the success of migratory practices, as an income

diversification strategy and as promoter in risk-taking behavior among farmers, depends heavily

on the initial asset holdings of the household. This means that if rural policies are not

implemented to help farmers overcome the uncertainties linked to agriculture, asset-poor farm

households, unable to pay the costs of international migration, will be kept marginalized and in a

persistent poverty trap.

Miluka, et.al, (2007) study the case of Albania, where 54 percent of the population

resides in rural areas and agriculture still employs around 50% of the workforce. Following the

New Economics of Labor Migration (NELM) approach, the authors analyze the allocation of

labor and capital resources in the household as a consequence of migration. Their objective is to

measure the effect on agriculture of changes in labor supply availability and the gain in access to

working capital or credit, due to the inflow of remittances.

They first study the impact of migration on the family labor hours spent in agriculture,

finding that members of households with migrants abroad work significantly fewer hours in









agricultural production. However, women in migrant households work proportionately more

than men, when compared with their counterparts in non-migrant households.

Then they measure the impact of migration on non-labor input expenses in agriculture to

measure the effect of migration on the investment in productivity-enhancing assets. They

conclude that migrant households do not seem to invest in more productive techniques such as

chemicals, fertilizers and machinery. Instead, migrant households are shifting toward livestock

production. This result is intuitive since the shift toward less labor-intensive activities such as

livestock production can be explained by the fact that male activities are being replaced by

female activities within the migrant household.

In the case of Mexico, Ortega-Sanchez and Findeis (2001) estimate the labor out-

migration impact on corn farmers of the central and southern regions of the country. However, as

opposed to the studies of Albania and Bangladesh, they study labor out-migration without

differentiating between internal and international migration. Instead they partition their sample

according to migration status (migrant households vs. non-migrant households), agricultural

environment (traditional, semi-modem and modern farming practices) and household typology

(classification of households by asset endowment or access to it) using discriminant analysis.

The idea is that differentiated access of land, agricultural machinery and family labor affects the

way labor out-migration impacts the production system.

In opposition to the previous studies that focus on the NELM approach, Ortega-Sanchez

and Findeis (2001) rely less on borrowing and liquidity constraints and focuses their research on

analyzing rural labor markets imperfections. Specifically, they note family labor faces two major

imperfections: the potential moral hazard problem linked to in-hired labor doing tasks that

require intensive effort; and the difficulty of replacing farmers with knowledge and









organizational leadership for traditional farming practices. Overcoming these market

imperfection problems requires costly supervision or adjustments to farm production. The

question that remains unanswered is the ability of migrant households to overcome these

imperfections.

They hypothesize that outmigration reduces the productivity of farm resources and causes

a malfunctioning of the agrarian production system. Estimating a sequential production function

Ortega-Sanchez and Findeis find that labor outmigration does have an impact on the household's

productivity. For instance, they find higher output and productivity levels in non-migrant

households that use traditional or semi-modern techniques compared to migrant households

using the same techniques. The same relationship is found for households with low-asset

endowments. This relationship, however, does not hold for migrant households using modern

techniques or with high-asset endowment. These findings support the idea that the impact labor

out-migration has on agricultural productivity depends not only on the household's initial

endowments but also on how labor-intensive the farming practices of the households and the

household's ability to substitute family labor with reciprocal or in-hire labor.

To analyze the substitution capacity of the households, Ortega-Sanchez and Findeis

(2001) calculate the partial elasticity of substitution, differentiating by household migration

status. They find similar elasticity ratios across inputs and tasks for households in both groups.

This finding suggests that the difference in output and productivity might be due to some sort of

inefficiencies in the migrant households. Such inefficiencies, however, are not observed in

migrant households using modern techniques or with high-asset endowment.

Another recent study carried out in Mexico by Taylor and Lopez-Feldman (2007),

focuses on the ways in which labor out-migration influences incomes and productivity of land









and human capital. Using the ENHRUM20 survey, they estimate a switching regression model

with cross-section income for 2002 and retrospective data on international migration dating for

1990. Their findings suggest that migration to the United States increases the per capital income

of households.

They find that households with a migrant in 1990 had higher marginal returns to land in

2002 than households that did not participate in migration. This finding suggests that a lapse of

time is required before the effects of migration on productivity can be observed. In the case of

human capital, they find that an additional year of farmer schooling has a significant and positive

effect on total income in households without a U.S. migrant and no effect in households with

migrants. This findings suggest that local wages is important in the migration decision of the

household.

Nonthakot and Villano (2008) pick up on the unresolved question regarding whether

remittance incomes enhance production enough to compensate for the reduced availability of

labor (Mochebelele and Winter-Nelson, 2000; Rozelle, Taylor and DeBrauw, 1999). They study

rural-urban migration in the northern part of Thailand, where seasonal migration has been a

common practice especially during the dry seasons.

They estimate a stochastic production function to evaluate the effects of migration on the

mean maize output. Their findings suggest that indeed labor shortages, caused by migration,

negatively affect maize production. Nevertheless, they also find that remittances as well as the

period of migration have a positive effect on decreasing technical inefficiencies. This means that

the longer the duration of migration, the greater the chances are that the migrant will send an

important amount of remittances to the household, allowing it to invest more income to improve

production efficiency.

20 The acronym stands for Mexican National Rural Household Survey .









They also find that age and education have a negative association with technical

inefficiency, inferring importance of experience and knowledge in improving farm management

practices. Finally, the migrant farms computed an average technical efficiency of 86 percent,

which is ten percent more than the coefficient registered for non-migrant farms. These findings

support the idea that migration can help relax liquidity and credit constraints, allowing

households to buy productivity-enhancing inputs such as chemicals and fertilizers and to hire

pre- and post-harvest labor for their farming operations in a timely manner.

Gender Productivity

Another important consequence of outmigration, in addition to shortages in labor and

remittances, has been the shift from male to female labor in agriculture. This shift in roles within

the migrant household has introduced the study of gender productivity into the labor out-

migration literature. In the gender productivity literature, studies have been done to analyze the

impact of gender discrimination on the allocation of agricultural inputs (Deere and Leon, 2003;

Doss and Morris, 2001); the effect on production of the intra-household allocation of resources

among the household's members (Udry, 1996); and the gender differences in production between

male- and female- headed households (Masterson, 2005; Holden, Shiferaw and Pender, 2001;

Jacoby, 1992; Lastarria-Cornhiel, 1988).

In the labor out-migration and agricultural productivity literature, the way in which the

gender effect is modeled depends on the availability of data on the ownership of land as

discussed in Quisumbing (1996). In the database used in our study 14% are female-headed

households; however, only 5.4% percent of these households work the land and less than 1% has

undergone a migratory experience, making it difficult to carry out inter-household comparisons.

In this case, gender effects can only be measured in terms of the intra-household allocation of

resources.









Most of the empirical work done on the intra-household gender effect of labor out-

migration has been carried out in Africa. According to Mabogunje (1989), for example, the

outmigration of farmers in the Sub-Saharan African agrarian economies has led to the

reorganization of the traditional labor supply institutions and the changing role and status of

women. It has been observed that agricultural production has had to adjust toward tasks that are

less labor intensive, and women have started assuming an active role in the decision-making of

the household.

Additional studies from South Africa coincide in the belief that farms without male labor

are at disadvantage compared to other households. Farm households in South Africa with the

male migrating to another place, for example, experience lower productivity per acre and per

worker due to the shortages in labor the migrant household face (Masterson, 2005). Mochebelele

and Winter-Nelson (2000) analyze the effects of gender on farms estimating the technical

efficiency coefficients for each group, taking into account migratory status. With an average

technical inefficiency of 0.24 for female and male managers in the migrant sample and of 0.37

and 0.35 for female and male managers in the non-migrant sample respectively, the author

concludes that within each migratory group, the gender-based estimates are not significantly

different from the sample estimates, suggesting no gender bias in technical inefficiency. The

findings that both male and female farm managers benefit from having a household member

away shows that the benefits of migration, through remittances, is not gender biased.

In the case of Latin America, and specifically Mexico, female labor is more oriented

toward household domestic activities making it much harder to analyze the efficiency of the

allocation of resources. Furthermore, the division of labor in agriculture tends to be

complementary, female and male. In addition, as mentioned before, the availability of data on









female-headed households undergoing a migratory experience is scarce. For these reasons, the

literature in this topic is still very limited.

Conclusion

This chapter presents an overview of the evolution of the Mexican migration during the

twentieth century and summarizes the exiting literature on labor out-migration and its impact on

agricultural productivity. In general, no consensus has been reached on the way migration

influences the farming practices and decisions of the household. However, it has been noticed

that the household's initial endowments as well as the type of migration entail different effects of

labor out-migration in the sending community. Furthermore, it has been found that, in general,

remittances are being used to relax credit constraints and improve the farm management

practices of the household. On the other hand, the study of labor out-migration on gender is still

limited.

In the case of Mexico, the analysis of the way in which migration affects productivity in

rural households is still limited. Existing studies suggest important differences between migrant

and non migrant households. Furthermore it has been found that the effect of labor out-migration

on agricultural productivity depends not only on the household's initial endowments but also on

how labor-intensive the farming practices of the households are, and the household's ability to

substitute the family labor with reciprocal or in-hire labor (i.e. the way the rural labor market

works).

Chapter 4 describes the database and explains the methodology that will be used in this

study to analyze labor out-migration in Mexico and its potential impact on agricultural labor

productivity.









CHAPTER 4
DATA ANALYSIS AND METHODOLOGY

Introduction

Building on the theoretical foundations and substantial empirical research highlighted in

the previous chapters, my hypothesis is that labor productivity in migrant households is greater

compared to the labor productivity of non migrant households. Adopting a NELM theoretical

framework and reasoning (Castles, 2002; Stark, 1991), my primary hypothesis is that, as a

household strategy to manage uncertainty and market imperfections, migrant households

maintain their agricultural production level by investing more in capital-intensive inputs to

compensate for the reduced labor force availability due to the migration of at least one of the

household members.

That is to say, among migrant families, the availability of labor measured as the total

number of days worked in agriculture is expected to fall as members in the household migrate.

My corollary hypothesis is that labor productivity, measured as the agricultural output generated

per day of work will be greater in migrant households compared to non migrant households. To

test these hypotheses, my study employs econometric techniques using the Mexican National

Rural Household Survey (ENHRUM).

The contribution of this study to the existing literature focuses on three main points.

First, we rely on the New Economics of Labor Migration (NELM) approach, using the household

as the unit of analysis to study the way labor out-migration influences the labor productivity of

rural households. Second, we estimate labor productivity accounting for the selectivity of

landholding. The idea behind this is that agricultural productivity can only be measured for those

households holding land, and until now no study recognized this selectivity when studying labor









productivity. Finally, we introduce into the labor productivity analysis the study of social

networks.

This chapter has the following structure. The first section describes in detail the survey

and the descriptive statistics of the sample. The second section describes the Heckman two-stage

procedure. The third section summarizes the variables introduced in the model. The fourth

section tackles the possibility of having endogenous variables in the model. The fifth section

presents conclusions.

Data and Descriptive Statistics

The Mexican National Rural Household Survey (ENHRUM)

The Mexican National Rural Household Survey (ENHRUM)21 is a survey conducted

among rural communities in Mexico,22 that is part of a project co-directed by the Colegio de

Mexico (Colmex) and the University of California at Davis.23 The goal of this project was to

obtain a representative survey of Mexican rural society and economy; this sample would enable

researchers to study the way in which the agricultural and trade reforms have impacted the

production, income and migration of rural households in Mexico.

It is a cross sectional survey which includes 8,520 individuals from 1,765 households in

14 states. According to Mexico's National Information and Census Office (INEGI), who

designed the sample the survey represents more than 80 percent of the rural population in

Mexico. It was conducted between January and March 2003 and collected detailed socio-

demographic and economic characteristics of the households as well as their labor and migratory






21 EHNRUM stands for Encuesta Nacional a Hogares Rurales de MIxcio.
22 The communities included in the sample contain a population between 500 and 2499 people.
23 For more information visit http://precesam.colmex.mx/ENHRUM/PAG%20PRIN ENHRUM .htm









experience24. It also captures the farming practices of the household, sources of income and

credit history, among other variables. In addition, it provides information on use of family labor

and consumption. The information captured in the household survey has been classified into 12

chapters: housing, household members, plot, crop, livestock, natural resources, other

expenditures and incomes, assets, credit and inheritance, household corner store (tienda) and

fishing.

In addition to the household survey, ENHRUM collected information on the surveyed

communities, such as major economic activities, possession land, land characteristics, overall

farming practices, use and access to natural resources, migratory patterns and governmental

programs, among others characteristics during the months of August and October of 2002. The

goal was to provide a generalized picture of the economic, social and political situation of the

surveyed communities. The communities were also grouped into five different regions defined

by the National Development Plan: Northeast, Northwest, Midwest, Central and South-

Southeast25.

Some drawbacks of the survey are the fact that it is cross sectional data, it does not

provide information on return migration or duration of trips, and some of the farming practices

were aggregated at the household level instead of the desired parcel/plot level. The first one will

not allow us to make inferences across time, such as inferences about whether or not migration

has made Mexican Agriculture more or less productive. The migratory history provided by

ENHRUM is not as rich when compared to the MMP. Finally, the lack of data at the parcel level

does not permit us to make any analysis at the parcel/plot or crop level.



24 A similar survey is The Mexican Migration Project (MMP), which has tracked information of migratory
experience of the head of the households since 1982. However, the MMP does not provide information about
agricultural practices.
25 For a list of the community codes refer to Table 4-1.









Sample Description

Households represent the unit of analysis in this study. According to the NELM

literature, migration becomes an intra-household strategy to overcome liquidity and other market

imperfection constraints (Castles, 2002; Stark, 1991). Hence, in order to analyze the impact of

labor out-migration on agricultural productivity the study needs to rely upon the household to

determine how households reallocates the remaining labor and capital resources once the

member of the household migrates.

The number of observations in the survey consists of 1765 households (n=1,765). An

important feature of this database, however, is that the number of observations in each chapter

varies greatly across households. In order to calculate the labor productivity of the household, for

example, information on production and labor employed are required. From those households

reporting information on plot characteristics (n=871), only 762 (n=762) reported their annual

production in the survey and 707 reported the family's labor during the crop cycle (n=707).

For the purpose of this study, those households that reported information on both

production and labor (n=707) were the only households taken into account.26 This group of

households is labeled sub-sample B and is the one used to measure the labor productivity at the

household level. The labor productivity of the head is also measured separately because the

household head constitutes an important asset in the family's labor force. The number of

households reporting information on the household's head labor productivity equals 667 and

represent the sub-sample A (n=667).

In order to avoid the non randomness nature of sub-sample A and B, we first estimate the

probability of households' having access to land and from those that have land, we then estimate


26 The remaining 164 households reporting information on plot characteristics but not specifically on output and
labor force are excluded.









the labor productivity of the household. We defined a household as a land holder if during the

survey the household reported information on plot characteristics otherwise it is considered a non

landholder. This is a strong assumption, but unfortunately no direct question about possession of

land was formulated in the survey. Something to keep in mind is that land tenure is not taken into

account under this definition. This means that the plot could be owned, rented or leased and the

household would still be considered a landholder.

Sample A consists of 1561 households, of which 894 are non landholders and 667 are

landholders. Sample B consists of 1601 households, of which 894 households are non

landholders and the remaining 707 are landholders. In sample A we are measuring the labor

productivity of the household head assuming the family's labor is a function of the labor force of

the head alone. In sample B we are measuring the labor productivity of the household. We are

assuming the family's labor is a function of the head, the son/daughter, the wife, the

grandchildren and the son/daughter-in-law. Once we described the two samples, we devote what

is left of this section to describe the demographic characteristics of our samples. Because sample

A forms part of sample B, and in an attempt to avoid duplications, we will focus on the

descriptive statistics of sample B in this chapter.

Of the 1601 households conforming sample B, 86.7% are male-headed and the remaining

13.3% are female-headed households. The average age of the household head is approximately

48 years27; while the average age of the household spouse is 41 years. 84% of the households are

married or live together. The level of education varies greatly across households overall, the

level of education can be considered as low. As shown in Figure 4-2., in approximately 17% of

the households, the head of the household has no education; in 21%, he/she finished elementary

school; in 40% he/she has some elementary school; only 9% finished middle school; and 2%

27 Figure 4-1 presents a breakdown of the household head age.









finished high school. In the case of the household spouse, in 14% of the households the spouse

has no education; in 61% he/she has elementary school; in 14.9%, middle school education; only

3.2% finished or not high school.

Despite the low education rate, the average experience of the household head (in any

sector, not exclusively agriculture) is approximately thirty three years. Furthermore, 24.2% of the

household heads were employed outside the agricultural sector during their first job. The average

number of household members is approximately five,28 with 29.1% of the households having at

least one child between zero and six years old. Finally, 17% of the households in the sample

speak an indigenous language at home. The following sections compare sub-samples according

to land holders and migratory experience.

Landholder and non landholder mean differences

For the 1601 households in sample B, we have plot level attributes for only 707 of the

households, what we have defined as landholders (n=707), the remaining households are

considered non landholders (n=894). Now we analyze the potential differences between these

two groups. To check for differences in the means, we run a ttest assuming equal variance.

When comparing household demographics, we observe that landholders are relatively

older and more numerous than those without land. As shown in Table 4-2, the average age of the

landholder head is 51 compared to 46 in a non landholder house. It seems that households

speaking an indigenous language are more likely to be landholders. For example, only 6.4% of

non landholders speak an indigenous language compared with 30.6% of landholders. In terms of

marital status, landholders are more likely to have a partner than non landholders.

Non landholders have on average higher education than landholders, as shown in Figure

4-4. Despite the fact that differences between no schooling and elementary school are not

28 Figure 4-3 shows the number of members in the households.









observed across heads, differences are observed across partners. The partners of landholders are

more likely to have no schooling (18.81%) and elementary school (64.64%) compared to non

landholders. Furthermore, the non landholder head is on average more likely to have completed

middle education (12.2%) than non landholder heads (5.7%). The same relationship holds for the

head partner.

On the other hand, children between zero and six years old are more likely to be present

in non landholders households (32.6%) compared to landholders households (24.8%). We also

find that the first employment sector of the head affects the likelihood of holding land. For

instance, only 9.8% of those that started working outside the agricultural sector hold land in the

sample.

In the specific case of the households' liquidity constraints, it is not evident which group

faces less of a liquidity constraint. Although landholders receive on average more loans from the

bank (6.2%) than non landholders (2%), non landholders are more likely to have an account at a

bank (12.6%) than landholders (8.91%).

According to our proxy of income accounting for the annual expenses of the household,

we observe that non landholders have greater home expenses during the year $8.454 than

landholders $7,350. Furthermore, the program of Progresa, which aims to alleviate poverty in

rural Mexico, has on average a wider coverage in those households holding land (47.4%) than in

those not holding land (25.4%). These two variables give some insight that households not

holding land are indeed better off than those households holding land.

When we measure the spatial distribution of the households we notice that households

living in the Midwest and Northern part of the country are less likely to be landholders than

those households located in the central or Southern part of the country. As shown in Figure 4-5,









33.5% of the landholders live in the South, while another 31.40%, in the Central region of the

country. Only 22.7% of non landholders live in the Northwest, another 26% in the Northeast and

22.8% in the Midwest. Despite this fact, the distance from the community to the closest town

however does not seem to affect the likelihood of holding land.

In the case of international migration, 19.2% of landholders and 28.8% of non

landholders live in communities where the migrant population represents a significant percentage

of the total population, specifically more than 20%. This evidence suggests that in fact

landholders are less likely to migrate to another country than non landholders; however, one

must be careful with this evidence since this variable is being measured at the community level.

At the household level there is no significant difference between the two groups in the share of

households migrating to the United States, 19% of the migrant households are landholders and

19.4%, of the migrant households are non landholders.

However, when we analyze national migration differences across groups, we observe

than indeed landholders are more likely to migrate to another part of the country (51.5%) in the

search for a job than those not holding land (43.5%). An explanation for this finding is that

households holding land are more likely to seek seasonal work in agriculture in other parts of the

country and try to diversify their sources of income working partially in non-farm activities,

which most of the time are located outside the home community.

The landholder sample

Landholder sample contains 707 households that were those households that reported

information on plot as well as labor productivity during the 2002 survey. According to the

sample, on average the size of land a household holds is 9.82 acres, while the average cultivated

area is 5.12 acres. One needs to be careful when analyzing these statistics, since farmers in









Mexico are not homogenous. For instance, the annual value of output of these households ranges

between zero and more than $69,212,744 pesos.

According to the information obtained in the survey, the majority of the farmers rely

heavily on rainfalls for crop production. Only 24.1% of the households reported access to an

irrigation system. In terms of tenancy rights, 82% of the households reported to own at least one

plot and 55.3% to have ejidal rights over at least one plot. Of the households with ejidal rights,

55.30% were already registered in Procede at the time of the survey.

As explained above, an important limitation of the household survey is that it doesn't

capture the input information specific for each crop. So no direct inference can be made on what

agricultural inputs are being used to grow which crop. Since our analysis is based on the

household's strategy, the relationship between input use and crop pattern was studied at the

household-level.

This study analyzes three crops maize, beans and vegetables. As discussed in chapter 2,

maize and bean represent Mexico's two major staple crops. Vegetable is a high value crop that

after Mexico was admitted into NAFTA, was expected to gain in importance. In general, 77.9%

of the households that registered a positive output in 2002 grew maize, 24.6%, beans, and 12.7%,

vegetables.

Many farmers in Mexico grow their crops only for subsistence. Indeed, in the survey

55.5% of the households reported no commercialization of their 2002 production. However, for

those that reported sales, the average share of crop traded represents 45.4% of their total

production. The raising of livestock is also a common practice in rural Mexico (Davis, 2000).

The average number of cattle per household equals 2.35.









Chemical use such as fertilizers and pesticides is a common farming practice. On average

62.8% of the households reported to have used fertilizers and 50.4% pesticides during 2002.

However, only 23.5% of the households used high yield varieties (HYV). In terms of labor input,

the average number of days a household spends in agriculture is approximately 88 days in a year,

the head of the household spends the most time (54 days), followed by the wife (13 days),

son/daughter (11 days), son/daughter-in-law (6 days), and finally the grandson (4 days).

As mentioned in Chapter 2, two governmental programs promote the adoption of HYV,

technical assistance and training among farmers, Procampo and Alliance for the Countryside. In

the case of Procampo, the survey reports that 54.3% of the landholding households received a

direct income transfer from this program during 2002. Alliance for the Countryside on the other

hand has only reached 36.2% of the communities in the sample.

Migrant and non migrant mean differences

This section analyzes the household's migratory pattern for the whole sample (n=1601).

For the purpose of this research, a household is classified as a migrant household if at least one

of its members reported having a migratory experience to the United States in 2002. The data

reveals that 19.2% of the households had an international migrant in this year. During 2002,

when the survey was carried out, 4.8% of these were living permanently in the host country.

There are two main reasons why we follow this classification of migration. First, we are

measuring labor productivity at one point in time (2002), so information on labor availability is

needed for that specific year. Second, although the survey presents information on the migratory

history of the household (1980-2002) the survey does not capture the return date of the migrant,

so we cannot differentiate between those who have returned from those who haven't at the time

of the survey.









I summarized the mean differences in the demographic variables between the migrant and

non migrant households in Table 4-3. We observe, for instance, significant differences in age.

The migrant household heads are on average older (52) than non migrant households (47). The

migrant household partner is also on average older (46) than the partner in non migrant

households (40). No difference in marital status was found across groups.

Households where an indigenous language is the primary language spoken in the

household are less likely to undergo an international migration (3.9%) compared to other

households where the predominant language is Spanish (20.2%). In addition, migrant households

have on average a larger family size but fewer children between the ages of 0 and 6 years old,

compared to non migrant households. For instance 30.7% of the non migrant households have

children compared to 22.5% of the migrant households. Education at the household head level is

only significantly different across groups at the middle school level. Non migrant households are

more likely to finish middle school (10.2%) compared to migrants households (5.5%). The

education of the partner also varies across groups. In migrant households partners are more likely

to have elementary school (67.1%) compared to non migrant households (59.6%) while in non

migrant households partners are more likely to have middle school (16.38%) than in non migrant

households (8.8%).

Spatial distribution of the households also varies across the two groups. As shown in

Figure 4-6., migrant households come predominantly from the Midwest and Northeastern

regions. These findings support previous findings on Mexican migration (Massey, 1997; 1994)

that affirm that states such as Zacatecas and Guadalajara, which are located in the Midwest,

experience a high migratory flow to the United States.









Access to credit is an important limitation faced by many households in rural Mexico.

Contrasting the credit situation of both groups, it seems migrant households are more likely to

overcome liquidity constraints. Migrant households, for instance, have on average more

accounts at a bank. The sending of remittances increases the migrant households' likelihood of

having an account at a bank. No significant difference, however, was found across groups when

analyzing their ability to get a loan from a bank. When analyzing the household's annual

expenses, migrant households spend on average $11,450 while non migrant households only

spend $7,159.

Description of landholding households accounting for migratory status

This section focuses on the differences in farming practices between migrant and non

migrant households. The analysis takes into consideration only the landholders (n=707)

subsample. Table 4-4 summarizes the results of the comparison between migrant and non

migrant landholders. At a 90% confidence level we find that migrants possess more land (on

average 14. 90 acres), compared to non migrants (8.64 acres). Furthermore, migrant households

have on average a larger cultivated area, 7.48 acres, compared to non migrant households, 4.56.

Interesting enough, however, there are no differences in output among migrant and non migrant

households.

Another surprising result is that on average the number of days the migrant household

dedicates to agricultural production is not significantly different from those of non migrant

households. This condition holds for all the members of the household (head, wife, son/daughter,

son/daughter-in-law, grandchildren). This is an unexpected result, because with the migration to

the United States of one of its members, we would expect a significant reduction in the number

of days the migrant household spends working in agriculture. Although the differences are not

significant, we observe that the head, son and son-in-law in the migrant household spend on









average, less days in agriculture than non migrant households. However, the wife and grandson

in migrant households spend a little more time in agriculture than non migrant households.

Usage of non labor inputs, however, presents differences across groups. Specifically,

migrant households are more likely to adopt capital-intensive and productivity-enhancing inputs

than non migrant households. The migrant households, for example, spend on average 81% more

on fertilizers and more than the double on seed purchases than the non migrant households.

Furthermore, the migrant group is more likely to use machinery (55.2%) during the crop cycle

and HYV (32.1%) than the non migrant group (41.7%) and (22.7%) respectively. The usage of

pesticide is also larger on migrant households but at a significant level of only p
No difference in cropping patterns was found when we measured the total production of

each crop. However, when considering the predominant crop grown by the household, we find

that non migrant households are more likely to assign more than 50% of their production to

growing maize, while migrant households, to grow bean. Crop commercialization is also

different across groups but only at a significant level of p<0.10. Migrant households sell a

greater percentage of their total harvest (25.1%) in comparison to non-migrant households

(19.3%). The inverse relationship is observed in the volume left for subsistence. In the case of

livestock assets, we observe as found in the literature (Miluka, et.al. 2007), that migrant

households have a greater accumulation of livestock assets than non-migrant households. For

instance, the average number of cattle in a non-migrant household is approximately two, while in

migrant households that number increases to almost five.

Participation in governmental programs differs greatly between migrant and non migrant

households. For instance, migrant households are more likely to be enrolled in the Procampo

program (62.6%) compared to non migrants (52.4%), while non migrant households have greater









chances of receiving an income transfer from Progresa (36.4%) than migrant households

(29.6%). Precede was not significantly different across groups.

Heckman Two-Stage Procedure

In the previous section two important characteristics of the sample became evident. First,

the information on farm management practices is only available for those households with a plot

at the time of the survey (44.16% of the households). Second, we are making the assumption that

only those households that reported plot information are the ones that hold land. Given these

circumstances, using OLS to estimate a non-randomly selected sample would generate biased

estimators. The Heckman two-stage estimation procedure deals with the sample selection bias

and still analyzes the data by simple least squares methods (Heckman, 1979).29 For that reason

the Heckman two-stage procedure will be used in this research. Using this method will allow us

to draw conclusions based on the whole sample, taking into account not only the household's

agricultural productivity but their likelihood to have land.

The Heckman two-stage procedure is specified by a selection equation defined as

follows:

L, = 8'X, +y'M, +a'R, + A'C, +e, (4-1)

This equation is estimated by maximum likelihood as an independent probit model. In

this case, the dependent variable of the selection equation, L, takes the value of 1 if the

household i holds land and 0 otherwise. The independent variables, X, account for demographic

characteristics of the household, M accounts for the migratory experience of the household,

defined as the migration to the United States of at least one of the members in the household



29 This method has been commonly used to study female labor market participation and to evaluate programs in the
social science field correcting for the selectivity of the samples.









during 2002, R, is a vector of regional characteristics, C, is a vector of excluded repressor, and

e, is the error term.

Parameter estimates from the selection equation generate a vector of inverse Mills ratios.

This vector represents the estimated expected error and is introduced into the regression equation

as an explanatory variable. Only if the dependent variable from the selection equation equals 1,

will the regression equation is computed. Thus, the selection equation is the one that determines

whether an observation belongs to the regression equation or not (Heckman, 1979). In this study

for instance, the agricultural productivity of the household will only be computed if households

are landholders.

Variables included in C, are only used to estimate the first-stage of the estimation and

are excluded from the regression equation. If the same variables in the selection equation are

included in the regression equation, the estimates in the model become very imprecise. This

occurs due to the collinearity caused from adding the inverse Mills ratios into the regression

equation (Wooldridge, 2001). To avoid this imprecision in the parameter estimates, exclusion

restrictions need to be a function of the selection equation but not of the regression equation. The

next section discusses in detail the variables that will be included in the selection equation.

Y, = 8'X, + y'M, + a'R, +, '1, + A, + e, (4-2)

e, N(0, 2) (4-3)

The dependent variable in the regression equation Y is the ratio of the total output of the

household in pesos and the total days the household worked the land during 2002. The output is

expressed in Mexican pesos30 and the labor force in days per year. The independent variables in


30 The output variable was created using information on crop production obtained from the survey and information
on crop prices obtained from a Generic Index published by Banco de Mexico.









the regression equation include X, explanatory variables that account for demographic

characteristics of the household. M, represents the migratory experience of the household as

previously defined. R, is a vector of regional dummy variables. I, is a vector of variables that

are only observed when the household holds land. A, is the inverse mills ratio generated in the

selection equation and included as an extra explanatory variable in the regression equation and

e, is the error term. We assume the error term is normally distributed, with mean equal to zero

and variance equal to a constant C2 The inclusion of the inverse Mills ratio into the regression

equation removes the part of the error term correlated with the explanatory variable and deals

with the sample selection bias problem. Next we turn to the variables that will be included in the

regression equation.

Variables Description

Dependent Variables

Using sample A and sample B we estimate two models. In both, Model 1 and Model 2

the selection equation has as a dependent variable a binary variable taking the value of 1 if the

household holds or 0 if it does not hold land (land). As mentioned in previous sections, this

variable does not account for the form of land tenure; it only accounts for the fact that the

household reported information on a plot of agricultural land at the time of the survey. As shown

in Table 4-6, most of the households with land own at least one plot (82%), but the household

could also be renting or leasing the land at the time of the survey.

On the other hand, the regression equation uses as a dependent variable a ratio of the total

output of the household in pesos and the total days the household worked the land during 2002.

This ratio serves as a measure of the household's labor productivity. This dependent variable

measures the difference in labor productivity among the households. The existing literature on









labor out migration has found that migration generates a shortage in the labor supply of the

household. With this variable, we plan to measure if the shortages in labor affect the labor

productivity of the migrant households compared to the non migrant households.

We calculate the dependent variable in two different forms. Model 1 estimates the

household's head labor productivity (prod) using sample A. This variable is the ratio of the total

production of the household in pesos and the total days the household head worked the plot

during 2002. Model 2 estimates the household's total labor productivity (prodtotal) using

sample B. This variable is the ratio of the household's total annual production and the number of

days the members of the household worked the plot during the surveyed year. We assume the

household's labor force consists of the head, wife, children, son-in-law, daughter-in-law and

grandchildren.

There are two reasons why we want to measure the household head labor productivity

apart. First the head of the household is the member of the household who spends on average the

largest amount of time in agriculture (54 days). Second, the head is the member of the household

with the largest migratory participation to the United States during 2002 (65.2% of the migrants).

Independent Variables

Table 4-5 and Table 4-6 summarize statistics of the selection and regression equation. As

noted, Model 1 and Model 2 share the same independent variables. The dependent variable in the

selection equation is also the same for both models. The only variable that changes is the

dependent variable of the regression equation. Furthermore, the selection and regression

equations also share certain variables, such as the demographic, migratory and regional

variables. These variables are added in both equations because they affect not only the

probability of having land but also the labor productivity of the household. We will first describe









the variables both equations have in common and then explain separately the variables unique to

each equation.

We are including in the model the following demographic variables: gender (sex), marital

status (union), a dummy for children between zero and six years old (children), number of

people living in the household, including children (members), indigenous language spoken in the

household (indiglanguage), dummies for education level (elementary], middle], high] and up),

the age of the household head and the age squared (aged and aged2). The dummy for no school

is excluded from the model as comparison variable. The variable union accounts for both,

marriage and living together. We are keeping both marital statuses together, since we are

measuring the share of responsibilities within the household.

To avoid making the assumption that the demographic characteristics of the household

head are the characteristics of the entire household we are including in the model the

demographic characteristics of the spouse as well. These variables are education level

(wifeelementary, wifemiddle, wifehigh and wifeup), age (wifeaged) as well as age squared

(wifeaged2). The dummy for no school is also excluded from the model as comparison variable.

In addition to the demographic variables, a proxy variable for income has been created

and introduced into both equations (income)31. This variable is created summing up all the

utilities bills as well as other monthly expenses the households registered during 2002. It

includes the expenses on water, gas, wood, electricity, transportation, gasoline, television and

telephone. It is expected that households spending more are those that are better off.

A variable that controls for national migration will be included in the model

(natmigration]) as well. The migration variable that accounts for international migration is

(migration2002). This variable as described in chapter 4 and in the model encompasses those

31 The income variable was divided by a scalar of 1000 to avoid zeros in the estimated parameters.









households that have at least one member in the United States during the survey year. In order to

account for the dynamics of migration, a social network proxy (morethan20) will be used.

