Income distribution among farmers in northern Nigeria

INCOME DISTRIBUTION AMONG FARMERS IN NORTHERN NIGERIA:
EMPIRICAL RESULTS AND POLICY IMPLICATIONS*

by

Peter J. Matlon**

*Published under an Agency for International Development contract
with Michigan State University (AID/ta-C-1328). Field work was con-
ducted while the author was a research assistant in the Department of
Agricultural Economics at Cornell University.

**Assistant Professor, Department of Agricultural Economics,
Michigan State University.

TABLE OF CONTENTS

Page

1. INTRODUCTION . . . . .. 1

1.1. Rural Income, Growth, and Changes in Income
Distribution . . . . 3
1.2. Income Distribution in Africa ......... 5
1.3. Growth and