Originally, a social network index was built considering the migratory experience of the whole

family. However, this variable is potentially endogenous so a variable measuring the percentage

of migrants in the community will be used instead.

In Mexico, the farming practices vary greatly across regions. In the Northern part of the

country for example, where rainfalls are scarce, there is a well developed irrigation system and

agribusinesses is also well developed, while in the South agriculture relies on rainfall and

farmers' crops are mainly for subsistence. For this reason, regional dummy variables are also

incorporated into the analysis to account for these differences across regions (%v/,,/, central,

Midwest and northeast). The dummy for Northwest is excluded from the model as comparison

variable.

There are five exclusion variables introduced only in the selection equation. The selection

of these variables was made considering those variables that affect the household's likelihood of

holding land, but not its farming practices. The first exclusion variable that was chosen is

inheritance (inheritance). The idea is that many households could have inherited their land and

that is the reason why they reported plot information at the time of the survey. However,

inherited land does not affect the way the household works the land.

Household's head first employment sector (h/Ijfl %i) is the second exclusion restriction. It

is a dummy variable taking the value of one if the first sector the household head worked in was

not the agricultural sector. To control for the possible migration of the household head to other

places since his first employment, we take into account only those individuals that remained in

the same community where they were first employed. The idea is that employment outside the









agricultural sector decreases the probability of having land. For instance, being employed in the

non agricultural sector suggests that the father of the head was not a farmer or that within the

community there were other employment activities such as tourism, manufacturing and crafts

among others.

On the other hand, the other three exclusion variables were created at the community

level. The third exclusion variable is a dummy variable that accounts for those communities that

have ejidal land rights (perejidallandd). The idea is that communities with ejidal lands increase

the odds of households having land but not necessarily the odds of improving the farming

practices of the land. The fourth exclusion variable is also a dummy variable that accounts for

communities that have community land (sharedland). Community land is commonly used for

grazing animals or hunting. As before, the communities with shared land increase the probability

of holding land but do not affect the farming practices of the households. The fifth exclusion

variable is a dummy variable that accounts for those communities where agriculture is an

important source of income (incomeag). And households in communities dependent on

agriculture are more likely to hold land but not necessarily to be more productive.

Independent variables introduced only in regression equation include all those variables

related to land, such as: land size (totalacres andplot), land rights (ownland and ejidalland) and

soil quality (irrigation). Dummies for participation in governmental programs are also

incorporated in the model (procampo andprocede).

In the case of labor inputs, the labor input we are controlling for in the model is the

number of contracted workers (contracted), The non labor inputs included in the model are

application of fertilizers (fertilizer), usage of HYV (seed) and usage of machinery (machinery).

Endogeneity Issues









Endogeneity problem emerges if an independent variable included in the model is

correlated with unobservable variables in the error term. The existence of endogenous variables

in the model violates the assumption of the classical linear regression model that the explanatory

variable is uncorrelated with the stochastic disturbance term (Gujarati, 2003). In our research, the

observable variable, which is migratory experience, can be correlated with unobservable

variables not taken into account in the model that affect the likelihood of holding land and

consequently the farming practices of the household. .

If the endogeneity problem is not taken into account, the OLS estimators are not only

biased but also inconsistent (Gujarati, 2003). Previous works on labor out-migration and

agricultural productivity have used instrumental variables to solve the endogeneity problem. For

instance Mendola (2008) instruments migration using: the education level of the highest

educated household member; the sample proportion of households in the village participating in

a migratory experience; and a family chain migration variable. Miluka, et.al. (2007) used

knowledge of the language of the destination country, the share of the male population between

the ages of 20 and 39, and the minimum distance between the household and the two border

crossings. Taylor and Lopez-Feldman (2007) used as instrumental variable historic migration

such as a dummy for participation in the Bracero program as well as dummies for internal and

international migration participation in the village. The inclusion of instrumental variables in the

sample selection model, however, requires complex methods that are beyond the scope of this

research.

We used a two-step procedure, as an alternative to the instrumental variable approach, to

account for the endogeneity problem of the migration variable. First, a separate logit model for

migration is run. Second, the fitted values are estimated and incorporated into the Heckman









model replacing the original migration variable. Although this procedure is not optimal, this

methodology serves as an alternative estimation procedure to solve the endogeneity of the

migration variable. The specification of the logit model is the following:

M, = 8'X, + 'V, + r'L, +e, (4-4)

Where M, takes the value of 1 if the household i has migratory experience to the United

States in 2002, and 0 otherwise. The independent variables, X, account for demographic

characteristics of the household. V, is a potential instrumental variable of migration. L, is a

dummy variable for holding land and e, is the error term.

This study uses distance as potential instrumental variable. Two different measures of

distance are tested. First, the distance from the capital city of the state where the community is

located to the city in the United States that reported to be the primary destination of migrants in

the community. Second, the distance from the closest city where the community is located to the

nearest border between Mexico and the United States.

Distance can be used as instrumental variable because this variable is correlated with the

migratory decision but at the same time uncorrelated with the error term. Distance discourages

migration by increasing the transaction costs of migration. The other way around, the closer the

community is to the border, the cheaper the transportation and transaction costs to cross. For

example, communities near the border have greater access to information on employment

opportunities and border enforcement laws making it easier to migrate. However distance does

not impact the likelihood of holding land and the farming practices used by the household.

Conclusion

This chapter analyzed the database and the main statistics of our sample. It also presented

the methodology of this research and the variables that will be included in our model to estimate









the labor productivity of the household. The descriptive statistics clearly suggests demographic

differences among households, especially between those holding and not holding land. In

addition, migrant land holders exhibited in general greater investment in farming practices such

as usage of chemicals and seeds than non migrant households. As opposed to our expectations,

no differences in labor inputs were found across households.

In the description of the sample it also became evident the non randomness nature of the

data and the importance to account for it in the model. A Heckman Two-Stage procedure is

carried out to account for this truncation in the data and the inclusion of the logit predicted

values into the sample selection model is used to account for the endogeneity of migration in the

model. In the next chapter, I present and analyze results from the estimation.













Table 4-1. The ENHRUM community codes

R Name State Name of Municipality Name of Community Code

Magdalena Tiacotepec Magdalena Tiacotepec 1 20 053 0001

San Juan Bautista Cuicatlan San Jose del Chilar 1 20 177 0007

San Juan Juquila Vijanos San Juan Juquila Vijanos 1 20 201 0001
Oaxaca
Santa Maria Comotlan Santa Maria Comotlan 1 20 400 0001

Santa Maria Peholes Duraznal 1 20 426 0007

Santiago Jocotepec San Miguel Lachixola 1 20 468 0008

Acultzingo Potrero, El 1 30 006 0012

South- Chicontepec Piocuayo 1 30 058 0101
Southeast Espinal San Francisco 1 30 066 0024
Veracruz
Minatitlan Rancho Nuevo Carrizal 1 30 108 0034

Papantla Caristay 1 30 124 0018

Uxpanapa Ninoes Heroes 1 30 210 0098

Chankom Xkopteil 1 31 017 0017

Hunucma Sisal 1 31 038 0004
Yucatan
Tekom Tekom 1 31 081 0001

Tizimin Sucopo 1 31 096 0069

Aculco Gunyo Poniente 2 15 003 0041

Axapusco San Pablo Xuchil 2 15 016 0015

Coatepec Harinas Tecolotepec 2 15 021 0031

Ixtapan de la Sal Salitre, El 2 15 040 0013
Edo. Mex.
Ixtapan del Oro San Martin Ocoxochitepec 2 15 041 0006

Ixtlahuaca San Isidro Boxipe 2 15 042 0021

Oro, El San Nicolas El Oro 2 15 064 0048

2 Middle Acambay Tixmadeje Barrio Dos 2 15 001 0112
2 Middle Acambay- Tixmadeje-Barrio-Dos
Cuetzalan del Progreso Santiago Yancuitlalpan 2 21 043 0037

Naupan Cueyatla 2 21 100 0004

Pantepec Ejido Carrizal Viejo 2 21 111 0008

Santa Isabel Cholula Santa Ana Acozautla 2 21 148 0004
Puebla
Tecamalchalco Laguna, La 2 21 154 0005

Tlacuilotepec Rincon, El 2 21 178 0013

Tzicatlacoyan San Bernardino Tepenene 2 21 193 0008

Xicotepec Santa Rita 2 21 197 0022

Acambaro Maguey, El 3 11 002 0032

Ciudad Manuel Doblado Calzada del Tepozan 3 11 008 0015

Irapuato Laguna Larga 3 11 017 0096

Leon Patina, La 3 11 020 0394
3 Midwest Guanajuato
Leon Ibarrilla 3 11 020 0340

San Diego de la Union Sauceda, La 3 11 029 0120

San Luis de la Paz Covadonga 3 11 033 0037

Valle de Santiago San Nicolas Quiriceo 3 11 042 0116













Compostela Puerta de la Lima, La 3 18 004 0141

Xalisco Aquiles Serdan 3 18 008 0004
Nayarit
Santiago Ixcuintla Tambor, El 3 18 015 0071

Bahia de Banderas Sayulita 3 18 020 0092
3 Midwest
Loreto Tierra Blanca 3 32 024 0043

Ojocaliente Cerrito de la Cruz 3 32 036 0009
Zacatecas
Villa de Cos Sarteneja 3 32 051 0061

Villa Garcia Copetillo, El 3 32 052 0013

Ensenada Nuevo Centro de Poblacion Padre Kino 4 02 001 0170

Ensenada Nuevo Uruapan 4 02 001 0598

B.C.N. Mexicali Represa Aurelio Benansini 4 02 002 0516

Mexicali Ejido Colima I 4 02 002 0143

Mexicali Ejido Xochimilco 4 02 002 0292

Culiacan Agua Caliente de los Monzon 4 25 006 0098

Escuinapa Cristo Rey 4 25 009 0023

Guasave San Jose de Guayparime 4 25 011 0759
4 Northwest Sinaloa
Mazatlan Castillo, El 4 25 012 0162

Navolato Bledal, El 4 25 018 0023

Navolato Campo Balbuena 4 25 018 0199

Empalme Mi Patria es Primero 4 26 025 0033

Hermosillo Victoria, La 4 26 030 0669
Sonora Huatabampo Sirebampo 4 26 033 0085

Opodepe Querobabi 4 26 045 0068
Villa Pesqueira Villa Pesqueira 4 26 068 0001

Namiquipa Namiquipa 5 08 048 0001

Namiquipa Cruces 5 08 048 0031

Balleza General Carlos Pacheco 5 08 007 0054
Chihuahua
Doctor Belisario Dominguez San Lorenzo 5 08 022 0001

Guerrero Rancho de Santiago 5 08 031 0091

Juarez Millon, El 5 08 037 0643

Canatlan Nicolas Bravo 5 10 001 0083

5 Northeast Durango Colonia Hidalgo 5 10 005 0187
Lerdo Salamanca 5 10 012 0042
Durango
Nazas Perla, La 5 10 015 0020

San Dimas Vencedores 5 10 026 0110

Santiago Papasquiaro Cazadero, El 5 10 032 0024

Gonzalez San Antonio Rayon 5 28 012 0128

Matamoros Ebanito, El 5 28 022 0121
Tamaulipas
Matamoros Ranchito y Refujio, El 5 28 022 0265
San Fernando Punta de Alambre 5 28 035 0460

Source ENHRUM Codebook











Table 4-2. Landholder and non landholder mean differences

Description Total No land
Head age 48.1263 45.5797
Union 0.8401 0.8188
Indigenous language 0.1705 0.0638
Head no school 0.1699 0.1588
Head elementary school 0.2086 0.2069
Head middle school 0.0931 0.1219
Partner age 41.2492 38.4240
Partner no school 0.1399 0.1018
Partner elementary school 0.6102 0.5817
Partner middle school 0.1493 0.1801
Children 0.2911 0.3255
Members in household 4.8413 4.5749
First employment sector 0.2423 0.3568
Income $7,983 $8,485
Inheritance 0.2561 0.1879
Loan 0.0387 0.0201
Account 0.1099 0.1264
Distance 8.4484 8.5664
Migration in community 0.2455 0.2875
International migration 0.1918 0.1935
National migration 0.4703 0.4351
Progresa 0.3510 0.2539
Midwest 0.1993 0.2282
Northeast 0.2005 0.2595
Northwest 0.1824 0.2274
Central 0.2036 0.1163
South 0.2142 0.1186


Note income is expressed in Mexican pesos
* statistically significant at a 90% confidence level
** statistically significant at a 95% confidence level


I


I


Land
51.3187
0.8670
0.3055
0.1839
0.2107
0.0566
44.8215
0.1881
0.6464
0.1103
0.2475
5.1782
0.0976
$7,350
0.3423
0.0622
0.0891
8.2896
0.1924
0.1895
0.5149
0.4738
0.1627
0.1259
0.0622
0.3140
0.3352


P value
0.0000 **
0.0086 **
0.0000 **
0.1855
0.8521
0.0000 **
0.0000 **
0.0000 **
0.0083 **
0.0001 **
0.0006 **
0.0000 **
0.0000 **
0.0321 **
0.0000 **
0.0000 **
0.0178 **
0.4457
0.0000 **
0.8410
0.0015 **
0.0000 **
0.0011 **
0.0000 **
0.0000 **
0.0000 **
0.0000 **










Table 4-3. Migrant and non migrant mean differences

Description Total Non migrant Migrant P value
Head age 48.13 47.15 52.27 0.0000 **
Union 0.8401 0.8331 0.8697 0.1156
Indigenous language 0.1705 0.2017 0.0391 0.0000 **
Head no school 0.1699 0.1662 0.1857 0.4133
Head elementary school 0.2086 0.2110 0.1986 0.6344
Head middle school 0.0931 0.1020 0.0554 0.0114 **
Partner age 41.2492 40.2220 45.5785 0.0000 **
Partner no school 0.1399 0.1352 0.1596 0.2688
Partner elementary school 0.6102 0.5958 0.6710 0.0152 **
Partner middle school 0.1493 0.1638 0.0879 0.0008 **
Children 0.2911 0.3068 0.2248 0.0044 **
Members in household 4.8413 4.6468 5.6612 0.0000 **
Income $7,983 $7,159 $11,450 0.0000 **
Inheritance 0.2561 0.2604 0.2378 0.4140
Loan 0.0387 0.0371 0.0456 0.4876
Account 0.1099 0.0997 0.1531 0.0071 **
Midwest 0.1993 0.1592 0.3681 0.0000 **
Northeast 0.2005 0.1862 0.2606 0.0003 **
Northwest 0.1824 0.1963 0.1238 0.0031 **
Central 0.2036 0.2110 0.1726 0.1339
South 0.2142 0.2407 0.0749 0.0000 **
Note income is expressed in Mexican pesos
* statistically significant at a 90% confidence level
** statistically significant at a 95% confidence level











Table 4-4. Migrant landholder and non migrant landholder mean differences


Description
Farm sizze
Cultivated area
Ejidalland
Irrigation
Total days worked
Total days head worked
Total days spouse worked
Total days son worked
Total days son-in-law worked
Total days grandson worked
Contracted workers
Machinery
HYV


Total
9.8235
5.1178
0.5530
0.2405
88.2588
54.1556
12.9901
11.2178
6.1726
3.7228
16.5958
0.4427
0.2448


Non migrant
8.6359
4.5645
0.5375
0.2339
89.9616
55.4049
12.8953
11.9808
6.4642
3.2164
16.7055
0.4171
0.2267


Chemical purchase $1,874.1170 $1,645.9600
Pesticide purchase $3,074.2280 $2,136.3180
Seed purchase $188.2628 $150.0360
Output $1,249,276 $1,251,436
Maize $285,268 $297,026
Bean $216,972 $243,778
Vegetables $675,938 $606,808
Mainly production of maize 0.5149 0.5462
Mainly production of bean 0.0509 0.0401
Mainly production of vegetable 0.0537 0.0506
Traded volume 0.2031 0.1927
Subsistence volume 0.3904 0.4048
Cattle 2.3567 1.6638
Procampo 0.5431 0.5236
Precede 0.4696 0.4817
Progresa 0.3510 0.3640
Alianza 0.3616 0.3570
Note the monetary variables are expressed in Mexican pesos
* statistically significant at a 90% confidence level
** statistically significant at a 95% confidence level


I


I


Migrant
14.9021
7.4838
0.6194
0.2687
80.9776
48.8134
13.3955
7.9552
4.9254
5.8881
16.1269
0.5522
0.3206
$2,978.8790
$6,710.2350
$347.8818
$1,240,042
$223,097
$132,508
$1,033,784
0.3806
0.0970
0.0672
0.2505
0.3248
5.2248
0.6269
0.4179
0.2964
0.3811


P value
0.0890 *
0.0000 **
0.0863 *
0.3968
0.4995
0.3822
0.9015
0.2562
0.4979
0.2126
0.8866
0.0045 **
0.0247 **
0.0038 **
0.0870 *
0.0170 **
0.9813
0.7368
0.6912
0.6122
0.0005 **
0.0070 **
0.4450
0.0841 *
0.0157 **
0.0000 **
0.0307 **
0.1835
0.0257 **
0.4303














Table 4-5. Selection equation variables statistics

Variable Description Mean Std. Dev. Min Max Obs
land the household holds land (1 =yes) 0.4416 0.4967 0 1 1601


Demographics
sex
children
union
members
indiglanguage
elementary
middle 1
highly
up
aged
aged2
wifeelementary
wifemiddle
wifehigh
wifeup
wifeaged
wifeaged2
income
Migration
natmigration 1
migration2002
morethan20
Region


Gender (1 male)
Children from 0 to 6 in the house (1 yes)
Marital status (1= married or living together)
Number of people in the household
Language spoken (1- indigenous language)
Schooling (1 elementary finished or not)
Schooling (1 middle school finished or not)
Schooling (1 high school finished or not)
Schooling (1-techinical, college or graduate)
Age ofthe head
Age of the head squared
Spouse schooloing( 1-elementary finished or not)
Spouse schooloing(l=middle finished or not)
Spouse schooling(1=high school finished or not)
Spouse schooling(l-technical, college or graduate)
Age of the spouse
Age of the spouse squared
Proxy for household income


National migration (1-yes)
Migratory experience in 2002 (1-yes)
Social network proxy


south Region (1 south-southeast)
central Region (1-middle)
midwest Region (1-middlewest)
northeast Region (1-northeast)
northwest Region (1-northwest)
Exclusion Variables
hhfirst first labour sector (1- no agricultural sector)
perejidallandd Community has ejidal land (1 yes)
sharedland Community has shared land (1 yes)
incomeag Agriculture important source of income (1-yes)
inheritance Received inheritance (1-yes)
Note All monetary variables are expressed in Mexican pesos


0.8670
0.2911
0.8401
4.8413
0.1705
0.6084
0.1318
0.0369
0.0387
48.1263
2558.1190
0.6102
0.1493
0.0319
0.0394
41.2492
1933.6430
$7,983


0.4703
0.1918
0.2455


0.2142
0.2036
0.1993
0.2005
0.1824


0.2561
0.0843
0.7893
0.5184
0.6202


0.3397
0.4544
0.3666
2.1555
0.3762
0.4883
0.3384
0.1885
0.1930
15.5604
1614.3200
0.4878
0.3565
0.1757
0.1945
15.2412
1414.2090
$10,514


0.4993
0.3938
0.4305


0.4104
0.4028
0.3996
0.4005
0.3863


0.4366
0.2780
0.4053
0.4998
0.4855


1 1601
1 1601
1 1601
14 1601
1 1601
1 1601
1 1601
1 1601
1 1601
95 1601
9025 1601
1 1601
1 1601
1 1601
1 1601
90 1601
8100 1601
$149,100 1601


1 1601
1 1601
1 1601


1 1601
1 1601
1 1601
1 1601
1 1601


1 1601
1 1601
1 1601
1 1601
1 1601














Table 4-6. Regression equation variables statistics


Variable Description
prodO Labor productivity of the household he
prodtotalO Labor productivity of the household
Demographics
sex Gender (1 = male)
children Children from 0 to 6 in the house (1lyi
union Marital status (1= married or living tog
members Number of people in the household
indiglanguage Language spoken (1= indigenous langu
elementary Schooling elementaryy finished or n
middle 1 Schooling (1lmiddle school finished oi
highly Schooling (1lhigh school finished or n
up Schooling (1-techinical, college or gra
aged Age of the head
aged2 Age of the head squared
wifeelementary Spouse schooloing(1 elementary finish
wifemiddle Spouse schooloing(1=middle finished
wifehigh Spouse schooling(1=high school finish
wifeup Spouse schooling(l-technical, college
wifeaged Age of the spouse
wifeaged2 Age of the spouse squared
income Proxy for household wealth
Migration
natmigrationi National migration (1lyes)
migration2002 Migratory experience in 2002 (1lyes)
morethan20 Social network proxy
Region


ad





es)
ether)


age)
ot)
rnot)
ot)
duate)



ied or not)
or not)
ed or not)
or graduate)


south Region (1=south-southeast)
central Region (1lmiddle)
midwest Region (1lmiddlewest)
northeast Region northeasta)
northwest Region northweste)
Plot Characteristics
plot Amount of plots per household
ownland the household owns the land (1=yes)
ejidalland propietary rights (1=ejidal)
totalacres Total acres of land
irrigation pattern (1 irrigation)
fertilizer Input use (1=used fertilizer)
seed Input use (1=high yield variety)
machinery Input use (1=use machinery)
contracted Total contracted workers
Programs
procampo Participates in procampo (1lyes)
precede Enrolled in precede (1 yes)
The dependent variables prod and prodtotal are expressed in pesos/day


0.2142 0.4104 0 1 1601
0.2036 0.4028 0 1 1601
0.1993 0.3996 0 1 1601
0.2005 0.4005 0 1 1601
0.1824 0.3863 0 1 1601


1.7072 1.0648 1 8 707
0.8204 0.3842 0 1 707
0.5530 0.4975 0 1 707
9.8235 38.3945 0 537.5 707
0.2405 0.4277 0 1 707
0.6280 0.4837 0 1 707
0.2348 0.4242 0 1 707
0.4427 0.4971 0 1 707
16.5958 42.2539 0 488 707


0.5431 0.4985 0 1 707
0.4696 0.4994 0 1 707


Mean
$50,657
$34,929


0.8670
0.2911
0.8401
4.8413
0.1705
0.6084
0.1318
0.0369
0.0387
48.1263
2558.1190
0.6102
0.1493
0.0319
0.0394
41.2492
1933.6430
$7,983


0.4703
0.1918
0.2455


I


Std. Dev.
$308,471
$247,697


0.3397
0.4544
0.3666
2.1555
0.3762
0.4883
0.3384
0.1885
0.1930
15.5604
1614.3200
0.4878
0.3565
0.1757
0.1945
15.2412
1414.2090
$10,514


0.4993
0.3938
0.4305


Max
$5,775,027
$5,775,027


1
1
1
14
1
1
1
1
1
95
9025
1
1
1
1
90
8100
$149,100




















100
0
15-24 25-34 35-44 45-54 55-64 65-74 75-84 85-95
Age




Figure 4-1. Age of the household head


*0OO
03o *180
03o

*2%0

*20%
0400

0900


Figure 4-2. Education level of the household head







600r


S 400
s 200
II 200-


* elementary
* elementary school not finished
O middle school
O middle school not finished
* high school
* high school not finished
* technical
O college
* graduate
*noschool


1 2 3 4 5 6 7 8 9 10 >10
Members in the household





Figure 4-3. Number of members in the household


0 NJ 0 M




















m ~ 0.4-

a 0.3













0 -Education level
Figure 4-4. Comparison of education level by landholding status
o 0.1 o -












Figure 4-4. Comparison of education level by landholding status


0-
cii
ao0
0 E

00


0 3000
0 2500
0 2000
0 1500
0 1000
0 0500
0 0000


D Total
ONo land
*Land


Middlewest Northeast Northwest Middle South

Location of households


Figure 4-5. Landholder and non landholder spatial distribution









03500
0 000-'
0 250100





0 00000
Middlewest Northeast Northwest Central South
Location ofhousehold





Figure 4-6. Migrant and non migrant spatial distribution


0 Total
DNon Migrant
* Migrant


Ono land




* land









CHAPTER 5
RESULTS

Introduction

In this chapter, we present and analyze the results of econometric models using the

Mexican National Rural Household Survey (ENHRUM). As mentioned in the previous chapter

we differentiate between the labor productivity of the head (henceforth, Model 1) and the labor

productivity of the entire household (henceforth, Model 2). With this distinction we aim to

evaluate the New Economics of Labor Migration approach that states that in order to study the

way labor out-migration influences the labor productivity of rural households, the analysis needs

to rely upon the household as unit of analysis. In both models we estimate labor productivity

accounting for the selectivity of landholding and introduce a variable of social network to

capture how the formation of social networks impact the household labor productivity.

This chapter is structured as follows. We first run the OLS model and present the

results32. The second section summarizes the results found when running the Heckman model33.

The third section reviews the results obtained when the endogeneity of migration is taken into

account in the model. The fourth section presents conclusions.

Estimation of OLS

We first estimate the regression equation of Model 1 and Model 2 by OLS. Although

OLS does not take into consideration the sample selection problem as well as the potential

endogeneity of the migration variable, this model is the simplest way to estimate labor

productivity and will be used as a framework to compare the results found in more complex

models.


32 The models in this research were run using STATA
33 The inverse of the Mills' ratio was first computed to test for the specification of the model and was found
significant at a p<0.05. This result corroborates the existence of sample selection in our data and the need to account
for it. The Heckman model was run using robust standard errors









Table 5-1 and Table 5-3 summarize the results found estimating the household head labor

productivity (Model 1) and the household total labor productivity (Model 2) by OLS. An

important finding of Model 1 is that it is not our migratory variable (migration2002) but the

proxy for social network (morethan20) that is negative and statistically significant at a 95%

confidence level. This finding suggests that the more a community is involved in migration, the

less labor productive the household head is going to be. For instance, the labor productivity of

the household head is $91,393.64 less in communities where social networks are strong.

On the other hand, a head with high school education compared to a head with no

education is more productive by $111,730.50. The education of the spouse at elementary and

middle school level, however, has a negative impact on the labor productivity of the household

head compared to a spouse with no school. For example, a spouse with middle school education

is less labor productive by $121,921.20 compared to a spouse with no school. Internal migration

of the household head to other places in the country impacts negatively the labor productivity of

the household head by $46,457.16.

Measuring labor input we find that contracted is statistically significant at a 95%

confidence level. An increase in one unit of wage worker increases the labor productivity of the

household head by $740.28. This result suggests that contracted labor force can be used as

substitute or complement of the household head labor. In the case of non labor inputs, the usage

of fertilizers increases the labor productivity of the household head by $54,753.31 but the

utilization of machinery reduces it by $58,404.49. On the other hand, the governmental program,

Procede has a positive effect on the labor productivity of the household head increasing the labor

productivity of the household head by $68,513.92, but Procampo has a negative impact reducing

the labor productivity by $46,362.49. We also observe that the household head labor









productivity is higher in Midwest but lower in the Central region compared to the Northwestern

region.

In Model 2 the migratory variable is negative but not significant. The variable for social

network (morethan20) is again negative and statistically significant at a 95% confidence level.

However, the variable in Model 2 has a greater impact on labor productivity than in Model 1. In

communities where migration represents more than 20% of the population, household head labor

productivity is $93,465.52 less than in those communities where migration is not so important. It

can be argued that in those communities where migration is a common practice, the labor market

of the whole community is affected, decreasing the number of people available to work the land.

Other variables that were found statistically significant in Model 2 at a 95% confidence

level were wifemiddle, wifeup, fertilizer, machinery and Midwest. The variables that were

significant at a 90% confidence level were wifeelementary, middle], income, natmigration],

procampo andprocede. The major difference between the two model specifications is the

contracted variable, this variable loses statistical significance in Model 2. It seems wage workers

do not increase the labor productivity of the entire household. It can be argued that households

employing the labor force of the entire household are less likely to hire wage workers compared

to households where only the head of the households works the land. In this case, the hired in

labor acts more as a substitute than a complement for the household labor force. For this reason,

the contracted variable loses statistical significance in Model 2.

Heckman Two-Stage Estimation

Who holds Land?

As shown in Table 5-2 and 5-4, the selection equation from both models report similar

results. The gender of the household head (sex) has a statistically significant effect on the

likelihood of holding land. Female-headed households are less likely to own an agricultural









asset and work in agriculture. Similarly, age (aged) increases the odds of holding land, but the

squared age (aged2) has the opposite effect, meaning there is a breaking point where age no

longer affects the likelihood of landholding. Union is negative and statistically significant but

only in Model 2. This finding suggests that the marital status of the household affects the odds of

landholding at the household level but not at the household head level. Couples that are married

or living together are less likely to hold land. .

In both models, the sign of the indigenous language variable is positive and statistically

significant, as expected. This means that households where an indigenous language is spoken are

more likely to make a living from agriculture and consequently are more likely to have land than

those households where Spanish is spoken. While the dummy for higher education level (such as

technical, college and graduate) is negative and statistically significant, is positive and

statistically significant the dummy variable for elementary school. These findings suggest that

household heads with elementary school are more likely to have land compared to those with no

schooling; however, households with a technical or college degree are less likely to hold land

compared to households with no schooling. This means that households with higher education

prefer to work outside the agricultural sector where the return to education is higher.

In terms of spatial distribution, the soil quality in the Southern part of Mexico and the

weather conditions favor the growth of crops in this region. This fact is supported in our

findings. For instance, we find that the dummy variables for the Southern, Central and

Midwestern parts of the country are positive and statistically significant compared to the dummy

variable for the Northwestern region. On the other hand, the dummy variable for the

Northeastern part is also positive but not statistically significant compared to the Northwestern

region.









National migration has a negative impact at a 95% confidence level on the likelihood of

holding land only in Model 2. This finding suggests that households engaging in national

migration are less likely to hold land. International migration variable is negative but not

significant in both models. The social network variable however is positive and statistically

significant. It can be argued that communities where migration is a common practice have a

higher propensity to hold land.

On the other hand, the dummies that account for the existence of ejidal land rights

(perejidallandd) and the existence of community land (sharedland) have a positive effect on the

likelihood of holding land. These findings suggest that in those communities consisting of ejidos,

the probability was higher that the households obtained land. The same reasoning applies for

those communities sharing land.

Furthermore, the dummy that accounts for those communities where agriculture is an

important source of income (incomeag) is also positive and statistically significant, meaning that

those communities highly dependent on agriculture are more likely to have land holders than in

communities where agriculture is not an important source of income. Inheritance also increases

the odds of landholding. The ejidal land for instance, was commonly inherited through

generations by the male in charge of the household. That is why it is nor surprising to find a

positive relationship between inheritance and land. Finally, the first employment outside the

agricultural sector is the only variable with a negative value, reducing the odds of households

having land as predicted.

What affects Labor Productivity?

The results found in the regression equation differ between the two models. For this

reason, the results of Model Iwill be analyzed first followed by the results of Model 2.

Model 1 household head labor productivity (sample A)









As shown in Table 5-1, the education level of the household head (middle]) affects

positively the labor productivity of the head by $111,725.80. This relationship, however, is only

observable at a 90% confidence level. This finding suggests, as opposed to expected, that

agricultural work becomes more productive with the level of education of the farmer. On the

other hand, the education level of the spouse (wifemiddle) has a negative and statistically

significant impact on the household head labor productivity. A household having a spouse with

middle education is $122,025.20 less labor productive that a household having a spouse with no

school.

Our migratory variable (migration2002) is negative but not statistically significant.

National migration however (natmigrtion]) is negative and statistically significant at a 95%

confidence level and social networks (morethan20) at a 90% confidence level. In general, it can

be argued that migration indeed affect the labor productivity of the household head. Households

engaging in internal migration to other parts in the country are less labor productive than those

staying in the community by $46,431.76. Furthermore, in communities where social networks

are strong, the household head is also less labor productive by $91,564.94.

Similarly to the results found in OLS, the usage of fertilizer increases the labor

productivity of the household head by $54,768.26, but the utilization of machinery reduces it by

$58,439.23. On the other hand, contracted is still positive and statistically significant. In this

case, an additional wage worker increases the labor productivity of the household head by

$740.03.

Regional dummy variables were not significant meaning that the location of the

community does not play a role in determining the labor productivity of the household head. In

terms of governmental programs, Procede is positive and statistically significant at a 95%









confidence level while Procampo is negative but only statistically significant at a 90%

confidence level. Households enrolled in Procede increase the labor productivity of the

household head by $68,562.32. However, households enrolled in Procampo seem to reduce the

labor productivity of the head by $46,336.40. In this model we also found an inverse relationship

between number of plots (plot) and labor productivity at a 95% confidence level. Although farm

size is not a topic in this study, further research in this area is recommended.

Model 2 household labor productivity (sample B)

The results obtained in Model 2 are summarized in Table 5-3. Middle education level of

the household head (middle]) as well as the age of the head squared (aged2) lost statistical

significance in this model. The dummy variable accounting for middle education level of the

spouse (wifemiddle) remains negative and statistically significant at a 95% confidence level,

while the variable accounting for elementary school (wifeelementary) also remains statistically

significance at a 90% confidence level. It seems that as the education level of the spouse

increases, reduces the labor productivity of the household. For instance, households having a

spouse with elementary education are $43,954.47 less labor productive and households having a

spouse with middle education are $84,443.23 less labor productive than households that have a

spouse with no school.

On the other hand, the proxy variable used to measure income is positive and statistically

significant, suggesting that households with less liquidity constraints are those that achieve

higher labor productivity. For instance, an increase of $1,000.00 in the expenditure level of the

household increases the labor productivity of the household by $1,646.98.

We found that the variable of interest in this study (migration2002) is negative and

statistically significant in this model. As opposed to expected, migrant households are less labor

productive than households with no migratory experience by $28,655.31. On the other hand, the









variables that account for national migration (natmigration]) as well as social network

(morethan20) are also negative and statistically significant at a 90% and a 95% confidence level

respectively. In general, it looks like migration, either national or international, reduces the labor

productivity of the entire household.

Similarly to the results found in OLS, the usage of fertilizer increases the labor

productivity of the household by $49,196.56 but the utilization of machinery reduces it by

$41,890.97. It should be noticed, however, that at the household level the positive impact of

fertilizer compensates for the lost of productivity in machinery, while at the household head level

the net effect is negative. On the other hand, Procampo is still negative and statistically

significant at the 90% confidence level. In this case, households enrolled in Procampo are

$36,933.66 less productive than those not enrolled in the program.

Two regional dummy variables gain significance in this model. For instance, the dummy

variables for South and Central are negative and statistically significant in Model 2. Households

living in the Southern region are $71,102.08 less labor productive than those living in the

Northwestern region. And households living in the Central region are $79,421.25 less labor

productive than the households residing in the Northwester region. This result suggests the

existence of regional differences in the agricultural labor productivity of the households.

Addressing Migration Endogeneity

In order to address the endogeneity of migration, we decided to use as potential

instrumental variable the distance from the closest city where the community is located to the

nearest border between Mexico and the United States (kilometers). This variable, however, was

only found statistically significant using sample A, which represents the sample for the entire

household. This finding suggests the instrumental variable is vulnerable to changes in the sample









size and certainly might not be the optimal instrument for migration. However, given the scope

of this research this is the instrumental variable that will be used to estimate Model 2.

The estimation of the logit model for migration is shown in Table 5-5. The instrumental

variable (kilometers) is negative and statistically significant. This finding supports the idea that

distance discourages migration by increasing the transaction costs of migration. The predicted

values (ivmig) of the logit model are calculated and used to estimate the Heckman model.

As Table 5-6 Model 2a summarizes, the variables of interest morethan20 and

migration2002 lose their significance when the endogeneity of migration is taken into account,

meaning migration has no impact on the labor productivity of the household. Variables such as

middle], %\i/nh and central also lose significance. The regional dummies also lose significance.

The variables that remained statistically significant at a 95% confidence level were fertilizer,

wifemiddle and income. The variables that remained statistically significant but at a 90%

confidence level was wifeelementary, contracted, procampo, machinery and natmigraation].

We present results from the selection equation estimation in Table 5-7 Model 2a. The

results suggest that the predicted values of migration generate no significant differences in the

estimation. The only change occurs in the variable of ivmig. The variable remains negative but

gains statistical significance at a 95% confidence level. This means that households having a

migrant in the United States during 2002 are less likely to hold land.

There are different ways to interpret the lost of significance in the migration

(migration2002) and social network variable (morethan20) when the predicted values of

migration are included in the model. We would argue that the predicted value of migration

(ivmig) can possibly be correlated with the social network variable reducing the significance in

both variables. The argument behind is that households with strong social networks tend to be









those located nearer the border between Mexico and the United States. In order to test this

argument we removed from the model our social network variable (morethan20) to measure the

effect of the predicted value (ivmig) alone.

In Table 5-6 and Table 5-7, Model 2b we summarize the results. As expected, the

migration variable (migration2002) gains statistical significance in the regression equation and

loses statistical significance in the selection equation. These findings suggest that, once corrected

for endogeneity, migration reduces the labor productivity of the household by $367,465.30, but

has no effect on the likelihood of landholding.

Conclusion

This chapter described the results of the model. Different estimation methodologies were

used to evaluate the results. The results found in the Heckman selection model support the idea

that the likelihood of landholding depends not only on the demographic characteristics of the

household but on the specific location of the household. It was proven that communities with

ejidal land and community land increase the odds of landholding.

Our model suggests that migrant households are $28,655.31 less labor productive than

those households with no migratory experience. The formation of social networks in the

community also has a negative effect. However, when we solve for the endogeneity problem

using a potential instrumental variable the impact of international migration and social network

becomes less clear. For a better understanding of the impact of migration on labor productivity,

further research needs to be carried out, dealing more properly with the endogeneity problem of

migration. My future research includes finding new instrumental variables that are correlated

with migration but not with the formation of social networks to be able to account for the impact

of both, migration and social networks on labor productivity.











Table 5-1. Model 1 household head labor productivity
Model I OLS Estimates Model I Heckman selection model
Number of obs 667 Number of obs 1561
F(36, 630) 4.69 Censored obs 894
Prob > F 0.0000 Uncensored obs 667
R-squared 0.2113 Log pseudolikelihood ########
Adj R-squared 0.1662 Wald chi2(36) 49.84
Root MSE 2.80E+05 Prob > chi2 0.0623
Robust Robust
prodO Coef. t Coef. z
Std. Err. Std. Err.


sex 22156.04 58800.
union -48224.82 48051.
children -28451.91 33894.
members 3624.18 5928.
indiglanguage 18899.77 32021.
elementary 28989.99 31918.
middle 111730.50 50393.
highly 145698.60 95537.
up -106050.30 95377.
aged -6072.11 6611.
aged2 89.74 61.
wifeelementary -51338.69 31208.
wifemiddle -121921.20 48961.
wifehigh -55404.49 105966.
wifeup 417508.20 92128.
wifeaged -1660.19 5020.
wifeaged2 -18.35 49.
income 5271.19 1119.
natmigrationi -46457.16 24157.
migration2002 -35733.82 32126.
morethan20 -91393.64 42643.
contracted 740.28 273.
ejidalland -38637.08 26178.
ownland 24948.89 32831.
plot -17658.74 11593.
irrigation -31732.82 29143.
totalacres 92.51 302.
procampo -46362.49 26635.
precede 68513.92 28874.
seed 44213.59 29621.
fertilizer 54753.31 24507.
machinery -58404.49 24398.
south -88759.02 59996.
central -95113.27 57616.
midwest 95621.20 58076.
northeast 24942.84 61015.
cons 289831.40 176746.
* statistically significant at a 90% confidence level
** statistically significant at a 95% confidence level


80
15
03
29
58
02
88
32
49
69
00
07
85
40
15
46
76
29
57
28
25
81
17
79
58
47
37
78
97
04
97
91
48
47
67
85
00


0.38
-1.00
-0.84
0.61
0.59
0.91
2.22 **
1.53
-1.11
-0.92
1.47
-1.65 *
-2.49 **
-0.52
4.53 **
-0.33
-0.37
4.71 **
-1.92 *
-1.11
-2.14 **
2.70 **
-1.48
0.76
-1.52
-1.09
0.31
-1.74 *
2.37 **
1.49
2.23 **
-2.39 **
-1.48
-1.65 *
1.65 *
0.41
1.64


21530.38
-48098.81
-28444.99
3629.23
18449.74
28865.51
111725.80
145764.70
-105728.10
-6123.01
90.17
-51439.26
-122025.20
-55409.56
417548.70
-1654.77
-18.47
5265.83
-46431.76
-35726.83
-91564.94
740.03
-38725.38
24886.86
-17679.92
-31737.23
92.34
-46336.40
68562.32
44228.12
54768.26
-58439.23
-89662.71
-96269.61
95050.16
24609.01
293594.60


60281.05
63517.30
20871.31
2894.48
28300.47
34059.74
60700.14
231717.60
89321.10
7796.20
92.96
34044.48
49117.96
100788.60
396237.60
2687.34
39.72
3333.46
22447.38
35255.50
51326.15
298.32
26707.39
16179.14
8350.82
29104.16
199.17
25128.67
31486.36
31041.52
24479.08
26682.26
67552.83
70440.76
98307.66
86014.14
181212.20


0.36
-0.76
-1.36
1.25
0.65
0.85
1.84 *
0.63
-1.18
-0.79
0.97
-1.51
-2.48 **
-0.55
1.05
-0.62
-0.46
1.58
-2.07 **
-1.01
-1.78 *
2.48 **
-1.45
1.54
-2.12 **
-1.09
0.46
-1.84 *
2.18 **
1.42
2.24 **
-2.19 **
-1.33
-1.37
0.97
0.29
1.62











Table 5-2. Model 1 Heckman selection model (first-stage)
Robust
land Coef. S E s z
Std. Err.
sex 1.0653 0.1368 7.79 **
union -0.1306 0.1257 -1.04
children 0.0310 0.0789 0.39
members -0.0177 0.0153 -1.16
indiglanguage 1.0448 0.0791 13.21 **
elementary 0.1858 0.0841 2.21 **
middle -0.0936 0.1262 -0.74
highly 0.0485 0.2257 0.21
up -0.5337 0.2365 -2.26 **
aged 0.1041 0.0147 7.06 **
aged2 -0.0009 0.0001 -6.74 **
wifeelementary 0.1112 0.0820 1.36
wifemiddle 0.2050 0.1253 1.64
wifehigh 0.1192 0.2251 0.53
wifeup -0.2290 0.2023 -1.13
wifeaged -0.0005 0.0131 -0.04
wifeaged2 0.0002 0.0001 1.30
income 0.0133 0.0026 5.20 **
natmigrationi -0.0854 0.0593 -1.44
migration2002 -0.1093 0.0742 -1.47
morethan20 0.2315 0.0815 2.84 **
south 1.1042 0.1159 9.53 **
central 1.7134 0.1139 15.05 **
midwest 0.6826 0.1181 5.78 **
northeast 0.0908 0.1148 0.79
inheritance 0.3687 0.0629 5.87 **
hhfirst -0.3390 0.1416 -2.39 **
perejidallandd 0.4337 0.0699 6.20 **
incomeag 0.6555 0.0619 10.60 **
sharedland 0.3737 0.0673 5.55 **
cons -6.1941 0.3894 -15.91
statistically significant at a 90% confidence level
** statistically significant at a 95% confidence level











Table 5-3. Model 2 household labor productivity
Model 2 OLS Estimates Model 2 Heckman selection model
Number of obs 707 Number of obs 1601
F( 36, 670) 4.29 Censored obs 894
Prob > F 0.0000 Uncensored obs 707
R-squared 0.1874 Log pseudolikelihood -10452.8
Adj R-squared 0.1437 Wald chi2(36) 52.58
Root MSE 2.30E+05 Prob > chi2 0.0366
Robust Robust
prodtotal0 Coef. S. Efs t Coef. Std s z
Std. Err. Std. Err.


sex
union
children
members
indiglanguage
elementary 1
middle 1
highly
up
aged
aged2
wifeelemenetar
wifemiddle
wifehigh
wifeup
wifeaged
wifeaged2
income
natmigrationi
migration2002
morethan20
contracted
ejidalland
ownland
plot
irrigation
totalacres
procampo
procede
seed
fertilizer
machinery
south
central
midwest
northeast
cons


39708.74
-50293.66
-15422.31
2427.40
-5708.15
18974.67
74869.70
-58011.55
-68255.98
-5548.46
77.37
-43485.95
-83941.08
12680.68
425349.60
1380.37
-42.09
1670.64
-33910.61
-28648.77
-93465.52
277.93
-22836.02
19492.39
-10274.86
-15749.38
170.10
-37047.90
42227.89
22502.10
49146.51
-41698.15
-66894.88
-74207.25
99111.80
48548.02
190808.60


43032.38
37316.21
26898.12
4703.18
25181.74
24870.10
40384.59
75710.42
73248.88
4876.22
44.11
24804.56
38712.39
71455.62
71733.19
3676.32
36.46
893.34
19050.50
25184.21
33979.27
220.14
20761.91
25955.63
9260.78
22979.14
245.25
21064.19
22727.55
23200.80
19444.91
19263.46
47159.96
45421.68
46141.59
48494.34
135449.50


0.92
-1.35
-0.57
0.52
-0.23
0.76
1.85 *
-0.77
-0.93
-1.14
1.75 *
-1.75 *
-2.17 **
0.18
5.93 **
0.38
-1.15
1.87 *
-1.78 *
-1.14
-2.75 **
1.26
-1.10
0.75
-1.11
-0.69
0.69
-1.76 *
1.86 *
0.97
2.53 **
-2.16 **
-1.42
-1.63
2.15 **
1.00


206998.40 128378.20 1.61


* statistically significant at a 90% confidence level
** statistically significant at a 95% confidence level


37356.29
-49534.68
-15421.46
2436.32
-7645.38
18374.13
74834.65
-57592.38
-66858.30
-5745.15
78.90
-43954.47
-84443.23
12196.17
425457.90
1384.71
-42.39
1646.98
-33735.25
-28655.31
-94257.02
276.84
-23273.56
19266.95
-10363.87
-15741.79
169.27
-36933.66
42430.69
22602.03
49196.56
-41890.97
-71102.08
-79421.25
96582.44
47024.89


80298.22
54401.42
15954.31
2075.05
8991.74
24733.69
48948.84
69222.37
65656.63
5659.00
70.74
24167.21
37430.47
39422.10
372003.90
2375.37
27.81
751.02
19171.59
14614.80
41681.22
163.63
25967.88
12126.63
7418.71
23073.91
182.16
22155.06
27140.45
20652.55
20153.34
24872.56
35214.74
37540.04
85615.68
52231.18


0.47
-0.91
-0.97
1.17
-0.85
0.74
1.53
-0.83
-1.02
-1.02
1.12
-1.82 *
-2.26 *
0.31
1.14
0.58
-1.52
2.19 *
-1.76 *
-1.96 *
-2.26 *
1.69 *
-0.90
1.59
-1.40
-0.68
0.93
-1.67 *
1.56
1.09
2.44 *
-1.68 *
-2.02 *
-2.12 *
1.13
0.90











Table 5-4. Model 2 Heckman selection model (first-stage)
Robust
prodtotal0 Coef. S E s
Std. Err.


sex 0.8583
union -0.2304
children 0.0367
members -0.0108
indiglanguage 1.0073
elementary 0.2280
middle -0.0873
highly 0.0192
up -0.4961
aged 0.0865
aged2 -0.0007
wifeelementary 0.1053
wifemiddle 0.2297
wifehigh 0.2088
wifeup -0.1218
wifeaged 0.0097
wifeaged2 0.0000
income 0.0130
natmigrationi -0.1258
migration2002 -0.0734
morethan20 0.2191
south 1.1475
central 1.7027
midwest 0.6758
northeast 0.0993
inheritance 0.3528
hhfirst -0.2908
perejidallandd 0.4673
incomeag 0.6545
sharedland 0.3700
cons -5.8038
* statistically significant at a 90% confidence level
** statistically significant at a 95% confidence level


z


0.1221
0.1119
0.0771
0.0144
0.0752
0.0805
0.1223
0.2235
0.2310
0.0142
0.0001
0.0789
0.1203
0.1993
0.2130
0.0125
0.0001
0.0025
0.0577
0.0721
0.0805
0.1128
0.1135
0.1156
0.1131
0.0620
0.1309
0.0681
0.0599
0.0631
0.3783


7.03 **
-2.06 **
0.48
-0.75
13.39 **
2.83 **
-0.71
0.09
-2.15 **
6.11 **
-5.47 **
1.33
1.91
1.05
-0.57
0.78
0.40
5.09 **
-2.18 **
-1.02
2.72 **
10.17 **
15.00 **
5.85 **
0.88
5.69 **
-2.22 **
6.87 **
10.92 **
5.86 **
-15.34














Table 5-5. Logit model for migration


Number of obs
LR chi2(26)
Prob > chi2
Pseudo R2
Log likelihood
migratipn2002 Coef.
sex -0.3932
union 0.2942
children 0.2468
members 0.1429
indiglangu-e -1.2749
elementary 0.1594
middle -0.4495
highly -0.2057
up -0.7173
aged 0.0602
aged2 -0.0004
wifeelementary 0.0866
wifemiddle -0.2840
wifehigh 0.4914
wifeup 0.4192
wifeaged 0.0104
wifeaged2 -0.0001
income 0.0249
natmigrationi -0.0082
morethan20 1.1803
south 0.5353
central 0.8490
midwest 1.1595
northeast 0.3268
kilometers -0.0005
land -0.0641
cons -5.0046
* statistically significant at a 90% confidence level
** statistically significant at a 95% confidence level


1601
275.6900
0.0000
0.1762
-644.6571
z
-1.35
0.96
1.14
3.79 **
-3.66 **
0.81
-1.42
-0.43
-1.41
1.61
-1.27
0.41
-0.89
1.06
0.89
0.35
-0.22
3.82 **
-0.05
6.65 **
1.51
2.90 **
4.75 **
1.32
-2.00 **
-0.39
-4.99


Std. Err.
0.2917
0.3063
0.2172
0.0377
0.3480
0.1966
0.3159
0.4829
0.5096
0.0374
0.0003
0.2116
0.3192
0.4635
0.4690
0.0294
0.0003
0.0065
0.1528
0.1776
0.3553
0.2926
0.2439
0.2475
0.0003
0.1642
1.0023


*


*











Table 5-6. Model 2 household labor productivity solving for the endogeneity problem
Model 2 Heckman model Model 2 Heckman model
solving for endogeneity I solving for endogeneity II
Number of obs 1601 Number of obs 1601
Censored obs 894 Censored obs 894
Uncensored obs 707 Uncensored obs 707
Log pseudolikelihood -10449.8 Log pseudolikelihood -10454.9
Wald chi2(36) 55.62 Wald chi2(35) 53.22
Prob > chi2 0.0194 Prob > chi2 0.0249
Robust Robust
prodtotal0 Coef. z Coef. z
Std. Err. Std. Err.


sex
union
children
members
indiglanguage
elementary 1
middle 1
highly
up
aged
aged2
wifeelementary
wifemiddle
wifehigh
wifeup
wifeaged
wifeaged2
income
natmigrationi
ivmig
morethan20
contracted
ejidalland
ownland
plot
irrigation
totalacres
procampo
procede
seed
fertilizer
machinery
south
central
midwest
northeast
cons


30614.49
-38873.74
-8893.97
6220.25
-31123.46
22806.79
62648.85
-64244.12
-93667.19
-4581.28
72.04
-39486.65
-92463.46
24649.61
436487.80
1227.86
-40.64
2625.82
-33925.76
-254718.40
-40070.38
268.11
-24406.25
19368.83
-9743.90
-15370.02
152.96
-34960.07
43710.56
23338.20
47455.76
-40628.53
-50323.44
-49829.31
140533.00
58345.67
147747.30


77062.35
50761.51
14177.55
3844.22
19691.93
26713.46
46126.74
69979.59
74610.05
5406.57
68.66
22918.57
40865.54
42727.38
375889.20
2405.19
27.70
1233.11
19126.52
156318.70
25865.51
157.61
26181.74
12293.48
7171.46
23037.16
174.46
21093.03
26884.89
20566.25
19059.87
24114.23
36786.03
39932.84
108893.50
57551.03
115067.90


0.40
-0.77
-0.63
1.62
-1.58
0.85
1.36
-0.92
-1.26
-0.85
1.05
-1.72 *
-2.26 **
0.58
1.16
0.51
-1.47
2.13 **
-1.77 *
-1.63
-1.55
1.70 *
-0.93
1.58
-1.36
-0.67
0.88
-1.66 *
1.63
1.13
2.49 **
-1.68 *
-1.37
-1.25
1.29
1.01
1.28


26054.65
-33840.62
-5784.87
8177.06
-43080.13
24859.45
57586.65
-66250.80
-104607.80
-3996.13
68.79
-37201.08
-96850.80
31153.82
442094.00
1208.13
-40.43
3078.39
-32735.28
-367465.30

255.67
-24954.02
20135.76
-9307.14
-15953.96
139.56
-34946.04
44168.24
23360.09
45898.68
-40055.73
-38422.72
-34160.81
157611.70
57602.22
116272.10


76060.74
50033.45
13870.25
4114.40
21140.83
26971.08
46411.67
70120.06
76001.90
5391.99
68.49
22761.56
41370.82
44150.11
377044.20
2425.16
27.79
1287.29
18828.50
167205.40

156.01
26340.37
12322.57
7126.42
23147.02
173.10
21098.86
27012.47
20565.21
18802.11
24019.69
35430.63
38298.23
111280.30
57245.49
111967.10


0.34
-0.68
-0.42
1.99 **
-2.04 **
0.92
1.24
-0.94
-1.38
-0.74
1.00
-1.63
-2.34 **
0.71
1.17
0.50
-1.45
2.39 **
-1.74 *
-2.20 **

1.64
-0.95
1.63
-1.31
-0.69
0.81
-1.66 *
1.64
1.14
2.44 **
-1.67 *
-1.08
-0.89
1.42
1.01
1.04


* statistically significant at a 90% confidence level
** statistically significant at a 95% confidence level
ivmzg represents the predicted value of the logit model for migration











Table 5-7. Model 2 Heckman selection model solving for the endogeneity problem (first-stage)


Model 2 Heckman model
solving for endogeneity I
Robust
land Coef. Std. Err. z
sex 0.7874 0.1256 6.27 **
union -0.1714 0.1128 -1.52
children 0.0902 0.0807 1.12
members 0.0315 0.0203 1.55
indiglanguage 0.8543 0.0915 9.34 **
elementary 0.2743 0.0810 3.39 **
middle -0.1762 0.1280 -1.38
highly -0.0280 0.2217 -0.13
up -0.6779 0.2381 -2.85 **
aged 0.0964 0.0146 6.60 **
aged2 -0.0008 0.0001 -5.84 **
wifeelementary 0.1336 0.0802 1.67 *
wifemiddle 0.1692 0.1202 1.41
wifehigh 0.3215 0.2040 1.58
wifeup -0.0420 0.2145 -0.20
wifeaged 0.0111 0.0126 0.88
wifeaged2 0.0000 0.0001 0.39
income 0.0211 0.0036 5.82 **
natmigrationi -0.1139 0.0581 -1.96 **
ivmig -1.9068 0.5881 -3.24 **
morethan20 0.6039 0.1508 4.00 **
south 1.2547 0.1183 10.61 **
central 1.8850 0.1282 14.71 **
midwest 1.0033 0.1567 6.40 **
northeast 0.2095 0.1176 1.78 *
inheritance 0.3520 0.0619 5.69 **
hhfirst -0.3039 0.1310 -2.32 **
perejidallandd 0.4516 0.0690 6.54 **
incomeag 0.6503 0.0598 10.87 **
sharedland 0.3792 0.0632 6.00 **
cons -6.3319 0.4316 -14.67
* statistically significant at a 90% confidence level
** statistically significant at a 95% confidence level
ivmzg represents the predicted value of the logit model for migration


100


Model 2 Heckman model
solving for endogeneity II
Robust
Coef. Std. Err. z
0.8660 0.1219 7.10 **
-0.2408 0.1110 -2.17 **
0.0338 0.0782 0.43
-0.0160 0.0159 -1.00
1.0256 0.0787 13.03 **
0.2237 0.0812 2.76 **
-0.0853 0.1231 -0.69
0.0105 0.2228 0.05
-0.4846 0.2278 -2.13 **
0.0848 0.0142 5.97 **
-0.0007 0.0001 -5.39 **
0.1010 0.0791 1.28
0.2356 0.1209 1.95 *
0.1937 0.1999 0.97
-0.1368 0.2125 -0.64
0.0111 0.0126 0.88
0.0000 0.0001 0.29
0.0125 0.0029 4.37 **
-0.1390 0.0579 -2.40 **
0.1021 0.3009 0.34

1.0922 0.1084 10.07 **
1.6471 0.1084 15.19 **
0.6835 0.1294 5.28 **
0.1462 0.1143 1.28
0.3452 0.0617 5.60 **
-0.3042 0.1295 -2.35 **
0.4784 0.0676 7.08 **
0.6645 0.0598 11.11 **
0.3643 0.0631 5.77 **
-5.7139 0.3849 -14.84









CHAPTER 6
CONCLUSION

This study aimed to capture the impact of international labor-out migration on the

agricultural production of Mexican rural households. Using the Mexican National Rural

Household Survey (ENHRUM) database and estimating a Heckman Two-Stage model we were

able to capture the labor productivity of the household accounting for the selectivity of

landholding.

Our findings suggest that landholding households in rural areas are more likely to have a

low level of education, speak an indigenous language and be located in the central and southern

part of the country compared to non landholding households. The study supports Assies, (2008)

assessment of the existence of regional differences in the agricultural sector. We noticed that

households located in the Southern and Central part of the country are less labor productive by

$71,102.08 and $79,421.25 respectively, than those located in the Northwestern region. This

finding suggests the need to generate incentives to promote the agricultural production in the

southern part of the country as measure to close the existing regional gap in the agricultural

sector.

Another important factor determining the labor productivity of the household is the level

of education. For instance, we found that household heads that have middle education level are

more labor productive by $111,725.80 than household heads that have no school. The education

level of the spouse, however, has the opposite effect on labor productivity. As the education level

of the spouse reaches middle education, decreases the labor productivity of the household head

by $122,025.20 and of the entire household by $84,443.23. These findings suggest the

importance of improving the education level in rural areas. With this result it also became

evident the importance of combining the education level of women with their empowerment in









other markers such as the credit or land rental market as strategy to increase the labor

productivity of the household.

We did not find any change in the intra-household allocation of labor. The amount of

time the household wife and son spend on agriculture does not affect the labor productivity of the

head or of the entire household overall. Studying the role women play in the labor productivity

of the household we found that female-headed households are less likely to own land and work

in agriculture compared to male-headed households. Procampo, which is an income transfer

program, was found to have a negative impact on the labor productivity of the household. It

reduces the labor productivity of the household head by $46,336.40 and of the entire household

by $36,933.66. As Assies (2008) suggests, it seems that the program has been insufficient to help

the rural sector achieve competitiveness.

In our study we also noticed that ejidos remain an important type of tenure of ownership

in Mexico. We found that the existence of ejidos in the community as well as of community land

increases the household's odds of holding land. When measuring the effect of land tenancy status

on labor productivity, we observed no impact on labor productivity. Households participating in

the governmental program, Procede, achieved greater labor productivity of the household head

by $68,562.32. These results support the belief that small farmers, with the assistance of

governmental programs promoting productivity, will have greater incentives to achieve

competitiveness (Quintana, Borquez and Aviles, 1998).

On the other hand, we observed that hired in labor increases the labor productivity of the

household head by $740.03. However, we did not find the same relationship when analyzing the

labor productivity of the entire household. Future studies should measure how the family labor

force and wage workers complement or substitute themselves at the presence of migration in the









agricultural sector. The only capital-intensive resource that seems to increase the labor

productivity of the household is fertilizer. The usage of chemicals increases the labor

productivity of the household head by $54,768.26 and of the entire household by $49,196.56.

Further research on the way the technology adoption of HYV can increase the labor productivity

of the households is recommended.

Other important concept analyzed in this study was the formation of social networks.

Studying those communities where the percentage of migrants is larger than 20 per cent; we

found that households living in these communities are more likely to hold land. Our study,

however, found that the formation of social networks reduces the labor productivity of the

household head by $91,564.94 and of the entire household by $94,257.02. This result suggests

that the existence of social networks in the community can be leading to an overall reduction in

the labor force availability constraining even more the household's capability to substitute or

complement the family labor with waged workers. This finding, however, does not hold when

the potential instrumental variable of distance is added into the model. This can be explained by

the possible correlation existing between the predicted values of the instrumental variable and

the social network variable.

To analyze migration we differentiated between national and international migration.

Households with national migratory experience are less likely to hold land and are also to be less

labor productive. National migration reduces the labor productivity of the head by $46,431.76

and of the household by $33,735.25. In the case of international migration, we observed no

relationship between international migration and access to landholdings. However, we found as

Miluka et.al. (2007) that international labor out-migration has a negative impact on the









household labor productivity. International migration reduces the household labor productivity

by $28,655.31.

As mentioned before, this result suggests that migrant households are less labor

productive than non migrant households. This finding rejects the hypothesis in our study. In

general, it seems that migration, either national or international, reduces the labor productivity of

the entire household. An explanation for this result is that migrant households are not investing

enough in capital-intensive resources to compensate for the reduction in labor supply affecting

negatively the labor productivity of the household.

However, one needs to be cautious in interpreting these results. These findings do not

hold when the potential instrumental variable of distance is added into the model. Furthermore,

the coefficient of the predicted value is significantly greater than the coefficient of the original

migration variable, suggesting a greater impact of international migration on labor productivity.

In order to make better inferences on the way labor out-migration impacts labor productivity,

further research needs to be carried out, dealing more properly with the endogeneity problem of

migration.

The contribution of this study to the current literature can be summarized in three main

points. First, we tested the New Economics of Labor Migration approach and proved that

migration represents indeed a household strategy. Contrasting the way labor out-migration

influences the labor productivity of rural households we found that international migration

affects labor productivity at the household level but not at the household head level. Second,

taking into account landholding selectivity we reached the same result found in the case of

Albania (Miluka, et.al, 2007), that international migration is affecting negatively the labor

productivity of the household. Third, when we introduced the social network variable into the


104









model, we reaffirmed the finding that migration has a negative impact on labor productivity in

the agricultural sector.

Limitations of this research are three. First, the database presents some drawbacks to the

analysis as the fact that it is cross sectional data, it does not provide information on return

migration or duration of trips, and some of the farming practices were aggregated at the

household level instead of the desired parcel/plot level. Second, the study of labor productivity

and migration require complex methods that are beyond the scope of this research. Third, the

utilization of instrumental variables represents a challenge in the research since no econometric

method can guarantee that the selected instrumental variable is the correct one to solve for the

endogeneity problem.

Future research will include finding instrumental variables that are correlated with

migration but not with the formation of social networks to be able to account for the impact of

both, migration and social networks on labor productivity. We will also try to develop more

complex methods that are able to account for a common endogenous regressor in the selection

and regression equation in the sample selection model (Kim, 2006).









LIST OF REFERENCES


Angelucci, M. 2005. "U.S. Border Enforcement and the Net Flow of Mexican Illegal Migration."
Working Paper No. 1642, IZA.

Araujo, C. 2004. "Can Non-Agricultural Employment Reduce Rural Poverty? Evidence from
Mexico." Cuaderno de Economia 41:383-399.

Araujo, C., A. de Janvry, and E. Sadoulet. 2002. "Spatial pattern of non-agricultural employment
growth in rural Mexico during the 90s," Dept. Agr. Res. Econ, University of California,
Berkeley.

Assies, W. 2008. "Land Tenure and Tenure Regimes in Mexico: An Overview." Journal of
Agrarian Change 8(1):33-63.

Borjas, G. J. 1984. "The Labor Supply of Male Hispanic Immigrants in the United States."
International Migration Review 17(4):653-671.

Caglar, A. 2006. "Hometown associations, the rescaling of state spatiality and migrant grassroots
transnationalism." Global Networks 6 (1):1-22.

Carrion-Flores, C. E. 2007. "What Makes You Go Back Home? Determinants of the Duration of
Migration of Mexican Immigrants in the United States." Unpublished, University of
Arizona.

Castles, S. 2002. "Migration and Community Formation under Conditions of Globalization."
International Migration Review 36(4): 1143-1168.

Chami, R., C. Fullenkamp, and S. Jahjah. 2005. "Are immigrant Remittance Flows a Source of
Capital for Development?" International Monetary Fund 52(1).

Chiswick, B. R., and T. J. Hatton. 2003. "International Migration and the Integration of Labor
Markets." M. D. Bordo, A. M. Taylor, and J. G. Williamson, ed. Chicago/London,
University of Chicago Press, pp. 65-117.

Christensen, L. R. 1975. "Concepts and Measurement of Agricultural Productivity." American
Journal ofAgricultural Economics 57(5):910-015.

Cord, L., and Q. Wodon. 2001. "Do Agricultural Programs in Mexico Alleviate Poverty?
Evidence from the Ejido Sector." Cuaderno Economico 38(114).

Cornelius, W. A. 2001. "Death at the Border: Efficacy and Unintended Consequences of US
Immigration Control Policy." Population and Development Review 27(4):661-685.

Cornelius, W. A., and P. L. Martin. 1993. "The Uncertain Connection:Free Trade and Rural
Mexican Migration to the United States." International Migration Review 27(3):484-512.









Davis, B. 2000. "The adjustment strategies of Mexican ejidatarios in the face of neoliberal
reform." CEPAL Review.

Davis, B., G. Stecklov, and P. Winters. 2001. "Disaggregating Mexican migrant networks: The
parts are greater than the whole." University of New England, Graduate School of
Agricultural and Resource Economics & School of Economics, pp. 1-28.

Deere, C. D., and M. Leon. 2003. "The gender asset gap: Land in Latin America." World
Development 31(6):925-947.

De Janvry, A., and E. Sadoulet. 2001. "Income Strategies Among Rural Households in Mexico:
The Role of Off-farm Activities." World Development 29(3):467-480.

Donato, K. M. 1999. "A Dynamic View of Mexican Migration to the United States." Gender
Issues.

Donato, K. M. 1994. "U.S. Policy and Mexican Migration to the United States, 1942-92." Social
Science Quarterly 75(4).

Doss, C., and M. Morris. 2001. "How does gender affect the adoption of agricultural
innovations? The case of improved maize technology in Ghana." Agricultural Economics
25:27-39.

Dovring, F. 1979. "Gross and Net Productivity: A Problem of Aggregation." American Journal
ofAgricultural Economics 61(4):694-696.

Fernandez-Cornejo, J., and C. R. Shumway. 1997. "Research and Productivity in Mexican
Agriculture." American Journal ofAgricultural Economics 79(3):738-753.

Fernandez-Kelly, P. and D. S. Massey. 2007. "Borders for Whom? The Role of NAFTA in
Mexico-U.S. Migration." The ANNALS of the American Academy of Political and Social
Science 2007 610(98).

Gujarati, N. 2003. "Basic Econometrics." Fourth ed. Mc Graw Hill.

Hanson, G. H., and A. Spilimbergo. 1999. "Illegal immigration, border enforcement, and relative
wages: evidence from apprehensions at the US-Mexico Border." The American Economic
Review 89(5):1337-1357.

Hashida, E., and J. M. Perloff 1996. "Duration of Agricultural Employment." California
Agricultural Experiment Station.

Hayes, J., M. Roth, and L. Zepeda. 1997. "Tenure Security, Investment and Productivity in
Gambian Agriculture: A Generalized Probit Analysis." American Journal ofAgricultural
Economics 79(2):369-382.









Heckman, J. J.1979. "Sample Selection Bias as a Specification Error" Econometrica 47(1):153-
160.

Holden, S., B. Shiferaw, and J. Pender. 2001. Market Imperfections and Land Productivity in
the Ethiopian Highlands." International Food Policy Research Institute.

Iwai, N., R. Emerson, and L. M. Walters. 2006. "A markov chain analysis with self-selectivity."
American Agricultural Economics Association Annual Meeting, Long Beach, California.

Iwai, N., 0. Napasintuwong, and R. Emerson (2005) Immigration Polity and the Agricultural
Labor Market: The Effect on Job Duration International Agricultural Trade and Policy
Center.

Jacoby, S. H. 1992. "Productivity of Men and Women and the Sexual Division of Labor in
Peasant Agriculture of the Peruvian." Journal of Development Economics 37:265-287.


Jansen, H., J. Pender, A. Damon, and R. Schipper. 2006. "Land management decisions and
agricultural productivity in the hillsides of Honduras." International Association of
Agricultural Economists Conference, Gold Coast, Australia, International Food Policy
Research Institute (IFPRI), USA

Johnson, N. L. 2001. ""Tierra y Libertad": Will Tenure Reform Improve Productivity in
Mexico[s "Ejido" Agriculture?" Economic Development and Cultural change 49(2):291-
309.

Jones, R. 1995. "Immigration Reform and Migrant Flows: Compositional and Spatial Changes in
Mexican Migration after the Immigration Reform Act of 1986." Annals of the
Association ofAmerican Geographers 85(4):715-730.

Kim, K. i. 2006. "Sample selection models with a common dummy endogenous regressor in
simultaneous equations: A simple two-step estimation." Economics Letters 91:280-286.

Kimhi, A. 2003. "Plot size and maize productivity in Zambia: the inverse relationship re-
examined." The Hebrew University of Jerusalem, pp. 26.

Lastarria-Cornhiel, S. 1988. "Female farmers and agricultural production in El Salvador"
Development and Change 19(4):585-616.

Mabogunje, A. L. 1989. "Agrarian Responses to Outmigration in Sub-Saharan Africa."
Population and Development Review 15 (Supplement: Rural Development and
Population Institutions and Policy): 324-342.

Massey, D. S., and K. E. Espinosa. 1997. "What's Driving Mexico-U.S. Migration? A
Theoretical, Empirical and Policy Analysis." The American Journal of Sociology
102(4):939-999.









Massey, D. s., L. Goldring, and J. Durand. 1994. "Continuities in Transnational Migration: An
Analysis of Nineteen Mexican Communities." American Journal of Sociology
99(6):1492-1533.

Masterson, T. 2005. "Land Markets, Female Land Rights and Agricultural Productivity in
Paraguayan Agriculture." University of Massachusetts Amherst.

McKenzie, D., and H. Rapoport. 2004. "Network Effects and the Dynamics of Migration and
Inequality: Theory and Evidence from Mexico Mimeo." Stanford University.

Mendola, M. 2008. "Migration and technological change in rural households: Complements or
substitutes?" Journal of Development Economics 85:150-175.

Miluka, J., G. Carletto, B. Davis, and A. Zezza. 2007. "The Vanishing Farms? The Impact of
International Migration on Albanian Family Farming." Barcelona, Spain, ESA Working
Paper No. 07-09.

Mochebelele, M., and A. Winter-Nelson. 2000. "Migrant Labor and Farm Technical Efficiency
in Lesotho." World Development 28(1):143-153.

Moretti, E. 1999. "Social Networks and Migrations: Italy 1876-1913." International Migration
Review 33(3):640-657.

Nonthakot, P., and R. Villano. 2008. "Migration and Farm Efficiency: Evidence from Northern
Thailand." AARES 52nd Annual Conference.

Nonthakot, P., and R. Emerson. 2005. Farm Mechanization and the Farm Labor Market: A
socioeconomic Model of Induced Innovation." International Agricultural Trade and
Policy Center.

Orozco, M., and M. Lapointe. 2004. "Mexican Hometown Associations and Development
Opportunities" Journal of International Affairs 57(2).

Orrenius, P. M. and M. Zavodny. 2005. "Self-selection among undocumented immigrants from
Mexico." Journal of Development Economics 78:215-240.

Orrenius, P. M. 1999. "The role of family networks, coyote prices and the rural economy in
Western Mexico: 1965-1994, Federal Reserve Bank of Dallas.

Ortega-Sanchez, I., and J. Findeis. 2001. "Production Impacts of Labor Outmigration in Agrarian
Economies: The case of corn farmers in central and southern areas in Mexico."
American Agricultural Economics Association Annual Meeting.

Pender, J., E. Nkonya, P. Jagger, D. Sserunkuuma, and H. Ssali. 2003. Strategies to Increase
Agricultural Productivity and Reduce Land Degradation: Evidence from Uganda." 25th
International Conference of Agricultural Economists, Durban, South Africa.


109









Portes, A., C. Escobar, and A. W. Radford. 2007. "Immigrant Transnational Organizations and
Development: A Comparative Study" International Migration Review 41(1):242-281.

Quintana, R. D., L. C. Borquez, and R. P. Aviles. 1998. "Peasant Logic, Agrarian Policy,, Land
Mobility, and Land Markets in Mexico." Land Tenure Center, University of Wisconsin-
Madison.

Quisumbing, A. 1996. "Male-female differences in agricultural productivity: Methodological
issues and empirical evidence." World Development 24(10): 1579-1595.

Rozelle, S., E. Taylor, and A. DeBrauw. 1999. "Migration, Remittances, and Agricultural
Productivity in China" AEA Papers andProceedings pp. 287-291.

Rytina, N. 2002. "IRCA Legalization Effects: Lawful Permanent Residence and Naturalization
through 2001.The Effects of Immigrant Legalization Programs on the United States:
Scientific evidence on immigrant adaptation and impacts on U.S. economy and society."
The Cloister, Mary Woodward Lasker Center, NIH Main Campus.

Stark, 0. 1991. "The Migration of Labour." Cambridge, Basil Blackwell.

Stark, 0., and D. E. B. May. 1985. "The New Economics of Labor Migration." AEA Papers and
Proceedings.

Taylor, J. E., and A. Lopez-Feldman. 2007. "Does Migration Make Rural Households More
Productive? Evidence from Mexico." Working Paper, ESA.

Taylor, J. E., G. A. Dyer, and A. Yunez-Naude. 2005. "Disaggregated Rural Economy wide
Models for Policy Analysis." WorldDevelopment 33(10): 1671-1688.

Taylor, J. E., A. Yunez-Naude, and G. Dyer. 1999. "Agricultural Price, Employment, and
Migration in a Diversified Rural Economy: A Village-Town CGE Analysis from
Mexico." American Journal ofAgricultural Economics 81(3):653-662.

Tittonell, P., K.D. Shepherd, B. Vanlauwe, and K.E. Giller. 2007. "Unravelling the effects of
soil and crop management on maize productivity in smallholder agricultural systems of
western Kenya -An application of classification and regression tree analysis."
Agriculture, Ecosystems and Environment 123:137-150.

Todaro, M. P. 1980. "International Migration in Developing Countries: A Survey." R. A.
Easterlin, ed. Chicago, University of Chicago Press, pp. 361-404.

Udry, C. 1996. "Gender, agricultural production, and the theory of the household." The Journal
ofPolitical Economy 104(5): 1010-1046.

Vertovec, S. 2004. "Migrant Transnationalism and Modes of Transformation." International
Migration Review 38(3):970-1001.


110










Villa-Issa, M. R. 1990. "Performance of Mexican Agriculture: The Effects of Economic and
Agricultural Policies." American Journal ofAgricultural Economics 72(3):744-748.

Wilson, T. D. 1998. "Weak ties, strong ties: Network principles in Mexican migration." Human
Organization 57(4):394-403.

Wood, S., L. You, and X. Zhang. 2004. "Spatial Patterns of Crop Yields in Latin America and
the Caribbean." Cuadernos de Economia 41:361-381.

Yunez-Naude, A. 2003. "The Dismantling of CONASUPO, a Mexican State Trade." Agriculture
Colegio de Mexico.

Yunez-Naude, A., and F. Barceinas. 2000. "Efectos de la Desaparici6n de la CONASUPO en el
Comercio y en los Precios de los Cultivos Basicos." Estudios Econ6micos July-Dec: 189-
227.









BIOGRAPHICAL SKETCH

Melissa Ramirez was born in 1981, in Puebla, Mexico. The younger of two children, she

grew up in Puebla, graduating from the American School High School in 2000. During her

undergraduate studies, she had the opportunity to participate in a 1-year exchange program in

Tubingen, Germany. She earned her B.S. in economics from the Universidad de las Americas,

Puebla. In 2006 she gained an assistantship to come to the University of Florida to continue her

studies. Melissa graduated in summer 2008 with a Master of Science in food and resource

economics and a certificate in supply chain management.





PAGE 1

HOW DOES MIGRATION AFFECT AGR ICULTURAL LABOR PRODUCTIVITY? THE CASE OF MEXICAN RURAL HOUSEHOLDS By MELISSA A. RAMIREZ RODRIGUES A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE UNIVERSITY OF FLORIDA 2008 1

PAGE 2

2008 Melissa A. Ramirez Rodrigues 2

PAGE 3

To all those who follow their dreams 3

PAGE 4

ACKNOWLEDGMENTS My deepest gratitude and love go to my ment ors in life: my beloved parents, Agustin Ramirez and Aidil Rodrigues; and my dearest sister, Milena. Their wisdom, dedication and unconditional love helped me achieve everything I have in life and helped me fulfill my dreams. I would also like to thank my chair, Dr. Carmen Carrion-Flores; and my supervisory committee, Dr. Carmen Deere for the support and guidance that made all this work possible. 4

PAGE 5

TABLE OF CONTENTS page COPYRIGHT ...................................................................................................................................2 DEDICATION .................................................................................................................................3 ACKNOWLEDGMENTS ...............................................................................................................4 LIST OF TABLES ...........................................................................................................................7 LIST OF FIGURES .........................................................................................................................8 ABSTRACT .....................................................................................................................................9 CHAPTER 1 INTRODUCTION ..................................................................................................................10 2 OVERVIEW OF THE AGRICULTURAL SECTOR IN MEXICO ......................................13 Introduction .............................................................................................................................13 Review of Agricultural Policies in Mexico ............................................................................13 Land Reforms in Mexico .................................................................................................14 Domestic Market Intervention .........................................................................................16 International Trade Regulation ........................................................................................19 Implications for the Agricultural Sector .................................................................................21 Conclusion ..............................................................................................................................28 3 MIGRATION AND AGRICULTURAL PRODUCTIVITY .................................................29 Introduction .............................................................................................................................29 Mexican Migration Literature .................................................................................................30 Immigration Reforms in the United States ......................................................................30 The Bracero program ...............................................................................................30 Immigration Reform and Control Act (IRCA) .........................................................31 Border enforcement ..................................................................................................32 The Evolution of Mexican Migration ..............................................................................32 Migratory patterns ....................................................................................................32 Migratory flows ........................................................................................................35 Causes of Migration ........................................................................................................37 Impact of Migration in the Sending Communities ..........................................................38 Remittances ..............................................................................................................39 Hometown associations ............................................................................................41 Labor Out-Migration and the Agricultural Productivity Literature ........................................42 Changes in Farming Practices and Decisions ..................................................................44 Gender Productivity .........................................................................................................49 5

PAGE 6

Conclusion ..............................................................................................................................51 4 DATA ANALYSIS AND METHODOLOGY .......................................................................52 Introduction .............................................................................................................................52 Data and Descriptive Statistics ...............................................................................................53 The Mexican National Rural Household Survey (ENHRUM) ........................................53 Sample Description .........................................................................................................55 Landholder and non landholder mean differences ...................................................57 The landholder sample .............................................................................................59 Migrant and non migrant mean differences .............................................................61 Description of landholding households accounting for migratory status .................63 Heckman Two-Stage Procedure .............................................................................................65 Variables Description .............................................................................................................67 Dependent Variables .......................................................................................................67 Independent Variables .....................................................................................................68 Endogeneity Issues .................................................................................................................71 Conclusion ..............................................................................................................................73 5 RESULTS ...............................................................................................................................84 Introduction .............................................................................................................................84 Estimation of OLS ..................................................................................................................84 Heckman Two-Stage Estimation ............................................................................................86 Who holds Land? .............................................................................................................86 What affects Labor Productivity? ....................................................................................88 Model 1 household head labor productivity (sample A) ..........................................88 Model 2 household labor productivity (sample B) ...................................................90 Addressing Migration Endogeneity ........................................................................................91 Conclusion ..............................................................................................................................93 6 CONCLUSION .....................................................................................................................101 LIST OF REFERENCHES ..........................................................................................................106 BIOGRAPHICAL SKETCH .......................................................................................................112 6

PAGE 7

LIST OF TABLES Table page 4-1 The ENHRUM community codes ......................................................................................75 4-2 Landholder and non landholder mean differences .............................................................77 4-3 Migrant and non migrant mean differences .......................................................................78 4-4 Migrant landholder and non mi grant landholder mean differences ...................................79 4-5 Selection equation variables statistics ................................................................................80 4-6 Regression equati on variables statistics .............................................................................81 5-1 Model 1 household head labor productivity. .....................................................................94 5-2 Model 1 Heckman selection model (first-stage) ..............................................................95 5-3 Model 2 household labor productivity ..............................................................................96 5-4 Model 2 Heckman selection model (first-stage) ..............................................................97 5-5 Logit model for migration .................................................................................................98 5-6 Model 2 household labor productivity solving for the endogeneity problem ...................99 5-7 Model 2 Heckman selection model solving for the endogeneity problem (first-stage) ..100 7

PAGE 8

LIST OF FIGURES Figure page 4-1 Age of the household head .................................................................................................82 4-2 Education level of the household head ..............................................................................82 4-3 Number of members in the household ...............................................................................82 4-4 Comparison of education level by landholding status .......................................................83 4-5 Landholder and non landholder spatial distribution ..........................................................83 4-6 Migrant and non migrant spatial distribution .....................................................................83 8

PAGE 9

Abstract of Thesis Presen ted to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Science HOW DOES MIGRATION AFFECT AGRIC ULTURAL LABOR PRODUCTIVITY? THE CASE OF MEXICAN RURAL HOUSEHOLDS By Melissa A. Ramirez Rodrigues August 2008 Chair: Carmen Carrion-Flores Major: Food and Resource Economics Using micro-level agricultural data from the ENHRUM survey, I examine the impact of international labor-out migration on the agricu ltural production of Mexi can rural households. The study evaluates how households reallocate la bor and capital resources as consequence of labor out-migration and incorporates a productivity variable to measure the efficiency of this reallocation. Estimating a Heckman Two-Stage model we capture the labor productivity of the household accounting for the selec tivity of landholding. The results suggest that international labor-out migration and the formation of soci al networks have a negative impact on the household labor productivity. Migrant households are less labor productive than households with no migratory experience by 28,655.31 Mexican pesos. It seems that migrant households are not investing enough in capital-intensive resources to compensate for the reduction in labor supply. Changes in the intra-household allocation of labo r are not observed. The e ducation level of the household head and spouse, the tenancy status of the land and the locatio n of the household are other factors affecting the labor productivity of the household. 9

PAGE 10

CHAPTER 1 INTRODUCTION Migration is increasingly bei ng considered in the literature on agricultural productivity in developing countries, it has become a common practice for rural households worldwide. Migration has been found to affect a household s decisions in three important ways. First, migration reduces the labor availa bility of the household; second, it generates an increase in the households income through remittances sent by the migrant; and third, it strengthens the formation of social networks, which can be used to promote the migration of other household members. In the case of Mexico, migration is one of the off-farm activities that rural households rely upon heavily. According to the Consejo Nacional de Poblacin (CONAPO), the number of Mexicans engaging in a migratory experience to U.S. reached 3.3 million between 1990 and 2000. Furthermore the composition of the annual net flow to this country has increased by a factor of three in the last three decades, l eading to the formation of a Mexican migrant community in the U.S. that reached 26.7 million in 2003.1 Thus far the studies on migration have not reached a consensus on the way migration influences the farming practices and decisions of the household. However, it has been noticed that the households initial endowment s as well as the type of migr ation lead to different effects of labor out-migration on the sending communit y. Furthermore, it has been found that, in general, remittances are used to relax credit constraints and improve the farm management practices of the household. Analysis of the way in which Mexican in ternational migration affects the labor productivity in rural households is scarce. Exis ting studies suggest farm ing differences between 1 Approximately 9.9 million people represent the migrants that were born in Mexico and the remaining 16.8 millions represent the population already born in the United States but with Mexican heritage. 10

PAGE 11

migrant and non migrant households depend not only on the househol ds initial endowments but also on how labor-intensive the farming practi ces of the households ar e, and the households ability to substitute the family la bor with reciprocal or wage labor i.e.; it depends upon how rural labor markets work. This study central research question is the impact of internatio nal migration on the household labor productivity. The study aims to ev aluate the reallocation of labor and capital resources as a consequence of labor out-migr ation and measure the efficiency of this reallocation. My primary hypothesi s is that, as a household stra tegy to manage uncertainty and market imperfections, migrant households main tain their agricultura l production levels by investing more in capital-intens ive inputs to compensate for the reduced labor force availability due to the migration of at least one of the hous ehold members. My corollary hypothesis is that labor productivity, measured as the agricultural output generated per day of work, will be greater in migrant households compared to non migrant households. To test th ese hypotheses, my study employs econometric techniques using the Mexican National Rural Household Survey (ENHRUM), a nationally representative sa mple of 1765 household co-directed by the Colegio de Mxico (COLMEX) and the University of California at Davis in 2002/2003. The contribution our study to the existing litera ture focuses on three main points. First, we rely on the New Economics of Labor Migr ation (NELM) approach, using the household as the unit of analysis to study the way labor out-migration influences the labor productivity of rural households. Second, we estimate labor productivity accounting for the selectivity of whether or not a household has access to land. The idea behind th is is that agricultural productivity can only be measured for those households holdi ng land, and until now no study recognized this 11

PAGE 12

selectivity when studying labor productivity. Finally, we introdu ce into the labor productivity analysis the study of social networks. 12

PAGE 13

CHAPTER 2 OVERVIEW OF THE AGRICULTURAL SECTOR IN MEXICO Introduction The purpose of this chapter is to review the status of the agricultural sector on the past three decades. A review of past and current ag ricultural policies in Mexico is crucial in understanding the way these changes have curren tly influenced the agricultural sector. These agricultural policies have directly or indirectly changed farmers agricultural practices affecting how the agricultural sector operates as well as its productivity This chapter is structured as follows. The first section summarizes the domestic and international agricultural policies institutionalized in Mexico during the twentieth century. Th e second section reviews the implications of the policies for the agricultural se ctor. Finally, some conclusions are presented in the third and last section. Review of Agricultural Policies in Mexico At the beginning of the twentieth century ag riculture employed an important share of the labor force but most of the agricultural wo rkers were landless (Fernandez-Cornejo and Shumway, 1997; Villa-Issa, 1990). The concentra tion of land in the ha nds of a few and the inequalities among social classes were two factors leading to the Mexi can Revolution (1910-17). After the Revolution and during the 1920s the agricu ltural sector received little or no investment. It was not until the 1930s that there was a substantial increase in public investment in the agricultural sector, such as the cons truction of roads, irrigation systems and the intensification of the land reform. From the 1930s until the 1980s the government played a key role in the development of the sector, for instance, with the creation of the ejido and the institutionalization of CONA SUPO (Yunez-Naude, 2003; Yun ez-Naude and Barceinas, 2000). The next sub-sections describe these in detail. 13

PAGE 14

Land Reforms in Mexico Following the Mexican Revolu tion, the first land reform2 was particularly important because it not only reallocated the possession of the land, but also set the foundation for the contemporary agrarian system. In essence, this reform encompassed a new system of tenancy called the ejido system, which consisted in communal land possession but generally individual farming. The ejido was made up of the ejidatarios who are the farmers who have rights to the ejido land (called agrarian reform ri ghts). This type of ownership ha s remained effective until the present. During the government of Lzaro Crdenas (1934-1940) there was a large-scale redistribution of land. By 1940, the ejido sector possessed 22.5 per cen t of the agricultural land and 47.4 per cent of the arable land of the count ry (Assies, 2008). Two additional presidential periods characterized by important redistribution of land were those of Gustavo Diaz Ordaz (1964-1970) and Luis Echeverria (1970-1976). During these two periods, however, no significant amount of irrigate d land was redistributed. There were three major restrictions imposed on the way the ejido operated. First, there was a labor restriction, where the ejidatarios were not allowed to hire labor. Second, if the ejidatarios resided away from their allo cated land for more than two years, they ran the risk of losing their ejidal rights. Moreover, within this syst em, long-term production contracts with farmers outside the ejido were not allowed (Johnson, 2001). In practice, however these restrictions were not always followe d. For instance, i llegal renting of ejido land to commercial farms was a common practice among farmer s as was migration (Assies, 2008). With the creation of ejidos the government aimed to promot e productivity and satisfy the internal market for agricultural products. In its early stage, the land agrarian reform was backed 2 The first land reform consisted in the 1917 amendment of Article 27 of the Constitution. 14

PAGE 15

up by technical assistance, credit and supply of s eeds. However, with the instauration of an import-substitution industrializati on model in the early 1930s, the policy gradually shifted away from the agrarian sector toward the industrial sector favoring the provisio n of cheap food for an increasingly industrializing country (Assies, 2008). The social sector consisting of ejidos and agrarian communities was thus confined to the production of staples under pr ice regulation and subsidies. At the same time, however, policies promoting the investment in irrigation systems and capital-intensive production favored the development of the private sector and the pro duction of high value exports, giving rise to the formation of a dual agrarian structure and a deep ening in the existing regi onal differences that persists until nowadays (Assies, 2008). By the early 1990s the ejido system accounted for approximately 100 million hectares entailing half of the national farmland (Fer nandez-Cornejo and Shumway, 1997). Furthermore, the land was distributed to nearly 3 million peasants that represented about three quarters of total producers, grouped in 26,796 ejidos and 2,366 agrarian communities (Quintana, Borquez and Aviles, 1998).A typical ejido would consist of approximately 74 ejidatarios and possess some 2,000 hectares The average ejidatario would hold 9.2 hectares in tw o parcels and have access to 28 hectares of common land (Assies, 2008). One of the most radical policies institutiona lized in Mexico during the 1990s was the second land reform or counter reform. The 1992 am endment of Article 27 of the Constitution put an end to the land redistribution process existe nt in the country since the 1930s. This reform aimed to transform the collective possession of land into an individual possession, setting the conditions necessary to start the privatization pr ocess of land. It also laid the foundations for trade liberalization of the agricultural sector (Fernandez Co rnejo and Shumway, 1997). It was 15

PAGE 16

believed the reform was going to help to overcome the crisis in the sect or through the expansion consolidation of rental markets, increased productivity and the promotion of foreign investment (Assies, 2008). It has been noticed however, that in a context of globa lization and asymmetric free-trade relations (Assies, 2008, pg 33) th e agrarian crisis ha s only intensified. Under the 1992 reform, the ejidatarios were granted the opportuni ty to certify their land rights if the ejido consented to participate in Procede .3 They were also allowed to hire labor and grow any crop and market it wherever they wanted. Long-term production contracts with outsiders were made also feasib le. In a general way, this reform reintroduced a market oriented scheme into the agricultural sector, allowing farmer s to respond directly to market incentives and disincentives. One important feature of Procede is that the decision to part icipate in this certification program not necessarily resulted in the privatization of the ejido. A governmental report capturing information for 1992-2005, for example, show ed that in total only 1% of the social property entering the certificati on process achieved full private property status in this time period. Moreover, 60% of this pr ivatization was done for urbaniza tion purposes (Assies, 2008). This number clearly suggests that one of the main goals of the re form, to start the privatization process of the land, in order to capitalize the ejidos has not been achieved. Domestic Market Intervention The Mexican government has always played an active role in regulating the agricultural domestic market. For instance, in 1965 the govern ment creased a state trading enterprise (STE) called The National Company of Popular Subsistence (CONASUP O). The organizations main 3 Ejidatarios participating in PROCEDE had the right to legally sell, rent, sharecrop or mortgage their land. The decision to sell ejido land to outsiders, however, required the approval of two-thirds of the ejido general assembly, witnessed by a government representative. 16

PAGE 17

objective consisted in promoti ng the domestic market by s ubsidizing both producers and consumers and regulating international trad e through direct imports (Yunez-Naude 2003). In its first stages, CONASUPO was designed as an economic development tool to protect small staple-farmers as well as low-income consumers. To protect producers, CONASUPO absorbed the transaction costs farmers faced by reducing the number of intermediaries involved in purchase-sale transactions; it also guaranteed crop support prices4. It also promoted production subsidies, including inpu t subsidies for water, electric ity and fertilizers (FernandezCornejo and Shumway, 1997). CONASUPO also ma naged subsidiary programs for processi storing, distributing and selli ng the crops. At some point, CONASUPO exerted control over 30% of the total gross domestic agricultural production (Yunez-Naude, 2003). The most important crop, however, was corn, representing 56% of the total value of crops managed by CONASUPO (Yunez-Naude and Barcei ng, nas, 2000). 5 By the end of the 1980s some of the task s performed by CONASUPO began to decline; and by mid 1990s most of CONASUPOs programs were already dismantled6, privatized or transferred to farmers. For inst ance, the processing of corn was privatized and the processing of wheat to make bread was ended. In addition, the warehouses for basic crop storage belonging to CONASUPO were transferred to farmers and local authorities.7 Finally, price intervention was reduced to just corn and beans. Corn and beans were the last two staple cr ops administered by CONASUPO since these two crops, representing Mexicos two major stap le crops produced by the larger number of 4 The agricultural crops involved in the CONASUPOs pr ograms were barley, beans, copra, corn, cotton, rice, sesame, sorghum, soybeans, sunflower and wheat. 5 To help low income consumers, CONASUPO sold basic foods to rural and urban costumers at very low prices. Some of these goods includ ed: corn, flour, wheat pasta, edible oils and fluid milk (Yunez-Naude, 2003). 6 The only two entities that survived this dismantling proc ess were the LICONSA entity in charge of processing milk powder to produce fluid milk for access to the poor at subsidized prices; and the retail store DICONSA responsible for distributing basic food to low-income consumers at low prices. 7 Also, one of the extension programs called CECONCA, used for technical supports to farmers was also abolished. 17

PAGE 18

peasant households, required a longer transf ormation period. By th e end of 1995, CONASUPO was still a last resort buyer of corn and beans at minimum prices. It wa s also in charge of regulating the external trade of both crops. In 1998 howev er, CONASUPOs involvement in social programs to assist the poor was ending, un dermining the main reason for the existence of the company. CONASUPO was subsequently liqui dated during the Zedill o administration (19952000) (Yunez-Naude, 2003). In 1991 a new agency was created by the Ag ricultural Ministry cal led Support Services for Agricultural Marketing (ASERCA) 8. This agency emerged as a substitute for CONASUPO although ASERCA has no mandate with respect to price fixing commodity imports. Some of the tasks this agency carries out in clude marketing and the coordinati on of direct income transfer programs. Between 1992 and 1996 and under the supervision of ASERCA, two programs were developed, the Program of Direct Payments to the Countryside ( Procampo) 9 and Alliance for the Countryside. The goal of th ese two programs was to support agricultural producers witho interfering with the new rural market economy. Although the goal of these two programs is the same, they differ in the way they are managed and funded. Alliance for the Countryside is statemanaged while Procampo is managed by the federal government. Moreover, a portion of the Alliance for the Countryside program is funded using farmers resources. ut The Program of direct Payment to the Countryside (Procampo) replaced the traditional price support system by an income direct payment for farmers based on the number of acres devoted to the production of maize, beans, whea t, rice, cotton, soybeans, safflower, barley, and sorghum. The Alliance for the Countryside provide d farmers with financial aid, technical and 8 ASERCA stands in Spanish for Apoyo y Servicio a la Comercializacin Agropecuaria 9 This program was expected to last 15 years ending in 2008. 18

PAGE 19

marketing assistance and training. In essence, its objectives were to: (1) increase the investment in capital intensity technology; (2) support the transfor mation of agriculture toward areas with comparative advantage; (3) promote the creati on of distribution channels for the products commercialization (C ord and Wodon, 2001). With the elimination of CONASUPO along with the creation of Procampo and Alliance for the Countryside, Mexico was laying the f oundations for the trade liberalization of the agricultural sector. With these changes as well as other market oriented policies, Mexico was preparing itself to enter into the General Agre ement on Tariffs and Trade (GATT) as well as the North American Trade Agreement (NAFTA)10. International Trade Regulation International trade for agricultural products was also regula ted heavily by the Mexican government. For instance, CONASUPO also had an active role regulati ng international trade through direct imports in the early-mid 1980s11; but just as with the domestic market, CONASUPOs participati on in trade regulation th rough the direct import of basic crops began to decrease considerably in the following years.12 The government also controlled trade volumes, imposing tariffs, quotes and licensing requirements. Other policies (outside the agricultural sector ) that influenced the sectors performance were exchange rates policies, investment policy in the rural sector, as well as state investment in infrastructure, transportation a nd communication (V illa-Issa, 1998).13 10 Fernandez-Cornejo, 1997 11 CONASUPO accoun ted for 95% of total rice imports, 83% of corn imports and 68% of wheat import in 1983-88 period; for 99% of beans import from 1989 until 1993; and more than 95% of total sorghum and soybean imports at the beginning of the 1970s (Yunez-Naude and Barceinas, 2000). 12 For example, the rice imported by CONASUPO reduced from 25% in 1989-1993 to zero in 1994-1996. Also, its corn imports declined from 38% in 1989-93 to 16% 1994-96. And CONASUPO direct imports of beans and wheat reached cero by the period 1994-96 (Yunez -Naude, 2003) 13 For instance, from 1955 to 1972 the scarce amount of private investments in the rural sector, due to a lag of 19% in the farm prices, was offset by public investment. Duri ng that time, the exchange rate was also overvalued 19

PAGE 20

In the mid 1980s Mexico started a series of adjustments to the ex isting economic model. In 1986 Mexico became a member of the General Agreement on Tariffs and Trade (GATT) and in 1994 Mexico was admitted into NAFTA. The ad mittance of Mexico represented an important step toward a market-oriented strategy and cons olidated many of the structural changes that began in the early 1980s. The most important structural changes included the substitution of the import substitution model by a market oriented m odel with a diminished participation of the government. An important trade policy change was the shift from import licenses to tariff rate quotas (TRQs).14 The North American Free Trade Agreement (NAFTA) included two separate agreements, one between Mexico and the United States and the other one between Me xico and Canada. It was agreed that import levels below the consen ted quota would not be subject to tariffs. A 15 year period (1994-2008) was set to eliminate the over quota tariffs for corn, dry beans and milk powder (milk powder was not negotiated between Mexico and Canada). Mexico has also signed other FTAs with La tin American and European countries. After its incorporation as a full member of the WTO in 1995, Mexico agreed during the Uruguay Round to set a tariff base of 25% for almost all agricultural products, with the promise of reducing it an additional 1% by 2000. The basi c crops subject to TRQs in the NAFTA negotiation were also kept valid in this negotiation adding wheat to this type of trade regulation. Canada and United States however have larger quota access and lower above quota tariffs. In summary, this overview of the Mexican policy re forms is a starting point to understand the agricultural sector. The next section focuses in explaining the implications for the agricultural sector. 14 The crops that were changed from import licenses to TRQs during the NAFTA negotiation were: barley, beans, corn and milk powder. 20

PAGE 21

Implications for the Agricultural Sector All the changes taking place in the agricultural se ctor trade liberalization, reform of the ejido and retreat of Statehave le d to a new incentive structure a ffecting farmers behavior and consequently the way they operate This section analyzes the adju stment strategies farmers have adopted as a consequence of these changes in the ag ricultural sector. In general terms, the reforms were expected to have a positive impact on agricultural sector productivity. Some of th e expected results were: a decl ine of small, less productive farmers, who under this new scenario would be willing to sell their land and move out of agriculture; an increase in crop diversification toward more ma rketable crops; and finally, an increase of capital intensification in the agri cultural sector. For a number of reasons, however, the reforms have not produced the expect ed results (Johnson, 2001; Cord and Wodon, 2001; Davis, 2000; Assies, 2008). There are different and inconclusive answer s to this puzzling s ituation. According to Davis (2000) for instance, farmers are assuming a risk-averse strategy, in which they abstain themselves from incurring big changes and they diversify their sources of income in order to reduce uncertainty. From this point of view, farmers can assure their subsistence by remaining in the same crop production; not investing in technological inputs such as fertilizers, machinery and improved seeds; and keeping a secure source of money through off-farm activities or migrant remittances. Assies (2008), on the other hand, suggests the changes taking place in the agricultural sector trade liberalization, reform of the ejido and retreat of Statehave only deepened the crisis in the rural sector because of the inaccessibility to credit15, insurance, market, modern inputs and 15 The total amount of credit the rural s ector had access to decrease from 30 pe rcent to 20 percent in 1997 (Assies, 2008). 21

PAGE 22

technical assistance in th e rural sector. For a better understand ing of farmers behavior and its impact on productivity, I turn to how ownership of major agricultural asse ts has changed since the policies reforms were implemented. Landholdings experienced so me changes after the ejido reforms. In an analysis done by Davis (2000) for example, based on panel data for 1,287 ejido households, it was found that from 1994 to 1997, the amount of land owned by an individual increased on average by 25%, from 8 NRE16 hectares to 10 NRE hectares. The increase of owned land can be attributed to the fact that common land owned by ejidos was divided after the reform was implemented, as well as to an increase in land converted into agriculture. In spite of the expansion in average land ownership, the chan ges in land tenure appear to have had no impact on productivity. Johnson (2001) tested the hypothesis that farmers faced asset-based credit rationing, meaning the amount of credit offered to individuals was constrained by the lack of assets. This hypothe sis suggested that farmers did not invest in productive assets because of their inability to access the credit market. She found however no evidence to support this hypothesis, implying that the lack of collateral and credit is not the cause of low-capital use and low productivity observed in the agriculture. Th is finding is very important in the sense that it shows that the reform of tenancy in Mexico by itself will not have a positive impact in agricultural productivity. Crop prices have been changing through the ye ars. There is strong evidence suggesting that after the trade reform, the Mexican domes tic agricultural prices are indeed converging toward international prices. OECD estimates re flect how the reduction in the nominal protection of basic staples has proceeded over time. For ex ample, maize protection decreased from 109% to 51% from 1993 to 1994, to 24.13% in 1995. In th e case of yellow maize, the protection 16 National Rain fed Equivalents 22

PAGE 23

decreased from 77%, to 28% and then 5%. Some other crops following the same declining pattern are sorghum, soybean, wheat and barley. In the specific case of rice, the protection estimate decreased and became negative around 1991 (Yunez-Naude and Barceinas, 2000). There is also evidence sugges ting a decrease in the producti on of rain-fed agriculture, which is the realm of small and medium size producers. Between 1985 and 1990, for example, the principal products of rain-f ed agriculture decreased by 0.60 per cent per year and between 1990 and 1994, they fell by 4.35 percent per year. Maize production fell by 4.64 per cent annually and beans by 2.63 per cent (Assies, 2008). Little research has been done on the crop diversification topic. However, in general terms it appears farmers are not undergoing crop dive rsification after guara nteed prices were eliminated. Although the amount of land has increased, maize, bean and fodder crops remain the staples of most producers in most regions of the country. Between 1994 and 1997 for instance, Davis (2000) found that 75% of the surveyed ejido households planted only maize, while 19% intercropped maize with other crops, leaving th e growing of fruit and vegetables as well as fodder unchanged. These results sup port the hypothesis that farmer s are behaving as risk-averse agents investing in low price, riskless production instead of unde rtaking the risk of producing high value crops. On the other ha nd, it has also been argued that crop diversification has not materialized because a considerable number of small farmers produce for own consumption purposes, remaining indifferent to re ductions in the relative price gap. As mentioned above, the government has played an important role in providing technical support to farmers through two of its programs, CONASUPO in the past and now ASERCA. The government has focused on the diffusion of improved crop varieties, which has proven to be an important source of agricultural productivity (Wood, You and Zhang, 2004). Through 23

PAGE 24

Procampo, with a coverage rate equal to 80% of all ejidatarios the government encourages the use of high yield variety (HYV) s eeds among basic grains producers.17 And with less success, Alliance, which has reached only 12% of the ejidatarios, the government promotes investment in capital intensity technology and th e transformation of agriculture toward areas with comparative advantage (Cord and Wodon, 2001). Usage of technology, however, does not depend on governmental support alone. Evidence suggests that the characte ristics of the household as well as of the community, such as farm size, community infrastructure and household members education and income also play an important role in determining the use of technology in agricu lture. For instance, Davis (2000) found that after the reforms, larger farmers made more use of HYV seeds, chemicals, technical assistance and credit, while small and poor house holds were the less likely to invest in technology use. Wood, You and Zhang, (2004) found that over tim e, most of the ag ricultural R&D has favored irrigated production systems, where th e potential for technology spillovers is greater than for heterogeneous rain fed areas. Many agri cultural technologies are often location specific. This means that a large part of the agricultural research is directed to overcome site specific constraints in crop production such as, increasi ng plants tolerance to frost or drought, or increasing plant resistance against a specific pest or disease. Homogeneous areas have more potential for agricultural R&D spillovers, thus re search is more abundant on irrigated production systems. 17 Studies measuring the impact of Procampo on agricultural productivity have found that the program has increased the agricultural income of the households keeping farmer s growing their crops. However, the program has had little impact on the productivity itself. Assies (2008) for instan ce, suggests the program has been insufficient to help the rural sector make the transition to other commercial crops. 24

PAGE 25

In general, the increased use of technology in the agricultu ral sector remains a long run goal. The ejido reforms have not had a substantial effect on increasing capital intensification as expected. Government continues supporting the diffusion of technology among farmers, but because of its potential spillovers, the farmers taking advantage of this technology have mostly been large, modernized farms. New programs orie nted toward small farmer s are still needed in order to bring technology to less productive farmers a nd have a major impact in the rural sector. Returning to farmers risk-averse strategies, livestock accumulation seems to be an important strategy for farmers because it keeps their savings relatively liquid but also protects the household from macroeconomic shocks such as in flation or devaluation. In addition, livestock and livestock derivatives consumed at home repr esent a fundamental part of the household diet. Finally, animal by-products such as the sheeps wool also represent an important source of income. Evidence suggests a clear increase in livestock accumulation after the ejido reforms were implemented. Davis (2000) found that on average the number of heads of cattle owned per household increased by almost 20% from 1994 to 1997. About half of the households surveyed had poultry, followed by pigs. Milk was produced by 25% of the households and eggs by 38%. Off-farm activities have always represented an important source of income for many rural households. Evidence from El Salvador, Mexico and Ecuador suggest that nonagricultural employment generates 40 to 50% of a rura l households income (A raujo, 2004), representing from 38% on the largest farms to 77% on the smallest (Araujo, de Janvry and Sadoulet, 2002). Furthermore, about 60% of rural Mexican househ olds have some family member working offfarm (Davis, 2000). 25

PAGE 26

Studies reveal that demogr aphic characteristics of the household members such as gender, age, ethnicity (de Ja nvry and Sadoulet, 2001) and seco ndary education (Araujo, de Janvry and Sadoulet,.2002) play an important role stimulating off-farm activities. The location of the community including the pr oximity of the community to an economic center, and the availability of roads connecting the community al so affect the propensity to be involved in offfarm activities (Araujo, de Janvry and Sadoulet, 2002). In Mexico, the main sources of off-farm in come come from nonagricultural employment, followed by other income, which includes governme ntal direct income transfers and welfare programs, and remittances. The ejido reforms have had an important effect on this specific asset. Evidence suggests that, after the reform, the di versification toward off-farm activities has considerably increased. Davis (2000) found that the proportion of families participating in offfarm activities rose by 33% between 1994 and 1997, encompassing up to 60% of the ejido households. The success of Procampo has also increased the de pendency of many farmers on this source of off-farm income (direct income transfers). The impact of off-farm activit ies on agricultural productivity has not been widely studied. However, it seems agricultural production and offfarm activities are negative correlated. This means the share of total household income deri ved from off-farm activities is inversely correlated to a farms size. The exception to this pattern is rem ittance, which is frequently found among medium size farms (de Janvry and Sadoulet, 2001). Furthermore, evidence suggests road availability as well as proximity to an economic center influence the effect agricultural output has on off-farm activities. Fo r instance, if there is road nearby and the distance between the rural community and the economic center is not so large, high value agricultural output would promote off-farm activities through service and 26

PAGE 27

manufacturing employment growth. On the other hand, if the community is isolated, high value agricultural output woul d shrink off-farm employment (Ara juo, de Janvry and Sadoulet, 2002). One of the major strategies of rural households is intern ational migration. Through the years this phenomenon has been expanding in the rural sector. In his survey of ejido households, Davis (2000) found that around 45% of the house holds had either a family member with migratory experience to the U.S. or children or siblings living there. Moreover, around 50% of the households with more than 5 NRE hectares reported some connection with the U.S. Researchers have noticed that the formation of social networ ks over the years has promoted migration through the reduction of the risk and transaction costs embedded in the migratory experience. Researchers studying the impact of the reforms anticipated an increase in out-migration in the agricultural sector. Studies measuring th e impact of NAFTA, for example, predicted a decrease in rural employment and wages, gene rating an emigration of as high as 800,000 people from the rural sector, migrating mostly to th e United States (Cornelius and Martin, 1993). However, to date this prediction has not materiali zed. As a matter of fact, the agricultural sector continues employing around 20% of the populati on (Taylor, Yunez-Naude and Dyer,.1999; Davis, 2000). Lastly, an important consequence of intern ational migration in the sending communities is remittances, which represents an important source of income in rural areas. In 2003, for instance, Mexican immigrants living in the US sent $14 billion in remittances to their relatives in Mexico (Orozco and Lapointe, 2004). Different st udies have been carried out in the impact of remittances on rural Mexico, reaching no consensu s on its impact. Chapter 3 discusses in detail the findings of these studies. 27

PAGE 28

Conclusion This chapter presented an overview of the main changes in the Mexican agricultural sector until the present time. As opposed to past policies, current polic ies attempt to set the necessary conditions for a market oriented stra tegy. However, it seems trade liberalization and the retreat of the State, have not made producers more responsive to market signals as expected. In addition, the ejido reforms alone have not created enough incentives to increase the productivity in the agricultural sector. The way the agricultural sector will achieve competitiveness within the international markets remains an enigma. Some believe that big private entrepreneurs will bring competitiveness to the sector; others believe sm all farmers who are now land owners and with government assistance will be able to gain compe titiveness in the international market, bringing new forms of self-employment and poverty alle viation (Quintana, Borquez and Aviles, 1998). Research in this area is still limited. As mentioned before, the impact off-farm ac tivities on agricultural productivity has not been widely studied. The next chapter will be devo ted to migration, one of the off-farm activities rural households rely upon heavily, and its linkage to the agricultur al sector in rural Mexico. 28

PAGE 29

CHAPTER 3 MIGRATION AND AGRICULTURAL PRODUCTIVITY Introduction Migration is being increasingly considered in the literature on agricultural productivity in developing countries. This is because migrati on has become a common practice for many rural households world-wide. Migrati on has been found to affect a households decisions in three important ways. First, migration reduces the labor availability of the household; second, it generates an increase in the households inco me through remittances sent by the migrant; and third, it strengthens the formati on of social networks, which can be used to promote the migration of other household members. The aim of this chapter is to review the existing literature on this topic. In order to understand the im pact of migration on the agri cultural sector of the home country, specifically the way in which the households structure changes due to the migration of one of its members and its impact on producti vity, it is fundamental to study not only the demographic characteristics of the migrant and the household, but also to understand the composition of the migratory flows, the macroec onomic factors inducing the migration (Orrenius and Zavodny, 2005;Cornelius, 2001;Jones, 1995;Donato, 1999;1994), and the inherent dynamics of migration (Davis, Stecklov and Winters, 2001; Massey and Es pinosa, 1997;Massey, Goldring and Durand, 1994). In the specific case of Mexico, migration ha s become a common practice in rural Mexico. According to the CONAPO the number of Mexi cans engaging in a migratory experience to the U.S. reached 3.3 million between 1990 and 2000. Furthermore, remittances have become an important part of the Mexican economy reachin g the second place in source of foreign currency after oil exports. The structure of this chap ter is the following. Th e literature on Mexican 29

PAGE 30

migration is reviewed first, followed by the l iterature on labor out-migration and agricultural productivity. Mexican Migration Literature Immigration Reforms in the United States The Mexican agricultural sector has always been very clos e linked to the agricultural sector of the United States. Some factors expl aining this relationship include the geographical closeness between the countries, the similarities in climate, the bonds among relatives and economic factors. In the specific case of the labor market, the immigration la ws have also played an important role in connecti ng the agricultural labor markets of the two countries. The major U.S. immigration reforms during the tw entieth century are described below: The Bracero program Migration of Mexican farmers to work in US fields has been a common practice since the 1940s. The first major Mexican migr atory flow took place during the Bracero Accord, which was implemented between 1942 and 1964 to face the s hortages of agricultural labor in the United States as consequence of World War II. The Bracero program allowed Mexicans to migrate temporarily for agricultural employment in th e United States encouraging seasonality in migration flows, with cyclical movements across countries (Donato, 1994). This program comprised approximately 4.5 million Mexican agricultural workers in total (Massey and Espinosa, 1997). During the same period, the U.S. Congress al so passed the Immigration and Nationality Act (INA) of 1952 promoting the allocation of visas to relatives of US citizens and bracero workers believed not to have an adverse impact on the US labor market. Consequently, many relatives of Mexican farmers enrolled in the Bracero program were able to apply and get visas. 30

PAGE 31

After the Bracero program ended, there was a decline in the number of visas issued to Mexicans. For instance, prior to 1965, there were no numerical li mits to the legal entry of Mexicans; in 1965 Mexico was placed under a hemispheric quota of 120,000, meaning Mexico had to compete with other Latin American and Caribbean countries for visas. In 1976, it was placed under a country quota of 20,000; in 1978 it was included under a global ceiling of 290,000; and in 1980 the global ceili ng was reduced to 270,000 (Masse y and Espinosa, 1997). Immigration Reform and Control Act (IRCA) Decrease in opportunities to en ter the country legally, led to an increase in illegal migration to the United Sates. Indeed, the perc entage of migrants leaving Mexico illegally increased from 37 percent during the Bracero program to 53 percent in 1965-68. Taking into account this situation and in an attempt to reduce undocumented mi gration to the United States, in 1986 the U.S. Congress passed the Immigr ation Reform and Control Act (IRCA). This Act generated several measures to st op the illegal migratory flow. These included increased border enforcement, employer sanctions against those who knowingly hired undocumented migrants, a supplemental guest wo rker program, a modification of the H-2 program, and amnesty to migrants already reside nt in the United States (Donato, 1994; Iwai, Emerson and Walters, 2006). There were two groups of immigrants that we re eligible for legaliz ation under IRCA: the first group was formed by those who had resided in the United States since before January 1, 1982; the second group were seasonal agricultural workers enrolled in th e Special Agricultural Worker (SAW) Program and employed for a mini mum of 90 days in the year prior to May, 1986. Three million Mexicans applied for legalization, a nd nearly 2.7 million were granted permanent residence (Rytina, 2002; Orrenius and Zavodny, 2005). Of this total, approximately, 1.3 million belonged to the second group (Iwai Emerson and Walters, 2006). 31

PAGE 32

Border enforcement In spite of the endless effort s to stop illegal immi gration, the number of illegal migrants entering the United States continues to increase In 1992, Donato (1994) found that 73 percent of the migrants undergoing a first trip entered th e United States without documents. Moreover, studies suggest that th e Mexican illegal population in the United States has grown from 1.1 million in 1980 to 2 million in 1990 and 4.8 million in 2000, with an average annual growth of 90,000 in the 1980s and 280,000 in the 1990s. From the total population of unauthorized residents in the United States, Mexicans account for 69 percent of the undocumented residents (Angelucci, 2005). In the search to stop illegal entry, the govern ment has turned to border enforcement to decrease illegal immigration, esp ecially since the 1986 I RCA. In the last two decades, the U.S. government has raised the enforcement budget of the U.S. Border Patrol from $290 million in 1980 to $1.7 billion in 1998 and more than $2 billion in 2006. Two additional pieces of immigration legislation passed af ter IRCA related to border enfo rcement were the Immigration Act (IA) of 1990 and the Illegal Immigration and Responsibility Act (IIRA) of 1996 (CarrionFlores, 2007). The current debate regarding bu ilding a wall on the southern border and the huge expenses inherent in this projec t, questions the effectiveness of this measure to stop people from entering the country illegally. The Evolution of Mexican Migration Migratory patterns Researchers working on Mexican migration have observed three migratory patterns: permanent migration, characterized by a long interval migration (more than five years), where migrants normally achieve legal status; temporar y migration, characteri zed by shorter trips done mostly illegally; and return migration, characte rized by the return of the migrant to the home 32

PAGE 33

country for either a period of time or forever. Temporary migration if done continuously is known as cyclical or repeating migration and is also associat ed with illegal migration. In the literature, permanent migrants are commonly more skilled, with better jobs and opportunities than temporary migrants. When an i ndividual first migrates to the foreign country, he usually has little or no skills valuable in th e foreign labor market; only after some years when the migrant has learned certain skills such as the foreign language and has acquired some experience, can he aspire to a better job and look for a legal status (Borjas, 1984). When the phenomenon of migration occurs both the receiving and the sending country undergo a change. Most of the research studying the impact of migration on the United States has focused on analyzing legal migrants, based on permanent migration. This is due to data constraints: data availability still plays an impor tant role in defining the unit of study and most data on illegal immigration is limited. In the case of Mexico, data availability is a constraint because most of the databases on migration are not nationally representative. Also they only keep track of temporary migrants who return to th e country of origin at the time of the survey. So much of the analysis on migration in Mexico focuses on temporary migration. Cyclical migration has been the migrator y pattern dominating the Mexican migration literature. Since the implementation of the Bracero program and until the 1990s, studies have found that Mexican migrants migrate to the Unit ed States repeatedly. This cyclical pattern becomes evident in a study carried out by Ange lucci (2005), where Mexican migratory inflows and out flows were calculated. In the study it was found that the migratory inflow of Mexicans between 1972 and 1993 rose to 1,265,000 people, while the outflow was around 95% of the annual inflows (Angelucci, 2005). 33

PAGE 34

Many researchers view Mexican migration to the United States as a self-perpetuating process. As the amount of prior U.S. experien ce grows, and the number of trips to the U.S. increases, so does the likelihood of repeat migration. Apparently, th e nature of cyclical migration is associated with the changes in the fam ily life cycle: increasi ng for young, unmarried men, falling with marriage, and then increasing ag ain as children grow and the households consumption needs rise (Massey and Espinosa, 1997). Kinship ties also play an important role in the migratory decision. Once the migrant has achieved the reunification of the family in the host country, th e probability of returning to Mexico is reduced significantly. Better-educated migrants are more likely to shorten their trips compared with less-educated migrants. The migran ts geographic location of origin also affects the duration of the trip. It seems distance is positively correlated with the duration of the trip. Furthermore, migrants coming from rural areas also spend more time in the United States compared to those coming from urban areas (Carrion-Flores, 2007). Finally, the cyclical migrator y pattern of Mexican migrants has begun to change in the past two decades, as border control has become more rigorous. Apparently, Mexican migrants are very sensitive to changes in border enforcement because they perceive it as an increase in the cost of migration, reducing migratory inflows to the United States (Hanson and Spilimbergo, 1999; Orrenius, 1999 ) At the same time, it discourages recurrent returns to Mexico, and consequently lengthens the time spent in the Unit ed States. For instance, a one unit increase in border controls has proved to decrease the i ndividual likelihood of returning to Mexico by 31.8 percent. This means that if normally 46% of the illegal residents in the United States return to Mexico each year, an increase in one unit of border control redu ce the number of returns by 31% percent (Angelucci, 2005). 34

PAGE 35

Migratory flows In conjunction with the migratory pattern, the composition of the migratory flows has also experienced significant changes during the twentieth century. As mentioned above, economic conditions, social ties, and political issues such as border enforcement play a determining role in inducing or deterring the migration of Mexicans, primarily undocumented, into the United States. Economic conditions in both Mexico and the Un ited States have proven to influence the individual decision to migrate to the United States. For instance, an increase in the U.S. expected wage is commonly associated with an increase in the length of the trip (Carrion-Flores, 2007). Furthermore, a 10% decrease in the real Mexican manufacturing wages is associated with at least a 6% increase in attempted illegal border crossings (Hanson and Spilimbergo, 1999). And, older migrants are more responsive to increases in U.S. farm wages, while nonagricultural wages and the minimum wage in the United States have greater influence on sons migration decision (Orrenius and Zavodny, 2005). When measuring self-selectivity among undocumented immigrants from Mexico, research suggests that higher average U.S. wage s and higher minimum wages are associated with more and less-skilled immigration that lead to a negative selec tion process. Improved conditions in the Mexican economy lead to less migration but also to relatively lower education levels among those who do migrate. In gene ral, skilled workers seem to be more responsive to changes in the Mexican economy and the unskilled, more responsive to changes in the economy of the U.S. It seems skilled workers in Mexico are more tied to the Mexican economy through physical or human capital, making it more difficult to reac t to temporary changes. In addition, skilled workers, unlike unskilled workers, are able to use their savings as a measure to smooth consumption for a longer period of time (Orrenius and Zavodny, 2005). 35

PAGE 36

Demographics of the migratory flows have also experienced some changes throughout the years. In a study carried out on first-time migrants occupati onal decision, a shift toward nonagricultural jobs was found in the recent years. During the Bracero program around 76% of the migrants worked in agriculture on their first U.S. trip. After 1964 however, this percentage dropped by 30% percent. Since then, many migrants have shifted toward unskilled jobs such as manufacturing, service and construction. Furthe rmore the number of migrants employed in skilled jobs increased from 3 percent to 14 percent of total migrants between 1942 and 1992 (Donato, 1994). Immigration reforms in the United States ha ve been an important factor defining the demographics of the migratory flows. During the 1942-1964 period migration was comprised primarily of men over 15 years of age, with almost half of them being bracero workers. After 1964, when many bracero workers achieved legal status and we re able to sponsor their families, the composition of cohorts changed. Women and ch ildren were increasingly likely to leave on a first trip. As more restrictions were implemented a nd the likelihood of entering the United States legally reduced, the flow of wome n continued to increase while th e flow of children was reduced dramatically. For example, the percentage of migrants less than 15 years old dropped from 20 percent in 1977-1981 to 14 percen t in 1987-92. On the contrary, women migrating to the United States increased from 28 to 32 percent between 196976 and 1977-81 respectively. In addition, studies suggest border enforcemen t is not only affecting the age and gender of the composition of the migrator y flows, but also their level of education. The idea behind is that border enforcement represents an increase in the cost of migration, making it more difficult for unskilled workers to raise that money and co nsequently limiting the migration to only those 36

PAGE 37

who can. Higher-skilled workers are more likel y to migrate, increasing positive selection among illegal immigrants (Orreni us and Zavodny, 2005). On the other hand, it has been found social ties affect migratory flows by reducing of migration costs. Using social network as a proxy for migration cost and dividing the sample into communities with low and high-migration costs, a study found that about 38% of households head living in low-cost communities have ever migrated to the United States; while 30% have done so in average-cost communities; and only 17% in high cost communities (Orrenius and Zavodny, 2005). Causes of Migration There is no consensus in the literature on migration about what causes individuals to migrate. Different approaches have been deve loped and employed in different contexts. The most widely used until recently, however, is an economic decision-making framework in which the individual migration decision is based on comparing the expected net present value of income in the destination country and in the place of origin. Todaro (1980) formalizes this framework and predicts that migration occurs only when the expected ne t present value of the earnings (net of transportation cost), weighted by the proba bility of employment at the destination country is positive (Chisw ick and Hatton, 2003; Moretti, 1999). Other approaches have been developed as alternatives to unders tand the causes of migration. Most of these models re-introduce the importance of the so cial context as an explanatory tool of migration. For instance, a sociological ap proach of migration relies on components such as cultural18 and social capital to understa nd migration decisions (Castles, 18 Cultural capital is defined as the knowledge acquired abou t other societies and the work opportunities they offer, including information about the labor market and the liv ing conditions. Social capital is more commonly used specially in terms of social networks and refers to the connections established among relatives, friends or people in a community to reduce the transaction costs and risks of migrating (Castles,2002). 37

PAGE 38

2002). Complementary reasons inducing migrati on include a demographi c approach (Massey, Goldring and Durand, 1994). However, the model that has gained acceptance and become an important conceptual framework for migration in the recent years is the New Economics of Labor Migration approach (NELM) (Stark and Bloom, 1985). This approach studies the prospective migrant as a social agent involved in a familys and community decision-making. The migration decision is linked to the familys strategy to manage uncertainty, diversify the income portfolio and a lleviate liquidity constraints through remittances (Castles, 2002; Stark, 1991). In consequence, the model suggests migr ants, although separated physically, maintain relationships with thei r families during the migratory process. In the specific case of Mexico, Mexican migr ation to the United States has proven to be determined by factors other than just the econom ic condition of the two countries. Furthermore, as noted previously, a common pract ice of rural households is the diversification of their sources of income, which suggests households entail a risk-averse complementary income generation strategy to confront incomplete or non-existent markets (Dav is, Stecklov and Winters, 2001). For this reason, this research will use the N ELM approach as a framework to model Mexican migration. Impact of Migration in the Sending Communities Sociologists have extensively studied the re lationship between migration and community development, paying special attention to th e links migrant establish with people and communities located in nations other than those to which they migrate (Vertovec, 2004). From this point of view, physical barriers as well as the physical location of the migrant lack of importance and instead efforts are made to quant ify the participation of immigrants in the economic, political and cultural life of their country of or igin through the consta nt flow of ideas, 38

PAGE 39

money, and information. (Portes, Escobar and Radford, 2007). This stateless way of studying migration is known as transnationalism in the sociological literature. In his work, Castles reinforces the idea of transnationalism by reformulating the role of the immigrant under the new conditions of globa lization. He argues that a global world would also affect the way migrants are conceived. Migran ts will be each time mo re diverse in social and cultural attributes, and the types of migration will not be limited to the three types mentioned in the last section. New types of migration will include the repeated, circulatory and retirement migration. Also the worldwide use of internet and the improvements in transportation are expected to strengthen informal networks im proving communication and organization among its members. According to the transnationalism litera ture remittances and hometown associations represent two important aspects of migrant tran snationalism. Remittances Studies on the effects of migration in th e sending country bega n receiving special attention, as the amount of remittances sent by the migrants to their families and home communities increased and became significant in volume. For instance, in 2003, Mexican immigrants living in the US sent $14 billion dollars in remittances to their relatives in Mexico. Also, the estimated amount of annual remittances in the world-wide is over $100 billion dollars (Orozco and Lapointe, 2004). Given this inflow, that promises to incr ease in the coming years, researchers are motivated to study the use and impact the remittances have on the households economic activities, taking into ac count that in most cases, migrants came from rural areas. Until now, however, the research on remittances has led to contradictory results and no consensus has been reached on the impact of remittances in the sending country. 39

PAGE 40

Positive evidence on development suggests rem ittances are directly invested in the development of small businesses such as manufact uring and craft companies, as well as in the purchase of productive inputs such as land, seeds, fertilizers and livestock. On the other hand, the negative findings indicate that remittances are not invested in productive activities but on the contrary, are spent on consumption goods such as food, cars, radio and television. Also, these are found to create big inequalitie s among community members and create a culture of economic dependency (Vertovec, 2004). Furthermore, ther e is the concern that remittances reduce the supply of labor by recipients in the labor ma rket, affecting the economic activity adversely (Chami, Fullenkamp and Jahjah, 2005). In another study, it has been found that the pr oductive use of remittances is positively associated with education. In general terms, better educated migrants ar e more likely to have their recipient families invest their remittances in housing or productive capital instead of spending it on consumption or nondurable goods (Durant et. al., 1996). When introducing remittances into a family context model, where the relationship between migrant and family is characterized by altruism, it has be en found remittances serve as co mpensatory transfers to help families overcome financial constraints crea ted by poor economic performances (Chami, Fullenkamp and Jahjah, 2005). Finally, in research using a disaggregat ed, rural economy wide modeling (DREM) approach, an increase in migrants remittances by 10% was simulated. This increase in direct transfers translated into a rise in international migration, whic h in turn drove up the cost of agricultural labor by 1% nega tively affecting cash crop and commercial maize production by between 0.5% and 2%. However, the income of household groups accessing remittances increased between 2% and 5%. An interesting fa ct is that remittances stimulate subsistence 40

PAGE 41

households consumption demand for maize, dr iving up the shadow price of maize and stimulating subsistence production (Tay lor, Dyer and Yunez-Naude, 2005). Hometown associations There is a long history of migrants collec ting money and sending it to home communities for collective benefits. However, it was not until the 1990s that the study of these associations increased. A hometown association can be defined as an organization of immigrants from the same town in a host country who meet for social and multi-aid purposes (Caglar, 2006). Activities performed by these associations vary greatly. Some are involved in charitable work, such as enhancement of the church or th e graveyard, while others focus on infrastructure improvement, such as building sewage treatment plants, providing electricity, paving roads, and improving health care and school f acilities. They can also serve as means of fundraising when natural disasters occur, or for th e celebration of the town patron. The characteristics of members are as di verse as the activities performed by these associations. In a study of associations from three Latin American-origin immigrant groups in the East Cost of the United States, Portes, Escobar and Radford, (2007) found that the personal characteristics of the immigrants play a determin ing role influencing the activities undertaken by an organization. Some of these ch aracteristics include the educati on, age and legal status of the immigrants, as well as their durat ion in the host country and their origins (rural or urban) For instance, migrants coming from rural areas tend to create associations not linked to politics while immigrants from urban origins tend to become mo re involved in the politi cs of their countries. Yet, it is difficult to generalize from these results. In combination with the immigrants backgr ound, the policies devel oped by the sending government can also determine the investment decisions of the associations. Some of the schemes and financial incentives that have been used to channel the hometown associations 41

PAGE 42

(HTAs) investments consist of reduction in tariffs on the importation of machinery and equipment, preferential access to capital goods as well as jo int-investments between local government and migrant organizations (Caglar, 2006). The Mexican Government has played a leadi ng role directing the course of the HTAs activities in the country. Its majo r effort culminated in the creation of the Institute of Mexican Abroad (IME) to promote the participation of th ese associations (Porte s, Escobar and Radford, 2007). Another initiative of the government was Th e Citizen Initiative Program 2x1 created in the 1990s, in which for each dollar raised by the hometown associations, the federal and state government each contributed a dollar to a commun ity project. In 2002, the program was changed into 3x1 to incorporate the municipal government s into this program (Orozco and Lapointe, 2004; Vertovec, 2004). In summary, the Mexican mi gration literature suggests that the demographics of the individual, the creation of soci al networks, and the economic and political condition of both countries affect the way in which the migratory process evolves. The literature also highlights the enormous heterogeneity that exists among th e migrants working in the United States. These factors need to be taken into account when modeling. Now we turn our attention to the agricultural productiv ity literature to study the way in which the decision of migration affects the farming practices of the migrant househol ds compared to non-migrant households. Labor Out-Migration and the Agricu ltural Productivity Literature There is an extensive body of literature that relies on agricultural productivity to measure the efficiency of the allocation of resources. Agricultural productivity is commonly defined as the ratio between total output and total input measured in a given time period (Christensen, 1975). There is no consensus on the way agri cultural productivity should be measured. 42

PAGE 43

Researchers use partial or tota l productivity, gross or net prod uctivity depending on the aim of the research and the availability of data (Dovring, 1979). Land productivity is a topic that has focuse d the attention of many researchers. The impact of tenure security on agricultural produ ctivity and the findings on farm size-productivity relationship are summarized by (Kimhi, 2003; Johnson, 2001; Hayes, Roth and Zepeda,1997). Land management decisions such as crop choice, pl anting dates, fertilizer use rates and capital use have also proven to impact crop yields.19 In the specific case of labor productivity, ther e is an extensive body of literature studying the impact of migration in the host count ry (Napasintuwong and Emerson, (2005); Iwai, Napasintuwong, and Emerson, (2005); Hashida and Perloff, (1996)) but the study of labor outmigration and its impact on the agricultural prod uctivity in the sending community is a small but increasing research area that has at tracted the interest of researcher s in the last decade. The idea of introducing migration in to the agricultural productivity analysis of the local of origin is that national as well as internationa l migration reduces the labor s upply in the community, affecting the farming practices decisions of the household, which in turn can lead to a change in productivity. This topic has recently gained the attenti on of agricultural econo mists due to several reasons: first, the growth in vol ume of migration; second, most mi grants come from rural areas where agriculture remains one of the primary economic activities; and third, migration, through the flow of remittances, is expected to alleviat e liquidity constraints, caused by credit and other markets imperfections in the rural economy, en abling farmers to invest more heavily in enhancing productive assets. 19 For a list of papers exploring the relationship between crop management and productivity see Pender et.al. (2003), Jansen et.al. (2006) and Tittonell et.al. ( 2007). 43

PAGE 44

The literature on migration and agricultura l productivity in the home country can be divided into two main areas. Th e first basically focuses on the study of migration and the way it affects the farming practices of the household through changes in input use, the second is relatively new and introduces the concept of gender into the la bor out-migration analysis. We summarize this literature in below. Changes in Farming Practices and Decisions No consensus has been reached on the way mi gration influences the farming practices and decisions of the household. Migr ation is used in different circ umstances either as a strategy to enhance the productive use of inputs or as a mechanism for movi ng out of agriculture. Most of the empirical work; however, coincides in findi ng differences in the farming practices of migrants compared with non-migrants households. How migration is incorporated into the model also varies greatly from one study to another. Some studies use a Total Factor Produ ctivity approach (TFP) in their analysis to estimate a production function and evaluate th e effects of migra tion on the crop output (Nonthakot and Villano, (2008); Ortega-Sanchez and Findeis, (2001)). Other studies adopt a partial productivity methodology and focus their anal ysis on the effect of migration on a specific farming practice (Mendola, 2008; Miluka, et.al. 2007). Using a cross-sectional household surve y, Mendola (2008) tests whether migration stimulates the use of high-yielding seed techno logy in rural Bangladesh where labor migration has been an enduring phenomenon. The empirical analysis of this paper, based on the NELM insights, addresses the fact that farm households typically face income uncertainty, and measures the effect of migration on risk-taking beha vior in agricultural production. An important contribution of the paper is the differentiation the author makes between temporary-domestic, 44

PAGE 45

permanent-domestic, and international migration to account for heterogeneous household migration strategies. The author finds that households engaging in international migration, which normally have higher initial asset holdi ngs to support migration expenses, are more likely to employ modern farming technology, such as HYVs of ri ce, thereby achieving hi gher productivity after migration of a household member. Asset-poor fa rm households are more likely to engage on domestic migration, and rely more on conserva tive strategies, which do not drive production increases. Mendola (2008) argues that th e success of migratory pr actices, as an income diversification strategy and as pr omoter in risk-taking behavior among farmers, depends heavily on the initial asset hold ings of the household. This means that if rural policies are not implemented to help farmers overcome the uncertainties linked to agriculture, asset-poor farm households, unable to pay the costs of international migration, will be kept marginalized and in a persistent poverty trap. Miluka, et.al, (2007) study the case of Albania, where 54 percent of the population resides in rural areas and agri culture still employs around 50% of the workforce. Following the New Economics of Labor Migration (NELM) appr oach, the authors analyze the allocation of labor and capital resources in th e household as a conse quence of migration. Their objective is to measure the effect on agriculture of changes in la bor supply availability and the gain in access to working capital or credit, due to the inflow of remittances. They first study the impact of migration on the family labor hours spent in agriculture, finding that members of households with migrants abroad work significantly fewer hours in 45

PAGE 46

agricultural production. However, women in migrant households work proportionately more than men, when compared with their counterparts in non-migrant households. Then they measure the impact of migration on non-labor input expenses in agriculture to measure the effect of migration on the invest ment in productivity-enha ncing assets. They conclude that migrant households do not seem to invest in more productive techniques such as chemicals, fertilizers and machin ery. Instead, migrant households are shifting toward livestock production. This result is intuitive since the shift toward less labor -intensive activities such as livestock production can be explained by the fact that male activities are being replaced by female activities within the migrant household. In the case of Mexico, Ortega-Sanchez a nd Findeis (2001) estimate the labor outmigration impact on corn farmers of the central and southern regions of th e country. However, as opposed to the studies of Albania and Banglad esh, they study labor out-migration without differentiating between internal and internationa l migration. Instead they partition their sample according to migration status (migrant households vs. non-migrant households), agricultural environment (traditional, semi-modern and m odern farming practices) and household typology (classification of households by asset endowment or access to it) using discriminant analysis. The idea is that differentiated access of land, agricultural machiner y and family labor affects the way labor out-migration impacts the production system. In opposition to the previous studies that focus on the N ELM approach, Ortega-Sanchez and Findeis (2001) rely less on borrowing and liquid ity constraints and focuses their research on analyzing rural labor markets imperfections. Specif ically, they note family labor faces two major imperfections: the potential moral hazard proble m linked to in-hired labor doing tasks that require intensive effort; and the difficulty of replacing farmers with knowledge and 46

PAGE 47

organizational leadership for traditional farming practices. Overcoming these market imperfection problems requires costly supervis ion or adjustments to farm production. The question that remains unanswered is the abilit y of migrant households to overcome these imperfections. They hypothesize that outmigration reduces the productivity of farm resources and causes a malfunctioning of the agrarian production sy stem. Estimating a sequential production function Ortega-Sanchez and Findeis find that labor outmigration does have an impact on the households productivity. For instance, they find higher ou tput and productivity levels in non-migrant households that use traditiona l or semi-modern techniques co mpared to migrant households using the same techniques. The same relatio nship is found for households with low-asset endowments. This relationship, however, does not hold for migrant households using modern techniques or with high-asset endowment. These findings support the idea that the impact labor out-migration has on agricultural productivity depends not only on th e households initial endowments but also on how labor-intensive the farming practices of the households and the households ability to substitute family labo r with reciprocal or in-hire labor. To analyze the substitution capacity of th e households, Ortega-Sanchez and Findeis (2001) calculate the partial el asticity of substitution, differe ntiating by household migration status. They find similar elas ticity ratios across inputs and task s for households in both groups. This finding suggests that the difference in outpu t and productivity might be due to some sort of inefficiencies in the migrant households. Such inefficiencies, however, are not observed in migrant households using modern techniques or with high-asset endowment. Another recent study carried out in Mexi co by Taylor and Lopez-Feldman (2007), focuses on the ways in which labor out-migratio n influences incomes and productivity of land 47

PAGE 48

and human capital. Using the ENHRUM20 survey, they estimate a switching regression model with cross-section income for 2002 and retrospective data on in ternational migration dating for 1990. Their findings suggest that migration to the United States increases the per capita income of households. They find that households with a migrant in 1990 had higher marginal returns to land in 2002 than households that did not participate in migration. This finding suggests that a lapse of time is required before the effects of migration on productivity can be obse rved. In the case of human capital, they find that an additional year of farmer schooling has a significant and positive effect on total income in households without a U.S. migrant and no eff ect in households with migrants. This findings suggest that local wages is important in the migration decision of the household. Nonthakot and Villano (2008) pick up on th e unresolved question regarding whether remittance incomes enhance production enough to co mpensate for the reduc ed availability of labor (Mochebelele and Winter-Nelson, 2000; Ro zelle, Taylor and DeBrauw, 1999). They study rural-urban migration in the northern part of Thailand, where seasonal migration has been a common practice especially during the dry seasons. They estimate a stochastic production function to evaluate the effects of migration on the mean maize output. Their findings suggest that indeed labor s hortages, caused by migration, negatively affect maize production. Nevertheless, th ey also find that remittances as well as the period of migration have a positiv e effect on decreasing technical in efficiencies. This means that the longer the duration of migration, the greater the chances are that the migrant will send an important amount of remittances to the household, a llowing it to invest more income to improve production efficiency. 20 The acronym stands for Mexican National Rural Household Survey 48

PAGE 49

They also find that age and education ha ve a negative association with technical inefficiency, inferring importance of experience and knowledge in improving farm management practices. Finally, the migrant farms computed an average technical efficiency of 86 percent, which is ten percent more than the coefficient registered for non-migrant farms. These findings support the idea that migration can help relax liquidity and credit constraints, allowing households to buy productivity-enhancing inputs such as chemicals and fertilizers and to hire preand post-harvest labor for their farming operations in a timely manner. Gender Productivity Another important consequence of outmigrati on, in addition to shortages in labor and remittances, has been the shift from male to female labor in agriculture. This shift in roles within the migrant household has introduced the study of gender productivity into the labor outmigration literature. In the gender productivity liter ature, studies have been done to analyze the impact of gender discrimination on the allocation of agricultural inputs (Deere and Leon, 2003; Doss and Morris, 2001); the effect on production of the intra-household allocation of resources among the households members (Udry, 1996); and th e gender differences in production between maleand femaleheaded households (Masterson, 2005; Holden, Shiferaw and Pender, 2001; Jacoby, 1992; Lastarri a-Cornhiel, 1988). In the labor out-migration and agricultural pr oductivity literature, the way in which the gender effect is modeled depends on the availabi lity of data on the ownership of land as discussed in Quisumbing (1996) In the database used in our study 14% are female-headed households; however, only 5.4% percent of these hous eholds work the land and less than 1% has undergone a migratory experience, making it difficult to carry out inter-household comparisons. In this case, gender effects can only be measur ed in terms of the intr a-household allocation of resources. 49

PAGE 50

Most of the empirical work done on the intr a-household gender effect of labor outmigration has been carried out in Africa. According to Mabogunje (1989), for example, the outmigration of farmers in the Sub-Saharan African agrarian economies has led to the reorganization of the traditional labor supply inst itutions and the changing role and status of women. It has been observed that agricultural pr oduction has had to adjust toward tasks that are less labor intensive, and women have started assuming an active role in the decision-making of the household. Additional studies from South Africa coincide in the belief that farms without male labor are at disadvantage compared to other households. Farm households in South Africa with the male migrating to another place, for example, experience lower productivity per acre and per worker due to the shortages in labor the mi grant household face (Mas terson, 2005). Mochebelele and Winter-Nelson (2000) analyze the effect s of gender on farms estimating the technical efficiency coefficients for each group, taking in to account migratory status. With an average technical inefficiency of 0.24 for female and ma le managers in the migrant sample and of 0.37 and 0.35 for female and male managers in the non-migrant sample resp ectively, the author concludes that within each migratory group, the gender-based estimates are not significantly different from the sample estimates, suggesting no gender bias in technical inefficiency. The findings that both male and female farm mana gers benefit from having a household member away shows that the benefits of migration, through remittances, is not gender biased. In the case of Latin America, and specifical ly Mexico, female labor is more oriented toward household domestic activities making it much harder to analyze the efficiency of the allocation of resources. Furthermore, the di vision of labor in agriculture tends to be complementary, female and male. In addition, as mentioned before, the availability of data on 50

PAGE 51

female-headed households undergoing a migratory experience is scarce. For these reasons, the literature in this topic is still very limited. Conclusion This chapter presents an overview of the e volution of the Mexican migration during the twentieth century and summarizes the exiting literature on labo r out-migration and its impact on agricultural productivity. In general, no cons ensus has been reached on the way migration influences the farming practices and decisions of the household. However, it has been noticed that the households initial endowment s as well as the type of migra tion entail different effects of labor out-migration in the sendi ng community. Furthermore, it has been found that, in general, remittances are being used to relax credit constraints and improve the farm management practices of the household. On the other hand, the study of labor out-migration on gender is still limited. In the case of Mexico, the analysis of the way in which migration affects productivity in rural households is still limited. Existing studies suggest important differences between migrant and non migrant households. Furthermore it has been found that the effect of labor out-migration on agricultural productivity depends not only on the households initial endowments but also on how labor-intensive the farming practices of th e households are, and the households ability to substitute the family labor with reciprocal or in-hire labor (i.e. the way the rural labor market works). Chapter 4 describes the database and explai ns the methodology that will be used in this study to analyze labor out-migra tion in Mexico and its potentia l impact on agricultural labor productivity. 51

PAGE 52

CHAPTER 4 DATA ANALYSIS AND METHODOLOGY Introduction Building on the theoretical foundations and subs tantial empirical research highlighted in the previous chapters, my hypothesis is that labo r productivity in migrant households is greater compared to the labor productiv ity of non migrant households. Adopting a NELM theoretical framework and reasoning (Castles, 2002; Star k, 1991), my primary hypothesis is that, as a household strategy to manage uncertainty and market imperfections, migrant households maintain their agricultural produc tion level by investing more in capital-intensive inputs to compensate for the reduced labor force availability due to the migration of at least one of the household members. That is to say, among migrant families, the av ailability of labor measured as the total number of days worked in agriculture is expect ed to fall as members in the household migrate. My corollary hypothesis is that labor productivit y, measured as the agri cultural output generated per day of work will be greater in migrant households compared to non migrant households. To test these hypotheses, my study employs econom etric techniques using the Mexican National Rural Household Survey (ENHRUM). The contribution of this study to the existing literature focuses on three main points. First, we rely on the New Economics of Labor Migration (NELM) approach, using the household as the unit of analysis to study the way labor out-migration influe nces the labor productivity of rural households. Second, we estimate labor pr oductivity accounting for the selectivity of landholding. The idea behind this is that agricultural productivity can only be measured for those households holding land, and until now no study recognized this se lectivity when studying labor 52

PAGE 53

productivity. Finally, we introdu ce into the labor productivity analysis the study of social networks. This chapter has the following structure. The first section describes in detail the survey and the descriptive statistics of the sample. The second section describes the Heckman two-stage procedure. The third section summarizes the va riables introduced in the model. The fourth section tackles the possibility of having endogenous variables in the model. The fifth section presents conclusions. Data and Descriptive Statistics The Mexican National Rural Household Survey (ENHRUM) The Mexican National Rural Household Survey (ENHRUM)21 is a survey conducted among rural communities in Mexico,22 that is part of a project co-directed by the Colegio de Mexico (Colmex) and the Univer sity of California at Davis.23 The goal of this project was to obtain a representative survey of Mexican rural society and econom y; this sample would enable researchers to study the way in which the agricu ltural and trade reforms have impacted the production, income and migration of rural households in Mexico. It is a cross sectional su rvey which includes 8,520 indivi duals from 1,765 households in 14 states. According to Mexicos National Information and Census Office (INEGI), who designed the sample the survey represents mo re than 80 percent of the rural population in Mexico. It was conducted between January and March 2003 and collected detailed sociodemographic and economic characteristics of the households as well as th eir labor and migratory 21 EHNRUM stands for Encuesta Nacional a Hogares Rurales de Mxcio 22 The communities included in the sample contain a population between 500 and 2499 people. 23 For more information visit http://precesam.colmex.mx/ENHR UM/PAG%20PRIN_ENHRUM_.htm 53

PAGE 54

experience24. It also captures the farming practices of the household, sources of income and credit history, among other variables. In addition, it provides information on use of family labor and consumption. The information captured in the household survey has been classified into 12 chapters: housing, household members, plot, crop, livestock, natura l resources, other expenditures and incomes, assets, credit and inheritance, house hold corner store ( tienda) and fishing. In addition to the household survey, ENHRUM collected information on the surveyed communities, such as major economic activities, possession land, land characteristics, overall farming practices, use and access to natural resources, migrat ory patterns and governmental programs, among others characteristics during the months of August and October of 2002. The goal was to provide a generalized picture of the economic, social and political situation of the surveyed communities. The communities were also grouped into five diffe rent regions defined by the National Development Plan: Northeast, Northwest, Midwest, Central and SouthSoutheast25. Some drawbacks of the survey are the fact that it is cross sec tional data, it does not provide information on return migration or durati on of trips, and some of the farming practices were aggregated at the household level instead of the desired parcel/plot level. The first one will not allow us to make inferences across time, such as inferences about wh ether or not migration has made Mexican Agriculture more or less pr oductive. The migratory history provided by ENHRUM is not as rich when comp ared to the MMP. Finally, the lack of data at the parcel level does not permit us to make any analysis at the parcel/plot or crop level. 24 A similar survey is The Mexican Migration Project (MMP), which has tracked information of migratory experience of the head of the households since 1982. However, the MMP does not provide information about agricultural practices. 25 For a list of the community codes refer to Table 4-1. 54

PAGE 55

Sample Description Households represent the un it of analysis in this st udy. According to the NELM literature, migration becomes an intra-household strategy to overcome liquidity and other market imperfection constraints (Castles 2002; Stark, 1991). Hence, in order to analyze the impact of labor out-migration on agricultural productivity the study needs to rely upon the household to determine how households reallocates the rema ining labor and capita l resources once the member of the household migrates. The number of observations in the survey consists of 1765 households (n=1,765). An important feature of this databa se, however, is that the number of observations in each chapter varies greatly across households. In order to calculate the labor productivity of the household, for example, information on production and labor employed are required. From those households reporting information on plot characteristics (n=871), only 762 (n=762) reported their annual production in the survey and 707 reported the fa milys labor during th e crop cycle (n=707). For the purpose of this study, those house holds that reported information on both production and labor (n=707) were the only households taken into account.26 This group of households is labeled sub-sample B and is the on e used to measure the labor productivity at the household level. The labor productivity of the head is also measured separately because the household head constitutes an important asset in the familys labor force. The number of households reporting information on the households head labor productivity equals 667 and represent the sub-sample A (n=667). In order to avoid the non randomness nature of sub-sample A and B, we first estimate the probability of households having access to land and from those that have land, we then estimate 26 The remaining 164 households reporting information on plot characteristics but not specifically on output and labor force are excluded. 55

PAGE 56

the labor productivity of the household. We defined a househol d as a land holder if during the survey the household reported information on plot characteristics otherwise it is considered a non landholder. This is a strong assumption, but unfort unately no direct question about possession of land was formulated in the survey. Something to keep in mind is that land tenure is not taken into account under this definition. This means that the plot could be owned, rented or leased and the household would still be considered a landholder. Sample A consists of 1561 households, of which 894 are non landholders and 667 are landholders. Sample B consists of 1601 hous eholds, of which 894 households are non landholders and the remaining 707 are landholders. In sample A we are measuring the labor productivity of the household head assuming the fam ilys labor is a function of the labor force of the head alone. In sample B we are measuri ng the labor productivity of the household. We are assuming the familys labor is a function of the head, the son/daughter, the wife, the grandchildren and the son/daughter-in-law. Once we described the two samples, we devote what is left of this section to describe the demogra phic characteristics of our samples. Because sample A forms part of sample B, and in an atte mpt to avoid duplications, we will focus on the descriptive statistics of sample B in this chapter. Of the 1601 households conforming sample B, 86.7% are male-headed and the remaining 13.3% are female-headed households. The average age of the household head is approximately 48 years27; while the average age of the household spous e is 41 years. 84% of the households are married or live together. The level of education varies grea tly across households overall, the level of education can be considered as low. As shown in Figure 4-2., in approximately 17% of the households, the head of the household has no e ducation; in 21%, he/she finished elementary school; in 40% he/she has some elementary school; only 9% finished middle school; and 2% 27 Figure 4-1 presents a breakdown of the household head age. 56

PAGE 57

finished high school. In the case of the househol d spouse, in 14% of the households the spouse has no education; in 61% he/she has elementary school; in 14.9%, middle school education; only 3.2% finished or not high school. Despite the low education rate, the average experience of the household head (in any sector, not exclusively agriculture) is approximately thirty three years. Fu rthermore, 24.2% of the household heads were employed outside the agricultu ral sector during their first job. The average number of household members is approximately five,28 with 29.1% of the households having at least one child between zero and six years old. Finally, 17% of the households in the sample speak an indigenous language at home. The following sections compare sub-samples according to land holders and migratory experience. Landholder and non landholder mean differences For the 1601 households in sample B, we have plot level attributes for only 707 of the households, what we have de fined as landholders (n=707), the remaining households are considered non landholders (n=894). Now we an alyze the potential diffe rences between these two groups. To check for differences in the mean s, we run a ttest assuming equal variance. When comparing household demographics, we observe that landholde rs are relatively older and more numerous than those without land. As shown in Table 4-2, the average age of the landholder head is 51 compared to 46 in a non landholder house. It seems that households speaking an indigenous language are more likely to be landholde rs. For example, only 6.4% of non landholders speak an indigenous language compared with 30.6% of landholders. In terms of marital status, landholders are more likely to have a partner than non landholders. Non landholders have on average higher educ ation than landholders as shown in Figure 4-4. Despite the fact that differences between no schooling an d elementary school are not 28 Figure 4-3 shows the number of members in the households. 57

PAGE 58

observed across heads, differences are observed ac ross partners. The partners of landholders are more likely to have no schooling (18.81%) a nd elementary school (64.64%) compared to non landholders. Furthermore, the non landholder head is on average more likely to have completed middle education (12.2%) than non landholder heads (5.7%). The sa me relationship holds for the head partner. On the other hand, children between zero and si x years old are more likely to be present in non landholders households (32.6%) compared to landholders households (24.8%). We also find that the first employment sector of the head affects the likeli hood of holding land. For instance, only 9.8% of those that started working outside the agricultural sector hold land in the sample. In the specific case of the households liquidity constraints, it is not evident which group faces less of a liquidity constraint. Although land holders receive on average more loans from the bank (6.2%) than non landholders (2%), non landholders are more likely to have an account at a bank (12.6%) than la ndholders (8.91%). According to our proxy of income accounting for the annual expenses of the household, we observe that non landholders have great er home expenses during the year $8.454 than landholders $7,350. Furthermore, the program of Progresa which aims to alleviate poverty in rural Mexico, has on average a wi der coverage in those households holding land (47.4%) than in those not holding land (25.4%). These two variables give some insight that households not holding land are indeed better off th an those households holding land. When we measure the spatial distribution of the households we notice that households living in the Midwest and Northern part of the country are less likely to be landholders than those households located in the central or Southe rn part of the country. As shown in Figure 4-5, 58

PAGE 59

33.5% of the landholders live in the South, while another 31.40%, in the Central region of the country. Only 22.7% of non landholders live in the Northwest, another 26% in the Northeast and 22.8% in the Midwest. Despite this fact, the di stance from the community to the closest town however does not seem to affect the likelihood of holding land. In the case of international migra tion, 19.2% of landholders and 28.8% of non landholders live in communities where the migrant population represents a significant percentage of the total population, specifica lly more than 20%. This evidence suggests that in fact landholders are less likely to migrate to a nother country than non landholders; however, one must be careful with this evidence since this vari able is being measured at the community level. At the household level there is no significant difference between the two groups in the share of households migrating to the United States, 19% of the migrant households are landholders and 19.4%, of the migrant households are non landholders. However, when we analyze national migra tion differences across groups, we observe than indeed landholders are more likely to migrate to another part of the country (51.5%) in the search for a job than those not holding land (43. 5%). An explanation for this finding is that households holding land are more likely to seek seasonal work in agriculture in other parts of the country and try to diversify their sources of in come working partially in non-farm activities, which most of the time are locate d outside the home community. The landholder sample Landholder sample contains 707 households that were those households that reported information on plot as well as labor productiv ity during the 2002 surv ey. According to the sample, on average the size of land a household ho lds is 9.82 acres, while the average cultivated area is 5.12 acres. One needs to be careful when analyzing these statistics, since farmers in 59

PAGE 60

Mexico are not homogenous. For instance, the annua l value of output of these households ranges between zero and more than $69,212,744 pesos. According to the information obtained in th e survey, the majority of the farmers rely heavily on rainfalls for crop production. Only 24.1% of the households reported access to an irrigation system. In terms of tenancy rights, 82 % of the households reported to own at least one plot and 55.3% to have ejidal rights over at least one pl ot. Of the households with ejidal rights, 55.30% were already registered in Procede at the time of the survey. As explained above, an important limitation of the household survey is that it doesnt capture the input information specific for each cr op. So no direct inference can be made on what agricultural inputs are being us ed to grow which crop. Since our analysis is based on the households strategy, the relations hip between input use and crop pattern was studied at the household-level. This study analyzes three crops maize, beans and vegetables. As discussed in chapter 2, maize and bean represent Mexicos two major st aple crops. Vegetable is a high value crop that after Mexico was admitted into NAFTA, was expect ed to gain in importance. In general, 77.9% of the households that registered a positive output in 2002 grew maize, 24.6%, beans, and 12.7%, vegetables. Many farmers in Mexico grow their crops onl y for subsistence. I ndeed, in the survey 55.5% of the households reported no commercia lization of their 2002 production. However, for those that reported sales, the average share of crop traded represents 45.4% of their total production. The raising of livestock is also a common practice in rura l Mexico (Davis, 2000). The average number of cattle per household equals 2.35. 60

PAGE 61

Chemical use such as fertilizers and pesticid es is a common farming practice. On average 62.8% of the households reported to have us ed fertilizers and 50.4% pesticides during 2002. However, only 23.5% of the households used high yield varieties (HYV). In terms of labor input, the average number of days a household spends in agriculture is a pproximately 88 days in a year, the head of the household spends the most time (54 days), followed by the wife (13 days), son/daughter (11 days), son/da ughter-in-law (6 days), and fi nally the grandson (4 days). As mentioned in Chapter 2, two government al programs promote the adoption of HYV, technical assistance and training among farmers, Procampo and Alliance for the Countryside. In the case of Procampo, the survey reports that 54.3% of the landholding households received a direct income transfer from this program during 2002. Alliance for the Countryside on the other hand has only reached 36.2% of the communities in the sample. Migrant and non migrant mean differences This section analyzes the hous eholds migratory pattern for the whole sample (n=1601). For the purpose of this research, a household is classified as a migrant household if at least one of its members reported having a migratory expe rience to the United States in 2002. The data reveals that 19.2% of the households had an in ternational migrant in this year. During 2002, when the survey was carried out, 4.8% of these we re living permanently in the host country. There are two main reasons why we follow this classification of migration. First, we are measuring labor productivity at one point in time (2002), so inform ation on labor availability is needed for that specific year. Second, although the survey presen ts information on the migratory history of the household (1980-2002) the survey does not capture the return date of the migrant, so we cannot differentiate between those who have returned from those w ho havent at the time of the survey. 61

PAGE 62

I summarized the mean differences in the demographic variables between the migrant and non migrant households in Table 43. We observe, for instance, si gnificant differences in age. The migrant household heads are on average older (52) than non migrant households (47). The migrant household partner is also on average older (46) than the partner in non migrant households (40). No difference in ma rital status was found across groups. Households where an indigenous language is the primary language spoken in the household are less likely to undergo an intern ational migration (3.9%) compared to other households where the predominant language is Spanish (20.2%). In addition, migrant households have on average a larger family size but fewer children between the ages of 0 and 6 years old, compared to non migrant households. For instan ce 30.7% of the non migrant households have children compared to 22.5% of th e migrant households. Education at the household head level is only significantly different across groups at th e middle school level. N on migrant households are more likely to finish middle school (10.2%) co mpared to migrants households (5.5%). The education of the partner also va ries across groups. In migrant hous eholds partners are more likely to have elementary school (67.1%) compared to non migrant househol ds (59.6%) while in non migrant households partners are more likely to have middle school (16.38%) than in non migrant households (8.8%). Spatial distribution of the households also varies across the two groups. As shown in Figure 4-6., migrant households come predomin antly from the Midwest and Northeastern regions. These findings support previous findi ngs on Mexican migrati on (Massey, 1997; 1994) that affirm that states such as Zacatecas and Guadalajara, which are located in the Midwest, experience a high migratory flow to the United States. 62

PAGE 63

Access to credit is an important limitati on faced by many households in rural Mexico. Contrasting the credit situation of both groups, it seems migrant households are more likely to overcome liquidity constraints. Migrant househ olds, for instance, have on average more accounts at a bank. The sending of remittances increases the migrant households likelihood of having an account at a bank. No significant diffe rence, however, was found across groups when analyzing their ability to get a loan from a bank. When analyzing the households annual expenses, migrant households spend on aver age $11,450 while non migrant households only spend $7,159. Description of landholding households accounting for migratory status This section focuses on the differences in farming practices between migrant and non migrant households. The analysis takes into consideration only th e landholders (n=707) subsample. Table 4-4 summarizes the result s of the comparison between migrant and non migrant landholders. At a 90% confidence level we find that migrants possess more land (on average 14. 90 acres), compared to non migrants (8.64 acres). Fu rthermore, migrant households have on average a larger cult ivated area, 7.48 acres, compared to non migrant households, 4.56. Interesting enough, however, there are no differen ces in output among mi grant and non migrant households. Another surprising result is that on averag e the number of days the migrant household dedicates to agricultural producti on is not significantly different from those of non migrant households. This condition holds for all the memb ers of the household (head, wife, son/daughter, son/daughter-in-law, grandchildren) This is an unexpected result, because with the migration to the United States of one of its members, we woul d expect a significant re duction in the number of days the migrant household spends working in agriculture. Although th e differences are not significant, we observe that the head, son and son-in-law in the migrant household spend on 63

PAGE 64

average, less days in agricu lture than non migrant households. However, the wife and grandson in migrant households spend a little more tim e in agriculture than non migrant households. Usage of non labor inputs, however, presen ts differences across groups. Specifically, migrant households are more likely to adopt cap ital-intensive and produc tivity-enhancing inputs than non migrant households. The migrant househol ds, for example, spend on average 81% more on fertilizers and more than the double on seed purchases than the non migrant households. Furthermore, the migrant group is more likely to use machinery (55.2%) during the crop cycle and HYV (32.1%) than the non migrant group (41.7 %) and (22.7%) respectiv ely. The usage of pesticide is also larger on mi grant households but at a signi ficant level of only p<0.10). No difference in cropping patterns was found when we measured th e total production of each crop. However, when considering the predominant crop grown by the household, we find that non migrant households are more likely to assign more than 50% of their production to growing maize, while migrant households, to grow bean. Crop commercialization is also different across groups but only at a significant level of p<0.10. Migrant households sell a greater percentage of their total harvest (25.1%) in comparison to non-migrant households (19.3%). The inverse relationship is observed in the volume left fo r subsistence. In the case of livestock assets, we observe as found in the literature (Miluka, et.al. 2007), that migrant households have a greater accumulation of livestock assets than non-migrant households. For instance, the average number of cattle in a non-migrant household is approximately two, while in migrant households that number increases to almost five. Participation in governmental programs diffe rs greatly between migrant and non migrant households. For instance, migr ant households are more likely to be enrolled in the Procampo program (62.6%) compared to non migrants (52.4%), while non migrant households have greater 64

PAGE 65

chances of receiving an income transfer from Progresa (36.4%) than migrant households (29.6%). Procede was not significantly di fferent across groups. Heckman Two-Stage Procedure In the previous section two important character istics of the sample b ecame evident. First, the information on farm management practices is only available for those households with a plot at the time of the survey (44.16% of the househol ds). Second, we are making the assumption that only those households that reported plot info rmation are the ones that hold land. Given these circumstances, using OLS to estimate a non-random ly selected sample would generate biased estimators. The Heckman two-stage estimation pro cedure deals with the sample selection bias and still analyzes the data by simple least squares methods (Heckman,1979).29 For that reason the Heckman two-stage procedure will be used in this research. Using this method will allow us to draw conclusions based on the whole samp le, taking into account not only the households agricultural productiv ity but their likelihood to have land. The Heckman two-stage procedure is specifi ed by a selection equation defined as follows: iii i i ieCRMXL '''' (4-1) This equation is estimated by maximum likelihood as an independent probit model. In this case, the dependent variable of the selection equation, takes the value of 1 if the household i holds land and 0 otherwise. The independent variables, account for demographic characteristics of the household, accounts for the migratory experience of the household, defined as the migration to the United States of at least one of the members in the household iLiXiM 29 This method has been commonly used to study female labor market participation and to evaluate programs in the social science field correcting for the selectivity of the samples. 65

PAGE 66

during 2002, is a vector of regional characteristics, is a vector of excluded repressor, and is the error term. iR XiCie Parameter estimates from the selection equation generate a vector of inverse Mills ratios. This vector represents the estimated expected er ror and is introduced into the regression equation as an explanatory variable. Only if the dependent variable from the selection equation equals 1, will the regression equation is computed. Thus, the selection equation is the one that determines whether an observation belongs to the regression equation or not (Heckman,1979). In this study for instance, the agricultural productivity of th e household will only be computed if households are landholders. Variables included in are only used to estimate the first-stage of the estimation and are excluded from the regression equation. If the same variables in the selection equation are included in the regression equation, the estimate s in the model become very imprecise. This occurs due to the collinearity caused from a dding the inverse Mills ra tios into the regression equation (Wooldridge, 2001). To avoid this imprecision in the parameter estimates, exclusion restrictions need to be a func tion of the selection equation but not of the regression equation. The next section discusses in detail the variables that will be included in the selection equation. iCiiii i i ieIRM Y ',0( N''' (4-2) )~2ei (4-3) The dependent variable in the regression equation is the ratio of the total output of the household in pesos and the total days the househ old worked the land during 2002. The output is expressed in Mexican pesosiY30 and the labor force in days per year. The independent variables in 30 The output variable was created using information on crop production obtained from the survey and information on crop prices obtained from a Generic Index published by Banco de Mexico. 66

PAGE 67

the regression equation include explanatory variables that account for demographic characteristics of the household. represents the migratory experience of the household as previously defined. is a vector of regional dummy variables. is a vector of variables that are only observed when the household holds land. iX Mi iRiIi is the inverse mills ratio generated in the selection equation and included as an extra explanatory variable in the regression equation and is the error term. We assume the error term is normally distributed, with mean equal to zero and variance equal to a constant The inclusion of the inverse Mills ratio into the regression equation removes the part of the error term corr elated with the explanatory variable and deals with the sample selection bias problem. Next we tu rn to the variables that will be included in the regression equation. ie2Variables Description Dependent Variables Using sample A and sample B we estimate two models. In both, Model 1 and Model 2 the selection equation has as a dependent variable a binary variable taking the value of 1 if the household holds or 0 if it does not hold land ( land ). As mentioned in previous sections, this variable does not account for th e form of land tenure; it only accounts for the fact that the household reported information on a plot of agricultu ral land at the time of the survey. As shown in Table 4-6, most of the households with land own at least one plot (82%), but the household could also be renting or leasing th e land at the time of the survey. On the other hand, the regression equation uses as a dependent variable a ratio of the total output of the household in pesos and the total days the household worked the land during 2002. This ratio serves as a measure of the household s labor productivity. Th is dependent variable measures the difference in la bor productivity among the households. The existing literature on 67

PAGE 68

labor out migration has found that migration gene rates a shortage in th e labor supply of the household. With this variable, we plan to meas ure if the shortages in labor affect the labor productivity of the migrant households co mpared to the non migrant households. We calculate the dependent variable in tw o different forms. Model 1 estimates the households head labor productivity ( prod) using sample A. This variable is the ratio of the total production of the household in pesos and the tota l days the household head worked the plot during 2002. Model 2 estimates the h ouseholds total labor productivity ( prodtotal ) using sample B. This variable is the ratio of the households total annual produ ction and the number of days the members of the household worked the pl ot during the surveyed year. We assume the households labor force consists of the head, wife, children, son-in-law, daughter-in-law and grandchildren. There are two reasons why we want to m easure the household head labor productivity apart. First the head of the household is the me mber of the household who spends on average the largest amount of time in agricultu re (54 days). Second, the head is the member of the household with the largest migratory participation to the United States duri ng 2002 (65.2% of the migrants). Independent Variables Table 4-5 and Table 4-6 summari ze statistics of the selecti on and regression equation. As noted, Model 1 and Model 2 share the same independe nt variables. The depe ndent variable in the selection equation is also the same for both mo dels. The only variable that changes is the dependent variable of the re gression equation. Furthermore, the selection and regression equations also share certain variables, such as the demographic, mi gratory and regional variables. These variables are added in both equations because they affect not only the probability of having land but also the labor produc tivity of the household. We will first describe 68

PAGE 69

the variables both equations have in common and then explain sepa rately the variables unique to each equation. We are including in the model the follo wing demographic variables: gender ( sex ), marital status ( union ), a dummy for children between zero and six years old ( children), number of people living in the household, including children ( members ), indigenous language spoken in the household ( indiglanguage), dummies for education level ( elementary1, middle1, high1 and up), the age of the household head and the age squared ( aged and aged2 ). The dummy for no school is excluded from the model as comparison variable. The variable union accounts for both, marriage and living together. We are keeping bo th marital statuses together, since we are measuring the share of responsib ilities within the household. To avoid making the assumption that the de mographic characteristics of the household head are the characteristics of the entire household we are including in the model the demographic characteristics of the spouse as well. These va riables are education level ( wifeelementary wifemiddle wifehigh and wifeup ), age ( wifeaged ) as well as age squared ( wifeaged2 ). The dummy for no school is also exclude d from the model as comparison variable. In addition to the demographic variables, a proxy variable for income has been created and introduced into both equations ( income )31. This variable is created summing up all the utilities bills as well as other monthly expe nses the households registered during 2002. It includes the expenses on water, gas, wood, elect ricity, transportation, gasoline, television and telephone. It is expected th at households spending more are those that are better off. A variable that controls for national migration will be included in the model ( natmigration1 ) as well. The migration variable that accounts for international migration is ( migration2002 ). This variable as described in chapter 4 and in the model encompasses those 31 The income variable was divided by a scalar of 1000 to avoid zeros in the estimated parameters. 69

PAGE 70

households that have at least one member in the United States during the survey year. In order to account for the dynamics of migra tion, a social network proxy ( morethan20) will be used. Originally, a social network index was built c onsidering the migratory experience of the whole family. However, this variable is potentially endogenous so a variable m easuring the percentage of migrants in the community will be used instead. In Mexico, the farming practices vary greatly across regions. In the Northern part of the country for example, where rainfa lls are scarce, there is a well developed irrigation system and agribusinesses is also well developed, while in the South agriculture relies on rainfall and farmers crops are mainly for subsistence. For this reason, regional dummy variables are also incorporated into the analysis to account for these differences across regions ( south central Midwest and northeast ). The dummy for Northwest is excl uded from the model as comparison variable. There are five exclusion variables introduced on ly in the selection equation. The selection of these variables was made considering those va riables that affect th e households likelihood of holding land, but not its farming practices. The fi rst exclusion variable that was chosen is inheritance ( inheritance ). The idea is that many households could have inherited their land and that is the reason why they re ported plot information at the time of the survey. However, inherited land does not affect the way the household works the land. Households head first employment sector ( hhfirst ) is the second exclus ion restriction. It is a dummy variable taking the va lue of one if the first sector the household head worked in was not the agricultural sector. To control for the pos sible migration of the household head to other places since his first employment, we take into account only those individuals that remained in the same community where they were first employed. The idea is that employment outside the 70

PAGE 71

agricultural sector decreases the probability of having land. For instance, being employed in the non agricultural sector s uggests that the father of the head was not a farmer or that within the community there were other employment activities such as tourism, manufacturing and crafts among others. On the other hand, the other three exclusi on variables were created at the community level. The third exclusion variab le is a dummy variable that acc ounts for those communities that have ejidal land rights ( perejidallandd). The idea is that communities with ejidal lands increase the odds of households having la nd but not necessarily the o dds of improving the farming practices of the land. The fourth exclusion vari able is also a dummy variable that accounts for communities that have community land (sharedland). Community land is commonly used for grazing animals or hunting. As before, the communitie s with shared land increase the probability of holding land but do not affect the farming pract ices of the households. The fifth exclusion variable is a dummy variable that accounts for those communi ties where agriculture is an important source of income ( incomeag). And households in communities dependent on agriculture are more likely to hold land but not necessarily to be more productive. Independent variables introduced only in regression equation include all those variables related to land, such as: land size ( totalacres and plot ), land rights ( ownland and ejidalland ) and soil quality ( irrigation ). Dummies for participation in governmental programs are also incorporated in the model ( procampo and procede ). In the case of labor inputs, the labor input we are controlling for in the model is the number of contracted workers ( contracted ), The non labor inputs in cluded in the model are application of fertilizers (fertilizer ), usage of HYV ( seed ) and usage of machinery ( machinery ). Endogeneity Issues 71

PAGE 72

Endogeneity problem emerges if an independe nt variable include d in the model is correlated with unobservable variab les in the error term. The ex istence of endogenous variables in the model violates the assumption of the classi cal linear regression mode l that the explanatory variable is uncorrelated with the stochastic disturbance term (Gujar ati, 2003). In our research, the observable variable, which is migratory experience, can be correlated with unobservable variables not taken into account in the model that affect th e likelihood of holding land and consequently the farming practices of the household. If the endogeneity problem is not taken in to account, the OLS estimators are not only biased but also inconsistent (Gujarati, 2003) Previous works on labor out-migration and agricultural productivity have used instrumental variables to solve the endogeneity problem. For instance Mendola (2008) instrument s migration using: the edu cation level of the highest educated household member; the sample proportion of households in the village participating in a migratory experience; and a fa mily chain migration variable Miluka, et.al. (2007) used knowledge of the language of the destination country, the share of the ma le population between the ages of 20 and 39, and the minimum dist ance between the household and the two border crossings. Taylor and Lopez-Feldman (2007) used as instrumental variable historic migration such as a dummy for participati on in the Bracero program as well as dummies for internal and international migration participation in the village The inclusion of instru mental variables in the sample selection model, however, requires comple x methods that are beyond the scope of this research. We used a two-step procedure, as an alternat ive to the instrumental variable approach, to account for the endogeneity problem of the migrati on variable. First, a se parate logit model for migration is run. Second, the f itted values are estimated and incorporated into the Heckman 72

PAGE 73

model replacing the original migration variable Although this procedure is not optimal, this methodology serves as an alternative estimati on procedure to solve the endogeneity of the migration variable. The specification of the logit model is the following: iiii ieLVXM''' (4-4) Where takes the value of 1 if the household i has migratory experience to the United States in 2002, and 0 otherwis e. The independent variables, account for demographic characteristics of the household. is a potential instrumental variable of migration. is a dummy variable for holding land and is the error term. iMiXiViLie This study uses distance as potential instrume ntal variable. Two di fferent measures of distance are tested. First, the di stance from the capital city of the state where the community is located to the city in the United States that reported to be the pr imary destination of migrants in the community. Second, the distance from the closes t city where the commun ity is located to the nearest border between Mexico and the United States. Distance can be used as instrumental variable because this variable is correlated with the migratory decision but at the same time uncorrela ted with the error term. Distance discourages migration by increasing the transaction costs of migration. The other way around, the closer the community is to the border, the cheaper the tran sportation and transacti on costs to cross. For example, communities near the border have greater access to information on employment opportunities and border enforcement laws making it easier to migrate. However distance does not impact the likelihood of holding land and th e farming practices used by the household. Conclusion This chapter analyzed the database and the main statistics of our samp le. It also presented the methodology of this research and the variables th at will be included in our model to estimate 73

PAGE 74

the labor productivity of the household. The descriptive sta tistics clearly suggests demographic differences among households, especially betw een those holding and not holding land. In addition, migrant land holders exhibited in genera l greater investment in farming practices such as usage of chemicals and seeds than non migr ant households. As opposed to our expectations, no differences in labor inputs were found across households. In the description of the sample it also b ecame evident the non randomness nature of the data and the importance to account for it in th e model. A Heckman Two-Stage procedure is carried out to account for this tr uncation in the data and the inclusion of the logit predicted values into the sample selection model is used to account for the endogeneity of migration in the model. In the next chapter, I present and analyze results from the estimation. 74

PAGE 75

Table 4-1. The ENHRUM community codes 1 RNameStateName of Municipality Name of Community Magdalena Tlacotepec Magdalena Tlacotepec 1200530001 San Juan Bautista CuicatlanSan Jose del Chilar 1201770007 San Juan Juquila VijanosSan Juan Juquila Vijanos 1202010001 Santa Maria Comotlan Santa Maria Comotlan 1204000001 Santa Maria Peoles Duraznal 1204260007 Santiago Jocotepec San Miguel Lachixola 1204680008 Acultzingo Potrero, El 1300060012 Chicontepec Piocuayo 1300580101 Espinal San Francisco 1300660024 Minatitlan Rancho Nuevo Carrizal 1301080034 Papantla Caristay 1301240018 Uxpanapa Nioes Heroes 1302100098 Chankom Xkopteil 1310170017 Hunucma Sisal 1310380004 Tekom Tekom 1310810001 Tizimin Sucopo 1310960069 Aculco Gunyo Poniente 2150030041 Axapusco San Pablo Xuchil 2150160015 Coatepec Harinas Tecolotepec 2150210031 Ixtapan de la Sal Salitre, El 2150400013 Ixtapan del Oro San Martin Ocoxochitepec 2150410006 Ixtlahuaca San Isidro Boxipe 2150420021 Oro, El San Nicolas El Oro 2150640048 Acambay Tixmadeje Barrio Dos 2150010112 Cuetzalan del Progreso Santiago Yancuitlalpan 2210430037 Naupan Cueyatla 2211000004 Pantepec Ejido Carrizal Viejo 2211110008 Santa Isabel Cholula Santa Ana Acozautla 2211480004 Tecamalchalco Laguna, La 2211540005 Tlacuilotepec Rincon, El 2211780013 Tzicatlacoyan San Bernardino Tepenene 2211930008 Xicotepec Santa Rita 2211970022 Acambaro Maguey, El 3110020032 Ciudad Manuel DobladoCalzada del Tepozan 3110080015 Irapuato Laguna Larga 3110170096 Leon Patia, La 3110200394 Leon Ibarrilla 3110200340 San Diego de la Union Sauceda, La 3110290120 San Luis de la Paz Covadonga 3110330037 Valle de Santiago San Nicolas Quiriceo 3110420116 Code Guanajuato Oaxaca Veracruz Yucatn Edo. Mex. Puebla Midwest SouthSoutheast 1 3 2Middle 75

PAGE 76

Compostela Puerta de la Lima, La 3180040141 Xalisco Aquiles Serdan 3180080004 Santiago Ixcuintla Tambor, El 3180150071 Bahia de Banderas Sayulita 3180200092 Loreto Tierra Blanca 3320240043 Ojocaliente Cerrito de la Cruz 3320360009 Villa de Cos Sarteneja 3320510061 Villa Garcia Copetillo, El 3320520013 Ensenada Nuevo Centro de Poblacion Padre Kino4020010170 Ensenada Nuevo Uruapan 4020010598 Mexicali Represa Aurelio Benansini 4020020516 Mexicali Ejido Colima I 4020020143 Mexicali Ejido Xochimilco 4020020292 Culiacan Agua Caliente de los Monzon 4250060098 Escuinapa Cristo Rey 4250090023 Guasave San Jose de Guayparime 4250110759 Mazatlan Castillo, El 4250120162 Navolato Bledal, El 4250180023 Navolato Campo Balbuena 4250180199 Empalme Mi Patria es Primero 4260250033 Hermosillo Victoria, La 4260300669 Huatabampo Sirebampo 4260330085 Opodepe Querobabi 4260450068 Villa Pesqueira Villa Pesqueira 4260680001 Namiquipa Namiquipa 5080480001 Namiquipa Cruces 5080480031 Balleza General Carlos Pacheco 5080070054 Doctor Belisario DominguezSan Lorenzo 5080220001 Guerrero Rancho de Santiago 5080310091 Juarez Millon, El 5080370643 Canatlan Nicolas Bravo 5100010083 Durango Colonia Hidalgo 5100050187 Lerdo Salamanca 5100120042 Nazas Perla, La 5100150020 San Dimas Vencedores 5100260110 Santiago Papasquiaro Cazadero, El 5100320024 Gonzalez San Antonio Rayon 5280120128 Matamoros Ebanito, El 5280220121 Matamoros Ranchito y Refujio, El 5280220265 San Fernando Punta de Alambre 5280350460 4Northwest Midwest 3 Tamaulipas Chihuahua Durango 5Northeast Source ENHRUM Codebook Zacatecas Nayarit B.C.N. Sinaloa Sonora 76

PAGE 77

Table 4-2. Landholder and no n landholder mean differences 2. Description TotalNo landLandP value Head age 48.126345.579751.31870.0000** Union 0.84010.81880.86700.0086** Indigenous language 0.17050.06380.30550.0000** Head no school 0.16990.15880.18390.1855 Head elementary school 0.20860.20690.21070.8521 Head middle school 0.09310.12190.05660.0000** Partner age 41.249238.424044.82150.0000** Partner no school 0.13990.10180.18810.0000** Partner elementary school 0.61020.58170.64640.0083** Partner middle school 0.14930.18010.11030.0001** Children 0.29110.32550.24750.0006** Members in household 4.84134.57495.17820.0000** First employment sector 0.24230.35680.09760.0000** Income $7,983$8,485$7,3500.0321** Inheritance 0.25610.18790.34230.0000** Loan 0.03870.02010.06220.0000** Account 0.10990.12640.08910.0178** Distance 8.44848.56648.28960.4457 Migration in community 0.24550.28750.19240.0000** International migration 0.19180.19350.18950.8410 National migration 0.47030.43510.51490.0015** Progresa 0.35100.25390.47380.0000** Midwest 0.19930.22820.16270.0011** Northeast 0.20050.25950.12590.0000** Northwest 0.18240.22740.06220.0000** Central 0.20360.11630.31400.0000** South 0.21420.11860.33520.0000** Note income is expressed in Mexican pesos statistically significant at a 90% confidence level ** statistically significant at a 95% confidence level 77

PAGE 78

Table 4-3. Migrant and non migrant mean differences 3 Description TotalNon migrant MigrantP value Head age 48.1347.1552.270.0000** Union 0.84010.83310.86970.1156 Indigenous language 0.17050.20170.03910.0000** Head no school 0.16990.16620.18570.4133 Head elementary school 0.20860.21100.19860.6344 Head middle school 0.09310.10200.05540.0114** Partner age 41.249240.222045.57850.0000** Partner no school 0.13990.13520.15960.2688 Partner elementary school 0.61020.59580.67100.0152** Partner middle school 0.14930.16380.08790.0008** Children 0.29110.30680.22480.0044** Members in household 4.84134.64685.66120.0000** Income $7,983$7,159$11,4500.0000** Inheritance 0.25610.26040.23780.4140 Loan 0.03870.03710.04560.4876 Account 0.10990.09970.15310.0071** Midwest 0.19930.15920.36810.0000** Northeast 0.20050.18620.26060.0003** Northwest 0.18240.19630.12380.0031** Central 0.20360.21100.17260.1339 South 0.21420.24070.07490.0000** Note income is expressed in Mexican pesos statistically significant at a 90% confidence level ** statistically significant at a 95% confidence level 78

PAGE 79

Table 4-4. Migrant landholder and non migrant landholder mean differences 4 Description TotalNon migrant MigrantP value Farm sizze 9.82358.635914.90210.0890* Cultivated area 5.11784.56457.48380.0000** Ejidalland 0.55300.53750.61940.0863* Irrigation 0.24050.23390.26870.3968 Total days worked 88.258889.961680.97760.4995 Total days head worked 54.155655.404948.81340.3822 Total days spouse worked 12.990112.895313.39550.9015 Total days son worked 11.217811.98087.95520.2562 Total days son-in-law worked6.17266.46424.92540.4979 Total days grandson worked3.72283.21645.88810.2126 Contracted workers 16.595816.705516.12690.8866 Machinery 0.44270.41710.55220.0045** HYV 0.24480.22670.32060.0247** Chemical purchase $1,874.1170$ 1,645.9600$2,978.87900.0038** Pesticide purchase $3,074.2280$ 2,136.3180$6,710.23500.0870* Seed purchase $188.2628$150.0360$347.88180.0170** Output $1,249,276$1,251,436$1,240,0420.9813 Maize $285,268$297,026$223,0970.7368 Bean $216,972$243,778$132,5080.6912 Vegetables $675,938$606,808$1,033,7840.6122 Mainly production of maize 0.51490.54620.38060.0005** Mainly production of bean 0.05090.04010.09700.0070** Mainly production of vegetabl e 0.05370.05060.06720.4450 Traded volume 0.20310.19270.25050.0841* Subsistence volume 0.39040.40480.32480.0157** Cattle 2.35671.66385.22480.0000** Procampo 0.54310.52360.62690.0307** Procede 0.46960.48170.41790.1835 Progresa 0.35100.36400.29640.0257** Alianza 0.36160.35700.38110.4303 Note the monetary variables are expressed in Mexican pesos statistically significant at a 90% confidence level ** statistically significant at a 95% confidence level 79

PAGE 80

Table 4-5. Selection equation variables statistics 5 VariableDescription MeanStd. Dev. MinMaxObs land the household holds land (1=yes) 0.44160.49670 11601 Demographics sex Gender (1 = male) 0.86700.33970 11601 childrenChildren from 0 to 6 in the house (1=yes) 0.29110.45440 11601 union Marital status (1= married or living together) 0.84010.36660 11601 membersNumber of people in the household 4.84132.15551 141601 indiglanguageLanguage spoken (1= indigenous language) 0.17050.37620 11601 elementary1Schooling (1=elementary finished or not) 0.60840.48830 11601 middle1Schooling (1=middle school finished or not) 0.13180.33840 11601 high1 Schooling (1=high school finished or not) 0.03690.18850 11601 up Schooling (1=techinical, college or graduate) 0.03870.19300 11601 aged Age of the head 48.126315.560415 951601 aged2 Age of the head squared 2558.11901614.320022590251601 wifeelementarySpouse schooloing(1=elementary finished or not) 0.61020.48780 11601 wifemiddleSpouse schooloing(1=middle finished or not) 0.14930.35650 11601 wifehighSpouse schooling(1=high school finished or not) 0.03190.17570 11601 wifeup Spouse schooling(1=technical, college or graduate) 0.03940.19450 11601 wifeagedAge of the spouse 41.249215.24127 901601 wifeaged2Age of the spouse squared 1933.64301414.20904981001601 income Proxy for household income $7,983$10,5140$149,1001601 Migration natmigration1National migration (1=yes) 0.47030.49930 11601 migration2002Migratory experience in 2002 (1=yes) 0.19180.39380 11601 morethan20Social network proxy 0.24550.43050 11601 Region south Region (1=south-southeast) 0.21420.41040 11601 central Region (1=middle) 0.20360.40280 11601 midwestRegion (1=middlewest) 0.19930.39960 11601 northeastRegion (1=northeast) 0.20050.40050 11601 northwestRegion (1=northwest) 0.18240.38630 11601 Exclusion Variables hhfirst first labour sector (1= no agricultural sector) 0.25610.43660 11601 perejidallanddCommunity has ejidal land (1=yes) 0.08430.27800 11601 sharedlandCommunity has shared land (1=yes) 0.78930.40530 11601 incomeagAgriculture important source of income (1=yes) 0.51840.49980 11601 inheritanceReceived inheritance (1=yes) 0.62020.48550 11601 Note All monetary variables are expressed in Mexican pesos 80

PAGE 81

Table 4-6. Regression equation variables statistics 6 VariableDescription MeanStd. Dev. MinMaxObs prod0 Labor productivity of the household head $50,657$308,4710$5,775,027667 prodtotal0 Labor productivity of the household $34,929$247,6970$5,775,027707 Demographics sex Gender (1 = male) 0.86700.33970 11601 childrenChildren from 0 to 6 in the house (1=yes) 0.29110.45440 11601 union Marital status (1= married or living together) 0.84010.36660 11601 membersNumber of people in the household 4.84132.15551 141601 indiglanguageLanguage spoken (1= indigenous language) 0.17050.37620 11601 elementary1Schooling (1=elementary finished or not) 0.60840.48830 11601 middle1Schooling (1=middle school finished or not) 0.13180.33840 11601 high1 Schooling (1=high school finished or not) 0.03690.18850 11601 up Schooling (1=techinical, college or graduate) 0.03870.19300 11601 aged Age of the head 48.126315.560415 951601 aged2Age of the head squared 2558.11901614.320022590251601 wifeelementarySpouse schooloing(1=elementary finished or not) 0.61020.48780 11601 wifemiddleSpouse schooloing(1=middle finished or not) 0.14930.35650 11601 wifehighSpouse schooling(1=high school finished or not) 0.03190.17570 11601 wifeupSpouse schooling(1=technical, college or graduate) 0.03940.19450 11601 wifeagedAge of the spouse 41.249215.24127 901601 wifeaged2Age of the spouse squared 1933.64301414.20904981001601 incomeProxy for household wealth $7,983$10,5140$149,1001601 Migration natmigration1National migration (1=yes) 0.47030.49930 11601 migration2002Migratory experience in 2002 (1=yes) 0.19180.39380 11601 morethan20Social network proxy 0.24550.43050 11601 Region south Region (1=south-southeast) 0.21420.41040 11601 centralRegion (1=middle) 0.20360.40280 11601 midwestRegion (1=middlewest) 0.19930.39960 11601 northeastRegion (1=northeast) 0.20050.40050 11601 northwestRegion (1=northwest) 0.18240.38630 11601 Plot Characteristics plot Amount of plots per household 1.70721.06481 8707 ownlandthe household owns the land (1=yes) 0.82040.38420 1707 ejidallandpropietary rights (1=ejidal) 0.55300.49750 1707 totalacresTotal acres of land 9.823538.39450537.5707 irrigationpattern (1=irrigation) 0.24050.42770 1707 fertilizerInput use (1=used fertilizer) 0.62800.48370 1707 seed Input use (1=high yield variety) 0.23480.42420 1707 machineryInput use (1=use machinery) 0.44270.49710 1707 contractedTotal contracted workers 16.595842.25390488707 Programs procampoParticipates in procampo (1=yes) 0.54310.49850 1707 procedeEnrolled in procede (1=yes) 0.46960.49940 1707 yp p The dependent variables prod and prodtotal are expressed in pesos/day 81

PAGE 82

0 100 200 300 400Number of households 15-2425-3435-4445-5455-6465-7475-8485-95 Age Figure 4-1. Age of the household head 21% 40% 9% 4% 2% 2% 1% 3% 0% 18% elementary elementary school not finished middle school middle school not finished high school high school not finished technical college graduate noschool Figure 4-2. Education level of the household head 0 200 400 600Number of households 12345678910>10 Members in the household Figure 4-3. Number of members in the household 82

PAGE 83

0 0.1 0.2 0.3 0.4 0.5Percentage of households noschool someelementary elementary somemiddle middleeduc somehigh higheduc technical college graduateEducation level no land land Figure 4-4. Comparison of education level by landholding status 0.0000 0.0500 0.1000 0.1500 0.2000 0.2500 0.3000Percentage of households MiddlewestNortheastNorthwestMiddleSouthLocation of households Total No land Lan d Figure 4-5. Landholder and non landholder spatial distribution 0.0000 0.0500 0.1000 0.1500 0.2000 0.2500 0.3000 0.3500 0.4000Percentage of households MiddlewestNortheastNorthwestCentralSouthLocation of household Total Non Migrant Migrant Figure 4-6. Migrant and non migrant spatial distribution 83

PAGE 84

CHAPTER 5 RESULTS Introduction In this chapter, we present and analyze th e results of econometric models using the Mexican National Rural Household Survey (ENHRUM ). As mentioned in the previous chapter we differentiate between the la bor productivity of the head (henceforth, Model 1) and the labor productivity of the entire household (henceforth, Model 2). With this distinction we aim to evaluate the New Economics of Labor Migration a pproach that states that in order to study the way labor out-migration influences the labor productivity of rural households, the analysis needs to rely upon the household as unit of analysis. In both models we estimate labor productivity accounting for the selectivity of landholding and introduce a variable of social network to capture how the formation of social networ ks impact the household labor productivity. This chapter is structured as follows. We first run the OLS model and present the results32. The second section summarizes the resu lts found when running the Heckman model33. The third section reviews the resu lts obtained when the endogeneity of migration is taken into account in the model. The fourth section presents conclusions. Estimation of OLS We first estimate the regression equati on of Model 1 and Model 2 by OLS. Although OLS does not take into consideration the samp le selection problem as well as the potential endogeneity of the migration variable, this m odel is the simplest way to estimate labor productivity and will be used as a framework to compare the results found in more complex models. 32 The models in this research were run using STATA 33 The inverse of the Mills ratio was first computed to test for the specification of the model and was found significant at a p<0.05. This result corroborates the existence of sample selection in our data and the need to account for it. The Heckman model was run using robust standard errors 84

PAGE 85

Table 5-1 and Table 5-3 summarize the result s found estimating the household head labor productivity (Model 1) and the household total labor productivity (Model 2) by OLS. An important finding of Model 1 is that it is not our migratory variable ( migration2002 ) but the proxy for social network ( morethan20) that is negative and statis tically significant at a 95% confidence level. This finding suggests that the mo re a community is involved in migration, the less labor productive the household h ead is going to be. For instan ce, the labor productivity of the household head is $91,393.64 less in communities where social networks are strong. On the other hand, a head with high school education compared to a head with no education is more productive by $111,730.50. The education of the spouse at elementary and middle school level, however, has a negative im pact on the labor productivity of the household head compared to a spouse with no school. For example, a spouse with middle school education is less labor productive by $121,921.20 compared to a spouse with no school. Internal migration of the household head to other places in the country impacts negatively the labor productivity of the household head by $46,457.16. Measuring labor input we find that contracted is statistically significant at a 95% confidence level. An increase in one unit of wage worker increases the labor productivity of the household head by $740.28. This result suggests that contracted labor force can be used as substitute or complement of the household head labor. In the case of non labor inputs, the usage of fertilizers increases the labor productiv ity of the household head by $54,753.31 but the utilization of machinery reduces it by $58,404.49. On the other hand, the governmental program, Procede has a positive effect on the labor productivity of the household head increasing the labor productivity of the household head by $68,513.92, but Procampo has a negative impact reducing the labor productivity by $46,362.49. We also observe that the househol d head labor 85

PAGE 86

productivity is higher in Midwest but lower in the Central region compared to the Northwestern region. In Model 2 the migratory variab le is negative but not significant. The variable for social network ( morethan20) is again negative and statistically significant at a 95% confidence level. However, the variable in Model 2 has a greater impact on labor pr oductivity than in Model 1. In communities where migration represents more th an 20% of the population, household head labor productivity is $93,465.52 less than in those commun ities where migration is not so important. It can be argued that in those communities where mi gration is a common prac tice, the labor market of the whole community is affect ed, decreasing the number of people available to work the land. Other variables that were f ound statistically significant in Model 2 at a 95% confidence level were wifemiddle, wifeup, fertilizer, machinery and Midwest. The variables that were significant at a 90% confidence level were wifeelementary, middle1, income, natmigration1, procampo and procede The major difference between the two model specifications is the contracted variable, this variable loses statistical significance in Model 2. It seems wage workers do not increase the labor productivity of the enti re household. It can be argued that households employing the labor force of the entire household ar e less likely to hire wage workers compared to households where only the head of the households works the land. In this case, the hired in labor acts more as a substitute than a complement for the household labor force. For this reason, the contracted variable loses statistical significance in Model 2. Heckman Two-Stage Estimation Who holds Land? As shown in Table 5-2 and 5-4, the selecti on equation from both models report similar results. The gender of the household head ( sex ) has a statistically significant effect on the likelihood of holding land. Female-headed househol ds are less likely to own an agricultural 86

PAGE 87

asset and work in agriculture. Similarly, age ( aged ) increases the odds of holding land, but the squared age ( aged2) has the opposite effect, meaning ther e is a breaking point where age no longer affects the likelihood of landholding. Union is negative and statistically significant but only in Model 2. This finding suggests that the ma rital status of the household affects the odds of landholding at the household level but not at the household head leve l. Couples that are married or living together are less likely to hold land. In both models, the sign of the indigenous langua ge variable is posit ive and statistically significant, as expected. This means that househ olds where an indigenous language is spoken are more likely to make a living from agriculture and consequently are more likely to have land than those households where Spanish is spoken. While the dummy for higher education level (such as technical, college and graduate) is negative and statistically significant, is positive and statistically significant the dummy variable for elementary school These findings suggest that household heads with elementary school are more likely to have land compared to those with no schooling; however, households with a technical or college degree are less likely to hold land compared to households with no schooling. This means that households with higher education prefer to work outside the agricultural sector where the return to education is higher. In terms of spatial distribution, the soil quality in the Southern part of Mexico and the weather conditions favor the growth of crops in this region. This fact is supported in our findings. For instance, we find that the dummy variables for the S outhern, Central and Midwestern parts of the country are positive and statistically si gnificant compared to the dummy variable for the Northwestern region. On the other hand, the dummy variable for the Northeastern part is also positive but not statistic ally significant compared to the Northwestern region. 87

PAGE 88

National migration has a negative impact at a 95% confidence leve l on the likelihood of holding land only in Model 2. This finding suggests that households engaging in national migration are less likely to hold land. International migration variable is negative but not significant in both models. The social network va riable however is positive and statistically significant. It can be argued that communitie s where migration is a common practice have a higher propensity to hold land. On the other hand, the dummies that account for the existence of ejidal land rights ( perejidallandd) and the existence of community land ( sharedland) have a positive effect on the likelihood of holding land. These findings suggest that in those communities consisting of ejidos, the probability was higher that the households obtained land. The same reasoning applies for those communities sharing land. Furthermore, the dummy that accounts for those communities where agriculture is an important source of income ( incomeag) is also positive and statistically significant, meaning that those communities highly dependent on agriculture are more likely to have land holders than in communities where agriculture is not an important source of income. Inheritance also increases the odds of landholding. The ejidal land for instance, was commonly inherited through generations by the male in charge of the househol d. That is why it is nor surprising to find a positive relationship between inheritance and land Finally, the first employment outside the agricultural sector is the only variable with a negative value, reducing the odds of households having land as predicted. What affects Labor Productivity? The results found in the regression equation differ between the two models. For this reason, the results of Model 1w ill be analyzed first followed by the results of Model 2. Model 1 household head labor productivity (sample A) 88

PAGE 89

As shown in Table 5-1, the education level of the household head ( middle1 ) affects positively the labor productivity of the head by $ 111,725.80. This relationship, however, is only observable at a 90% confidence level. This finding suggests, as opposed to expected, that agricultural work becomes more productive with the level of edu cation of the farmer. On the other hand, the education level of the spouse ( wifemiddle) has a negative and statistically significant impact on the househol d head labor productivity. A household having a spouse with middle education is $122,025.20 less labor productive that a household having a spouse with no school. Our migratory variable ( migration2002 ) is negative but not st atistically significant. National migration however ( natmigrtion1 ) is negative and statistic ally significant at a 95% confidence level and social networks ( morethan20) at a 90% confidence level. In general, it can be argued that migration indeed affect the labor productivity of the household head. Households engaging in internal migration to other parts in the country are less labor productive than those staying in the community by $46,431.76. Furtherm ore, in communities where social networks are strong, the household head is al so less labor productive by $91,564.94. Similarly to the results found in OLS, th e usage of fertilizer increases the labor productivity of the household head by $54,768.26, but the utilization of machinery reduces it by $58,439.23. On the other hand, contracted is still positive and statistica lly significant. In this case, an additional wage worker increases the labor productivity of the household head by $740.03. Regional dummy variables we re not significant meaning that the location of the community does not play a role in determining the labor productivity of the household head. In terms of governmental programs, Procede is positive and statistically significant at a 95% 89

PAGE 90

confidence level while Procampo is negative but only statis tically significant at a 90% confidence level. Households enrolled in Procede increase the labor productivity of the household head by $68,562.32. However, households enrolled in Procampo seem to reduce the labor productivity of the head by $46,336.40. In this model we also found an inverse relationship between number of plots ( plot ) and labor productivity at a 95% confidence level. Although farm size is not a topic in this study, further research in th is area is recommended. Model 2 household labor productivity (sample B) The results obtained in Model 2 are summarized in Table 5-3. Middle education level of the household head (middle1) as well as the age of the head squared ( aged2 ) lost statistical significance in this model. The dummy variable accounting for middle education level of the spouse ( wifemiddle) remains negative and statistically significant at a 95% confidence level, while the variable accounting for elementary school ( wifeelementary ) also remains statistically significance at a 90% confidence level. It seems that as the education level of the spouse increases, reduces the labor productivity of the household. For instance, households having a spouse with elementary education are $43,954.47 less labor productive and households having a spouse with middle education are $84,443.23 less labor productive than households that have a spouse with no school. On the other hand, the proxy variable used to measure income is positive and statistically significant, suggesting that househ olds with less liquidity cons traints are those that achieve higher labor productivity. For inst ance, an increase of $1,000.00 in the expenditure level of the household increases the labor produc tivity of the household by $1,646.98. We found that the variable of interest in this study ( migration2002 ) is negative and statistically significant in this model. As opposed to expected, migrant households are less labor productive than households with no migratory experience by $28,655.31. On the other hand, the 90

PAGE 91

variables that account fo r national migration ( natmigration1) as well as social network ( morethan20) are also negative and statistically signi ficant at a 90% and a 95% confidence level respectively. In general, it looks like migration, either national or international, reduces the labor productivity of the entire household. Similarly to the results found in OLS, th e usage of fertilizer increases the labor productivity of the household by $49,196.56 but the utilization of m achinery reduces it by $41,890.97. It should be noticed, however, that at the household level th e positive impact of fertilizer compensates for the lost of productivity in machinery, while at the household head level the net effect is negativ e. On the other hand, Procampo is still negative and statistically significant at the 90% confid ence level. In this case, households enrolled in Procampo are $36,933.66 less productive than those not enrolled in the program. Two regional dummy variables gain significance in this model. For instance, the dummy variables for South and Central ar e negative and statistically signi ficant in Model 2. Households living in the Southern region are $71,102.08 less labor productiv e than those living in the Northwestern region. And households living in the Central region ar e $79,421.25 less labor productive than the households residing in the Northwester region. Th is result suggests the existence of regional differences in the agricultural labor pr oductivity of the households. Addressing Migration Endogeneity In order to address the endoge neity of migration, we decided to use as potential instrumental variable the distance from the closest city where the commun ity is located to the nearest border between Mexico and the United States ( kilometers ). This variable, however, was only found statistically significant using sample A, which represents the sample for the entire household. This finding suggests th e instrumental variable is vulne rable to changes in the sample 91

PAGE 92

size and certainly might not be the optimal instru ment for migration. However, given the scope of this research this is the instrumental variable that will be used to estimate Model 2. The estimation of the logit model for migrati on is shown in Table 5-5. The instrumental variable ( kilometers) is negative and statistica lly significant. This findi ng supports the idea that distance discourages migration by increasing the transaction costs of migration. The predicted values ( ivmig) of the logit model are calculated and used to estimate the Heckman model. As Table 5-6 Model 2a summarizes the variables of interest morethan20 and migration2002 lose their significance when the endogene ity of migration is taken into account, meaning migration has no impact on the labor produc tivity of the household. Variables such as middle1 south and central also lose significance. The regiona l dummies also lose significance. The variables that remained statistically significant at a 95% confidence level were fertilizer, wifemiddle and income The variables that remained statistically significant but at a 90% confidence level was wifeelementary contracted procampo, machinery and natmigraation1. We present results from the selection equation estimation in Table 5-7 Model 2a. The results suggest that the predicte d values of migration generate no significant differences in the estimation. The only change occurs in the variable of ivmig. The variable remains negative but gains statistical significance at a 95% confidence level. This means that households having a migrant in the United States during 2002 are less likely to hold land. There are different ways to interpret th e lost of significance in the migration ( migration2002 ) and social network variable ( morethan20) when the predicted values of migration are included in the model. We woul d argue that the predicted value of migration ( ivmig ) can possibly be correlated with the social network variable reducing the si gnificance in both variables. The argument behind is that house holds with strong social networks tend to be 92

PAGE 93

those located nearer the border be tween Mexico and the United St ates. In order to test this argument we removed from the model our social network variable ( morethan20) to measure the effect of the predicted value ( ivmig ) alone. In Table 5-6 and Table 5-7, Model 2b we summarize the results. As expected, the migration variable ( migration2002 ) gains statistical significance in the regression equation and loses statistical significance in the selection e quation. These findings sugg est that, once corrected for endogeneity, migration reduces the labor productivity of the household by $367,465.30, but has no effect on the likelihood of landholding. Conclusion This chapter described the results of the m odel. Different estimation methodologies were used to evaluate the results. The results found in the Heckman selection model support the idea that the likelihood of landholding depends not only on the demogr aphic characteristics of the household but on the specific location of the household. It was proven that communities with ejidal land and community land increa se the odds of landholding. Our model suggests that migrant households are $28,655.31 less labo r productive than those households with no migratory experience. The formation of social networks in the community also has a negative effect. However, when we solve for the endogeneity problem using a potential instrumental variable the impact of international migration and social network becomes less clear. For a better understanding of the impact of migration on labor productivity, further research needs to be carried out, deal ing more properly with th e endogeneity problem of migration. My future research includes finding new instrumental variables that are correlated with migration but not with the fo rmation of social networks to be able to account for the impact of both, migration and social networks on labor productivity. 93

PAGE 94

Table 5-1. Model 1 household head labor productivity 7-1. Number of obs 667Number of obs 1561 F( 36, 630) 4.69Censored obs 894 Prob > F 0.0000Uncensored obs 667 R-squared 0.2113Log pseudolikelihood ######## Adj R-squared 0.1662Wald chi2(36) 49.84 Root MSE 2.80E+05Prob > chi2 0.0623 prod0 Coef. Robust Std. Err. tC o e f Robust Std. Err. z sex 22156.0458800.800.38 21530.3860281.050.36 union -48224.8248051.15-1.00-48098.8163517.30-0.76 children -28451.9133894.03-0.84-28444.9920871.31-1.36 members 3624.185928.290.61 3629.232894.481.25 indiglanguage18899.7732021.580.59 18449.7428300.470.65 elementary128989.9931918.020.91 28865.5134059.740.85 middle1111730.5050393.882.22**111725.8060700.141.84* high1 145698.6095537.321.53145764.70231717.600.63 up -106050.3095377.49-1.11-105728.1089321.10-1.18 aged -6072.116611.69-0.92 -6123.017796.20-0.79 aged2 89.7461.001.47 90.1792.960.97 wifeelementary-51338.6931208.07-1.65*-51439.2634044.48-1.51 wifemiddle-121921.2048961.85-2.49**-122025.2049117.96-2.48** wifehigh-55404.49105966.40-0.52-55409.56100788.60-0.55 wifeup 417508.2092128.154.53**417548.70396237.601.05 wifeaged -1660.195020.46-0.33 -1654.772687.34-0.62 wifeaged2 -18.3549.76-0.37 -18.4739.72-0.46 income 5271.191119.294.71**5265.833333.461.58 natmigration1-46457.1624157.57-1.92*-46431.7622447.38-2.07** migration2002-35733.8232126.28-1.11-35726.8335255.50-1.01 morethan20-91393.6442643.25-2.14**-91564.9451326.15-1.78* contracted 740.28273.812.70**740.03298.322.48** ejidalland-38637.0826178.17-1.48-38725.3826707.39-1.45 ownland 24948.8932831.790.76 24886.8616179.141.54 plot -17658.7411593.58-1.52-17679.928350.82-2.12** irrigation-31732.8229143.47-1.09-31737.2329104.16-1.09 totalacres 92.51302.370.31 92.34199.170.46 procampo-46362.4926635.78-1.74*-46336.4025128.67-1.84* procede 68513.9228874.972.37**68562.3231486.362.18** seed 44213.5929621.041.49 44228.1231041.521.42 fertilizer 54753.3124507.972.23**54768.2624479.082.24** machinery-58404.4924398.91-2.39**-58439.2326682.26-2.19** south -88759.0259996.48-1.48-89662.7167552.83-1.33 central -95113.2757616.47-1.65*-96269.6170440.76-1.37 midwest 95621.2058076.671.65*95050.1698307.660.97 northeast 24942.8461015.850.41 24609.0186014.140.29 cons 289831.40176746.001.64293594.60181212.201.62 Model 1 OLS Estimates Model 1 Heckman selection model* statistically significant at a 90% confidence level ** statistically significant at a 95% confidence level 94

PAGE 95

Table 5-2. Model 1 Heckman se lection model (first-stage) 8 land Coef. Robust Std. Err. z sex 1.06530.13687.79** union -0.13060.1257-1.04 children 0.03100.07890.39 members -0.01770.0153-1.16 indiglanguage1.04480.079113.21** elementary10.18580.08412.21** middle1 -0.09360.1262-0.74 high1 0.04850.22570.21 up -0.53370.2365-2.26** aged 0.10410.01477.06** aged2 -0.00090.0001-6.74** wifeelementary0.11120.08201.36 wifemiddle 0.20500.12531.64 wifehigh 0.11920.22510.53 wifeup -0.22900.2023-1.13 wifeaged -0.00050.0131-0.04 wifeaged2 0.00020.00011.30 income 0.01330.00265.20** natmigration1-0.08540.0593-1.44 migration2002-0.10930.0742-1.47 morethan200.23150.08152.84** south 1.10420.11599.53** central 1.71340.113915.05** midwest 0.68260.11815.78** northeast 0.09080.11480.79 inheritance 0.36870.06295.87** hhfirst -0.33900.1416-2.39** perejidallandd0.43370.06996.20** incomeag 0.65550.061910.60** sharedland 0.37370.06735.55** cons -6.19410.3894-15.91* statistically significant at a 90% confidence level ** statistically significant at a 95% confidence level 95

PAGE 96

Table 5-3. Model 2 household labor productivity 9 Number of obs 707Number of obs 1601 F( 36, 670) 4.29Censored obs 894 Prob > F 0.0000Uncensored obs 707 R-squared 0.1874Log pseudolikelihood -10452.8 Adj R-squared 0.1437Wald chi2(36) 52.58 Root MSE 2.30E+05Prob > chi2 0.0366 prodtotal0 Coef. Robust Std. Err. tC o e f Robust Std. Err. z sex 39708.7443032.380.92 37356.2980298.220.47 union -50293.6637316.21-1.35-49534.6854401.42-0.91 children -15422.3126898.12-0.57-15421.4615954.31-0.97 members 2427.404703.180.52 2436.322075.051.17 indiglanguage-5708.1525181.74-0.23 -7645.388991.74-0.85 elementary118974.6724870.100.76 18374.1324733.690.74 middle1 74869.7040384.591.85*74834.6548948.841.53 high1 -58011.5575710.42-0.77-57592.3869222.37-0.83 up -68255.9873248.88-0.93-66858.3065656.63-1.02 aged -5548.464876.22-1.14 -5745.155659.00-1.02 aged2 77.3744.111.75* 78.9070.741.12 wifeelemenetar y -43485.9524804.56-1.75*-43954.4724167.21-1.82* wifemiddle-83941.0838712.39-2.17**-84443.2337430.47-2.26** wifehigh 12680.6871455.620.18 12196.1739422.100.31 wifeup 425349.6071733.195.93**425457.90372003.901.14 wifeaged 1380.373676.320.38 1384.712375.370.58 wifeaged2 -42.0936.46-1.15 -42.3927.81-1.52 income 1670.64893.341.87*1646.98751.022.19** natmigration1-33910.6119050.50-1.78*-33735.2519171.59-1.76* migration2002-28648.7725184.21-1.14-28655.3114614.80-1.96** morethan20-93465.5233979.27-2.75**-94257.0241681.22-2.26** contracted 277.93220.141.26 276.84163.631.69* ejidalland-22836.0220761.91-1.10-23273.5625967.88-0.90 ownland 19492.3925955.630.75 19266.9512126.631.59 plot -10274.869260.78-1.11-10363.877418.71-1.40 irrigation-15749.3822979.14-0.69-15741.7923073.91-0.68 totalacres 170.10245.250.69 169.27182.160.93 procampo-37047.9021064.19-1.76*-36933.6622155.06-1.67* procede 42227.8922727.551.86*42430.6927140.451.56 seed 22502.1023200.800.97 22602.0320652.551.09 fertilizer 49146.5119444.912.53**49196.5620153.342.44** machinery-41698.1519263.46-2.16**-41890.9724872.56-1.68* south -66894.8847159.96-1.42-71102.0835214.74-2.02** central -74207.2545421.68-1.63-79421.2537540.04-2.12** midwest 99111.8046141.592.15**96582.4485615.681.13 northeast 48548.0248494.341.00 47024.8952231.180.90 cons 190808.60135449.501.41206998.40128378.201.61 Model 2 OLS Estimates Model 2 Heckman selection model* statistically significant at a 90% confidence level ** statistically significant at a 95% confidence level 96

PAGE 97

Table 5-4. Model 2 Heckman selection model (first-stage) 10 prodtotal0 Coef. Robust Std. Err. z sex 0.85830.12217.03** union -0.23040.1119-2.06** children 0.03670.07710.48 members -0.01080.0144-0.75 indiglanguage 1.00730.075213.39** elementary1 0.22800.08052.83** middle1 -0.08730.1223-0.71 high1 0.01920.22350.09 up -0.49610.2310-2.15** aged 0.08650.01426.11** aged2 -0.00070.0001-5.47** wifeelementary0.10530.07891.33 wifemiddle 0.22970.12031.91 wifehigh 0.20880.19931.05 wifeup -0.12180.2130-0.57 wifeaged 0.00970.01250.78 wifeaged2 0.00000.00010.40 income 0.01300.00255.09** natmigration1-0.12580.0577-2.18** migration2002-0.07340.0721-1.02 morethan20 0.21910.08052.72** south 1.14750.112810.17** central 1.70270.113515.00** midwest 0.67580.11565.85** northeast 0.09930.11310.88 inheritance 0.35280.06205.69** hhfirst -0.29080.1309-2.22** perejidallandd 0.46730.06816.87** incomeag 0.65450.059910.92** sharedland 0.37000.06315.86** cons -5.80380.3783-15.34* statistically significant at a 90% confidence level ** statistically significant at a 95% confidence level 97

PAGE 98

Table 5-5. Logit model for migration 11 Number of obs 1601 LR chi2(26) 275.6900 Prob > chi2 0.0000 Pseudo R2 0.1762 Log likelihood -644.6571 migratipn2002 Coef.Std. Err.z sex -0.39320.2917-1.35 union 0.29420.30630.96 children 0.24680.21721.14 members 0.14290.03773.79** indiglangu~e -1.27490.3480-3.66** elementary1 0.15940.19660.81 middle1 -0.44950.3159-1.42 high1 -0.20570.4829-0.43 up -0.71730.5096-1.41 aged 0.06020.03741.61 aged2 -0.00040.0003-1.27 wifeelementary 0.08660.21160.41 wifemiddle -0.28400.3192-0.89 wifehigh 0.49140.46351.06 wifeup 0.41920.46900.89 wifeaged 0.01040.02940.35 wifeaged2 -0.00010.0003-0.22 income 0.02490.00653.82** natmigration1 -0.00820.1528-0.05 morethan20 1.18030.17766.65** south 0.53530.35531.51 central 0.84900.29262.90** midwest 1.15950.24394.75** northeast 0.32680.24751.32 kilometers -0.00050.0003-2.00** land -0.06410.1642-0.39 cons -5.00461.0023-4.99* statistically significant at a 90% confidence level ** statistically significant at a 95% confidence level 98

PAGE 99

Table 5-6. Model 2 household labor productivity solving for the endogeneity problem 12 Number of obs 1601Number of obs 1601 Censored obs 894Censored obs 894 Uncensored obs 707Uncensored obs 707 Log pseudolikelihood -10449.8Log pseudolikelihood-10454.9 Wald chi2(36) 55.62Wald chi2(35) 53.22 Prob > chi2 0.0194Prob > chi2 0.0249 prodtotal0 Coef. Robust Std. Err. zC o e f Robust Std. Err. z sex 30614.4977062.350.40 26054.6576060.740.34 union -38873.7450761.51-0.77-33840.6250033.45-0.68 children -8893.9714177.55-0.63 -5784.8713870.25-0.42 members 6220.253844.221.62 8177.064114.401.99** indiglanguage-31123.4619691.93-1.58-43080.1321140.83-2.04** elementary122806.7926713.460.85 24859.4526971.080.92 middle1 62648.8546126.741.36 57586.6546411.671.24 high1 -64244.1269979.59-0.92-66250.8070120.06-0.94 up -93667.1974610.05-1.26-104607.8076001.90-1.38 aged -4581.285406.57-0.85 -3996.135391.99-0.74 aged2 72.0468.661.05 68.7968.491.00 wifeelementary-39486.6522918.57-1.72*-37201.0822761.56-1.63 wifemiddle-92463.4640865.54-2.26**-96850.8041370.82-2.34** wifehigh 24649.6142727.380.58 31153.8244150.110.71 wifeup 436487.80375889.201.16442094.00377044.201.17 wifeaged 1227.862405.190.51 1208.132425.160.50 wifeaged2 -40.6427.70-1.47 -40.4327.79-1.45 income 2625.821233.112.13**3078.391287.292.39** natmigration1-33925.7619126.52-1.77*-32735.2818828.50-1.74* ivmig -254718.40156318.70-1.63-367465.30167205.40-2.20** morethan20-40070.3825865.51-1.55 .. contracted 268.11157.611.70* 255.67156.011.64 ejidalland-24406.2526181.74-0.93-24954.0226340.37-0.95 ownland 19368.8312293.481.58 20135.7612322.571.63 plot -9743.907171.46-1.36 -9307.147126.42-1.31 irrigation-15370.0223037.16-0.67-15953.9623147.02-0.69 totalacres 152.96174.460.88 139.56173.100.81 procampo-34960.0721093.03-1.66*-34946.0421098.86-1.66* procede 43710.5626884.891.63 44168.2427012.471.64 seed 23338.2020566.251.13 23360.0920565.211.14 fertilizer 47455.7619059.872.49**45898.6818802.112.44** machinery-40628.5324114.23-1.68*-40055.7324019.69-1.67* south -50323.4436786.03-1.37-38422.7235430.63-1.08 central -49829.3139932.84-1.25-34160.8138298.23-0.89 midwest140533.00108893.501.29157611.70111280.301.42 northeast 58345.6757551.031.01 57602.2257245.491.01 cons 147747.30115067.901.28116272.10111967.101.04* statistically significant at a 90% confidence level ** statistically significant at a 95% confidence level ivmig represents the predicted value of the logit model for migrationModel 2 Heckman model solving for endogeneity I Model 2 Heckman model solving for endogeneity II 99

PAGE 100

Table 5-7. Model 2 Heckman selection model solving for the endogeneity problem (first-stage) 13 land Coef. Ro b ust Std. Err. zC o e f Ro b ust Std. Err. z sex 0.78740.12566.27**0.86600.12197.10** union -0.17140.1128-1.52 -0.24080.1110-2.17** children 0.09020.08071.12 0.03380.07820.43 members 0.03150.02031.55 -0.01600.0159-1.00 indiglanguage0.85430.09159.34**1.02560.078713.03** elementary10.27430.08103.39**0.22370.08122.76** middle1 -0.17620.1280-1.38 -0.08530.1231-0.69 high1 -0.02800.2217-0.13 0.01050.22280.05 up -0.67790.2381-2.85**-0.48460.2278-2.13** aged 0.09640.01466.60**0.08480.01425.97** aged2 -0.00080.0001-5.84**-0.00070.0001-5.39** wifeelementary0.13360.08021.67* 0.10100.07911.28 wifemiddle 0.16920.12021.41 0.23560.12091.95* wifehigh 0.32150.20401.58 0.19370.19990.97 wifeup -0.04200.2145-0.20 -0.13680.2125-0.64 wifeaged 0.01110.01260.88 0.01110.01260.88 wifeaged2 0.00000.00010.39 0.00000.00010.29 income 0.02110.00365.82**0.01250.00294.37** natmigration1-0.11390.0581-1.96**-0.13900.0579-2.40** ivmig -1.90680.5881-3.24**0.10210.30090.34 morethan200.60390.15084.00** .. south 1.25470.118310.61**1.09220.108410.07** central 1.88500.128214.71**1.64710.108415.19** midwest 1.00330.15676.40**0.68350.12945.28** northeast 0.20950.11761.78* 0.14620.11431.28 inheritance 0.35200.06195.69**0.34520.06175.60** hhfirst -0.30390.1310-2.32**-0.30420.1295-2.35** perejidallandd0.45160.06906.54**0.47840.06767.08** incomeag 0.65030.059810.87**0.66450.059811.11** sharedland 0.37920.06326.00**0.36430.06315.77** cons -6.33190.4316-14.67 -5.71390.3849-14.84* statistically significant at a 90% confidence level ** statistically significant at a 95% confidence level ivmig represents the predicted value of the logit model for migrationModel 2 Heckman model solving for endogeneity I Model 2 Heckman model solving for endogeneity II 100

PAGE 101

CHAPTER 6 CONCLUSION This study aimed to capture the impact of international laborout migration on the agricultural production of Mexican rural hous eholds. Using the Mexican National Rural Household Survey (ENHRUM) database and estimating a Heckman Two-Stage model we were able to capture the labor productivity of the household accounting for the selectivity of landholding. Our findings suggest that landholdi ng households in rural areas are more likely to have a low level of education, speak an indigenous langua ge and be located in the central and southern part of the country compared to non landholdi ng households. The study su pports Assies, (2008) assessment of the existence of regional differences in the agricultural sector. We noticed that households located in the Southern and Central part of the country ar e less labor productive by $71,102.08 and $79,421.25 respectively, than those lo cated in the Northwestern region. This finding suggests the need to generate incentive s to promote the agricu ltural production in the southern part of the country as measure to cl ose the existing regional gap in the agricultural sector. Another important factor determining the la bor productivity of the household is the level of education. For instance, we found that househ old heads that have middle education level are more labor productive by $111,725.80 than househol d heads that have no school. The education level of the spouse, however, ha s the opposite effect on labor produc tivity. As the education level of the spouse reaches middle education, decrease s the labor productivity of the household head by $122,025.20 and of the entire household by $84,443.23. These findings suggest the importance of improving the education level in rural areas. With this result it also became evident the importance of combining the educatio n level of women with their empowerment in 101

PAGE 102

other markers such as the credit or land rent al market as strategy to increase the labor productivity of the household. We did not find any change in the intra-household allocation of labor. The amount of time the household wife and son spend on agriculture does not affect the la bor productivity of the head or of the entire household overall. Studying the role women play in the labor productivity of the household we found that female-headed h ouseholds are less likely to own land and work in agriculture compared to male-headed households. Procampo, which is an income transfer program, was found to have a negative impact on the labor productivity of the household. It reduces the labor productivity of the househol d head by $46,336.40 and of the entire household by $36,933.66. As Assies (2008) suggests, it seems th at the program has been insufficient to help the rural sector achieve competitiveness. In our study we also noticed that ejidos remain an important type of tenure of ownership in Mexico. We found th at the existence of ejidos in the community as well as of community land increases the households odds of holding land. When measuring the effect of land tenancy status on labor productivity, we observed no impact on la bor productivity. Households participating in the governmental program, Procede, achieved greater labor productivity of the household head by $68,562.32. These results support the belief th at small farmers, with the assistance of governmental programs promoting productivity, will have greater incentives to achieve competitiveness (Quintana, Borquez and Aviles, 1998). On the other hand, we observed that hired in labor increases the labor productivity of the household head by $740.03. However, we did not fi nd the same relationship when analyzing the labor productivity of the entire household. Future studies sh ould measure how the family labor force and wage workers complement or substitute themselves at the presence of migration in the 102

PAGE 103

agricultural sector. The only capital-intensive resource that seems to increase the labor productivity of the household is fertilizer. The usage of chemical s increases the labor productivity of the household head by $54,768.26 and of the entire household by $49,196.56. Further research on the way th e technology adoption of HYV can increase the labor productivity of the households is recommended. Other important concept analyzed in this study was the formation of social networks. Studying those communities where th e percentage of migrants is larger than 20 per cent; we found that households living in these communitie s are more likely to hold land. Our study, however, found that the formation of social ne tworks reduces the la bor productivity of the household head by $91,564.94 and of the entire household by $94,257.02. This result suggests that the existence of social networks in the comm unity can be leading to an overall reduction in the labor force availability constraining even mo re the households capability to substitute or complement the family labor with waged worker s. This finding, however, does not hold when the potential instrumental variable of distance is added into the model. This can be explained by the possible correlation existing between the predic ted values of the instrumental variable and the social network variable. To analyze migration we differentiated betw een national and international migration. Households with national migratory experience are less likely to hol d land and are also to be less labor productive. National migration reduces the labor productivity of the head by $46,431.76 and of the household by $33,735.25. In the case of international migration, we observed no relationship between internati onal migration and access to lan dholdings. However, we found as Miluka et.al. (2007) that inte rnational labor out-migration has a negative impact on the 103

PAGE 104

household labor productivity. Intern ational migration reduces th e household labor productivity by $28,655.31. As mentioned before, this result suggests that migrant households are less labor productive than non migrant households. This fi nding rejects the hypothesis in our study. In general, it seems that migration, either national or international, reduces the labor productivity of the entire household. An explanation for this result is that migrant hous eholds are not investing enough in capital-intensive resour ces to compensate for the reduc tion in labor supply affecting negatively the labor productiv ity of the household. However, one needs to be cautious in interp reting these results. These findings do not hold when the potential instrument al variable of distance is adde d into the model. Furthermore, the coefficient of the predicted value is significan tly greater than the coefficient of the original migration variable, suggesting a greater impact of international migrati on on labor productivity. In order to make better inferences on the wa y labor out-migration imp acts labor productivity, further research needs to be carried out, deal ing more properly with th e endogeneity problem of migration. The contribution of this study to the current literature can be summa rized in three main points. First, we tested the New Economics of Labor Migration approach and proved that migration represents indeed a household stra tegy. Contrasting the way labor out-migration influences the labor productivity of rural hou seholds we found that international migration affects labor productivity at the household level but not at the household head level. Second, taking into account landholding se lectivity we reached the same result found in the case of Albania (Miluka, et.al, 2007), that internationa l migration is affecting negatively the labor productivity of the household. Third, when we in troduced the social network variable into the 104

PAGE 105

model, we reaffirmed the finding that migration has a negative impact on labor productivity in the agricultural sector. Limitations of this research are three. First, the database presents some drawbacks to the analysis as the fact that it is cross sectional data, it does not provide information on return migration or duration of trips, and some of the farming pract ices were aggregated at the household level instead of the desired parcel/plo t level. Second, the st udy of labor productivity and migration require complex methods that are beyond the scope of this research. Third, the utilization of instrumental variables represents a challenge in the research since no econometric method can guarantee that the select ed instrumental variable is th e correct one to solve for the endogeneity problem. Future research will include finding instrume ntal variables that are correlated with migration but not with the formation of social ne tworks to be able to account for the impact of both, migration and social networks on labor productivity. We will also try to develop more complex methods that are able to account for a common endogenous regre ssor in the selection and regression equation in the samp le selection model (Kim, 2006). 105

PAGE 106

LIST OF REFERENCHES Angelucci, M. 2005. U.S. Border Enforcement and the Net Flow of Mexican Illegal Migration. Working Paper No. 1642, IZA. Araujo, C. 2004. Can Non-Agricultural Empl oyment Reduce Rural Poverty? Evidence from Mexico. Cuaderno de Economia 41:383-399. Araujo, C., A. de Janvry, and E. Sadoulet. 2002. Spatial pattern of nonagricultural employment growth in rural Mexico during the 90s, Dept. Agr. Res. Econ, University of California, Berkeley. Assies, W. 2008. Land Tenure and Tenure Regimes in Mexico: An Overview. Journal of Agrarian Change 8(1):33-63. Borjas, G. J. 1984. The Labor Supply of Male Hispanic Immigrants in the United States. International Migration Review 17(4):653-671. Caglar, A. 2006. Hometown associations, the rescal ing of state spatiality and migrant grassroots transnationalism. Global Networks 6 (1):1-22. Carrion-Flores, C. E. 2007. What Makes You Go Back Home? Determinants of the Duration of Migration of Mexican Immigrants in the Un ited States. Unpublis hed, University of Arizona. Castles, S. 2002. Migration and Community Formation under Conditions of Globalization. International Migration Review 36(4):1143-1168. Chami, R., C. Fullenkamp, and S. Jahjah. 2005. Are immigrant Remittance Flows a Source of Capital for Development? International Monetary Fund 52(1). Chiswick, B. R., and T. J. Hatton. 2003. Inter national Migration and th e Integration of Labor Markets. M. D. Bordo, A. M. Taylor, and J. G. Williamson, ed. Chicago/London, University of Chicago Press, pp. 65-117. Christensen, L. R. 1975. Concepts and M easurement of Agricultural Productivity. American Journal of Agricultural Economics 57(5):910-015. Cord, L., and Q. Wodon. 2001. Do Agricultur al Programs in Mexico Alleviate Poverty? Evidence from the Ejido Sector. Cuaderno Economico 38(114). Cornelius, W. A. 2001. Death at the Border: Efficacy and Unintended Consequences of US Immigration Control Policy. Population and Development Review 27(4):661-685. Cornelius, W. A., and P. L. Martin. 1993. T he Uncertain Connection:Free Trade and Rural Mexican Migration to the United States. International Migration Review 27(3):484-512. 106

PAGE 107

Davis, B. 2000. The adjustment strategies of Mexican ejidatarios in the face of neoliberal reform. CEPAL Review Davis, B., G. Stecklov, and P. Winters. 2001. Disaggregating Mexican migrant networks: The parts are greater than the whole. Universi ty of New England, Graduate School of Agricultural and Resource Economics & School of Economics, pp. 1-28. Deere, C. D., and M. Leon. 2003. The ge nder asset gap: Land in Latin America. World Development 31(6):925-947. De Janvry, A., and E. Sadoulet. 2001. Income Strategies Among Rural Households in Mexico: The Role of Off-farm Activities. World Development 29(3):467-480. Donato, K. M. 1999. A Dynamic View of Mexican Migration to the United States. Gender Issues. Donato, K. M. 1994. U.S. Policy and Mexican Migration to the Unit ed States, 1942-92. Social Science Quarterly 75(4). Doss, C., and M. Morris. 2001. How does gender affect the adoption of agricultural innovations? The case of improved maize technology in Ghana. Agricultural Economics 25:27-39. Dovring, F. 1979. Gross and Net Produc tivity: A Problem of Aggregation. American Journal of Agricultural Economics 61(4):694-696. Fernandez-Cornejo, J., and C. R. Shumway. 1997. Research and Pr oductivity in Mexican Agriculture. American Journal of Agricultural Economics 79(3):738-753. Fernandez-Kelly, P. and D. S. Massey. 2007. Borders for Whom? The Role of NAFTA in Mexico-U.S. Migration. The ANNALS of the American Academy of Political and Social Science 2007 610(98). Gujarati, N. 2003. Basic Econometrics. Fourth ed. Mc Graw Hill. Hanson, G. H., and A. Spilimbergo. 1999. Illegal immigration, border enforcement, and relative wages: evidence from apprehensions at the US-Mexico Border. The American Economic Review 89(5):1337-1357. Hashida, E., and J. M. Perloff. 1996. Dur ation of Agricultural Employment. California Agricultural Experiment Station. Hayes, J., M. Roth, and L. Zepeda. 1997. Tenure Security, Investment and Productivity in Gambian Agriculture: A Generalized Probit Analysis. American Journal of Agricultural Economics 79(2):369-382. 107

PAGE 108

Heckman, J. J.1979. Sample Selecti on Bias as a Specification Error Econometrica 47(1):153160. Holden, S., B. Shiferaw, and J. Pender. 2001. Market Imperfections and Land Productivity in the Ethiopian Highlands. Internationa l Food Policy Research Institute. Iwai, N., R. Emerson, and L. M. Walters. 2006. A markov chain analysis with self-selectivity. American Agricultural Economics Association Annual Meeting, Long Beach, California. Iwai, N., O. Napasintuwong, and R. Emerson (2005) Immigration Polity and the Agricultural Labor Market: The Effect on Job Duration International Agricultural Trade and Policy Center. Jacoby, S. H. 1992. Productivity of Men a nd Women and the Sexual Division of Labor in Peasant Agriculture of the Peruvian. Journal of Development Economics 37:265-287. Jansen, H., J. Pender, A. Damon, and R. Sc hipper. 2006. Land management decisions and agricultural productivity in th e hillsides of Honduras. International Association of Agricultural Economists Conference, Gold Coast, Australia, International Food Policy Research Institute (IFPRI), USA Johnson, N. L. 2001. Tierra y Libertad: Will Tenure Reform Improve Productivity in Mexico[s "Ejido" Agriculture? Economic Development and Cultural change 49(2):291309. Jones, R. 1995. Immigration Reform and Migran t Flows: Compositional and Spatial Changes in Mexican Migration after the Im migration Reform Act of 1986. Annals of the Association of American Geographers 85(4):715-730. Kim, K. i. 2006. Sample selection models with a common dummy endogenous regressor in simultaneous equations: A simp le two-step estimation. Economics Letters 91:280-286. Kimhi, A. 2003. Plot size and maize productivity in Zambia: the inverse relationship reexamined. The Hebrew University of Jerusalem, pp. 26. Lastarria-Cornhiel, S. 1988. Female farmers and agricultural produc tion in El Salvador Development and Change 19(4):585-616. Mabogunje, A. L. 1989. Agrarian Responses to Outmigration in Sub-Saharan Africa. Population and Development Review 15 (Supplement: Rural Development and Population Institutions and Policy): 324-342. Massey, D. S., and K. E. Espinosa. 1997. What's Driving Mexico-U.S. Migration? A Theoretical, Empirical and Policy Analysis. The American Journal of Sociology 102(4):939-999. 108

PAGE 109

Massey, D. s., L. Goldring, and J. Durand. 1994. Continuities in Transn ational Migration: An Analysis of Nineteen Mexican Communities. American Journal of Sociology 99(6):1492-1533. Masterson, T. 2005. Land Markets, Female Land Rights and Agricultu ral Productivity in Paraguayan Agriculture. Universi ty of Massachusetts Amherst. McKenzie, D., and H. Rapoport. 2004. Network Effects and the Dynamics of Migration and Inequality: Theory and Evidence from Me xico Mimeo. Stanford University. Mendola, M. 2008. Migration and technological change in rural households: Complements or substitutes? Journal of Development Economics 85:150-175. Miluka, J., G. Carletto, B. Davis, and A. Zezza. 2007. The Vanishing Farms? The Impact of International Migration on Albanian Family Farming. Barcelona, Spain, ESA Working Paper No. 07-09. Mochebelele, M., and A. Winter-Nelson. 2000. Migrant Labor and Farm Technical Efficiency in Lesotho. World Development 28(1):143-153. Moretti, E. 1999. Social Networks and Migrations: Italy 1876-1913. International Migration Review 33(3):640-657. Nonthakot, P., and R. Villano. 2008. Migration and Farm Efficiency: Evidence from Northern Thailand. AARES 52nd Annual Conference. Nonthakot, P., and R. Emerson. 2005. Farm Mechanization and the Farm Labor Market: A socioeconomic Model of Induced Innovation. International Agricultural Trade and Policy Center. Orozco, M., and M. Lapointe. 2004. Mexican Hometown Associati ons and Development Opportunities Journal of International Affairs 57(2). Orrenius, P. M. and M. Zavodny. 2005. Selfselection among undocumented immigrants from Mexico. Journal of Development Economics 78:215-240. Orrenius, P. M. 1999. The role of family ne tworks, coyote prices and the rural economy in Western Mexico: 1965-1994, Federal Reserve Bank of Dallas. Ortega-Sanchez, I., and J. Findeis. 2001. Produc tion Impacts of Labor Outmigration in Agrarian Economies: The case of corn farmers in cen tral and southern areas in Mexico. American Agricultural Economics Association Annual Meeting. Pender, J., E. Nkonya, P. Jagger, D. Sserunkuuma, and H. Ssali. 2003. St rategies to Increase Agricultural Productivity and Reduce Land Degradation: Evidence from Uganda 25th International Conference of Agricultural Economists, Durban, South Africa. 109

PAGE 110

Portes, A., C. Escobar, and A. W. Radford. 20 07. Immigrant Transnational Organizations and Development: A Comparative Study International Migration Review 41(1):242-281. Quintana, R. D., L. C. Borquez, and R. P. Av iles. 1998. Peasant Logic, Agrarian Policy,, Land Mobility, and Land Markets in Mexico. La nd Tenure Center, University of WisconsinMadison. Quisumbing, A. 1996. Male-fema le differences in agricultur al productivity: Methodological issues and empirical evidence. World Development 24(10):1579-1595. Rozelle, S., E. Taylor, and A. DeBrauw. 1999. Migration, Remittances, and Agricultural Productivity in China AEA Papers and Proceedings pp. 287-291. Rytina, N. 2002. IRCA Legalization Effects: Lawful Permanent Residence and Naturalization through 2001.The Effects of Immigrant Lega lization Programs on the United States: Scientific evidence on immigrant adaptation and impacts on U.S. economy and society. The Cloister, Mary Woodward La sker Center, NIH Main Campus. Stark, O. 1991. The Migration of Labour. Cambridge, Basil Blackwell. Stark, O., and D. E. B. May. 1985. T he New Economics of Labor Migration. AEA Papers and Proceedings Taylor, J. E., and A. Lopez-Feldman. 2007. D oes Migration Make Rural Households More Productive? Evidence from Mexico Working Paper, ESA. Taylor, J. E., G. A. Dyer, and A. Yunez-Naude. 2005. Disa ggregated Rural Economy wide Models for Policy Analysis. World Development 33(10):1671-1688. Taylor, J. E., A. Yunez-Naude, and G. Dyer 1999. Agricultural Price, Employment, and Migration in a Diversified Rural Econo my: A Village-Town CGE Analysis from Mexico. American Journal of Agricultural Economics 81(3):653-662. Tittonell, P., K.D. Shepherd, B. Vanlauwe, and K.E. Giller. 2007. Unravelling the effects of soil and crop management on maize productivity in smallholder agri cultural systems of western Kenya -An application of classi fication and regressi on tree analysis. Agriculture, Ecosystems and Environment 123:137-150. Todaro, M. P. 1980. International Migration in Developing Countries: A Survey. R. A. Easterlin, ed. Chicago, University of Chicago Press, pp. 361-404. Udry, C. 1996. Gender, agricultural production, and the theory of the household. The Journal of Political Economy 104(5):1010-1046. Vertovec, S. 2004. Migrant Transnationalism and Modes of Transformation. International Migration Review 38(3):970-1001. 110

PAGE 111

Villa-Issa, M. R. 1990. Performance of Mexican Agriculture: The Effects of Economic and Agricultural Policies. American Journal of Agricultural Economics 72(3):744-748. Wilson, T. D. 1998. Weak ties, strong ties: Network principles in Mexican migration. Human Organization 57(4):394-403. Wood, S., L. You, and X. Zhang. 2004. Spatial Patterns of Crop Yields in Latin America and the Caribbean. Cuadernos de Economia 41:361-381. Yunez-Naude, A. 2003. The Dismantling of CONAS UPO, a Mexican State Trade. Agriculture Colegio de Mxico. Yunez-Naude, A., and F. Barceinas. 2000. Efectos de la Desaparicin de la CONASUPO en el Comercio y en los Precios de los Cultivos Bsicos. Estudios Econmicos July-Dec:189227. 111

PAGE 112

112 BIOGRAPHICAL SKETCH Melissa Ramirez was born in 1981, in Pueb la, Mexico. The younger of two children, she grew up in Puebla, graduating from the American School High School in 2000. During her undergraduate studies, she had th e opportunity to partic ipate in a 1-year exchange program in Tubingen, Germany. She earned her B.S. in ec onomics from the Universidad de las Americas, Puebla. In 2006 she gained an assi stantship to come to the Universi ty of Florida to continue her studies. Melissa graduated in summer 2008 with a Master of Science in food and resource economics and a certificate in supply chain management.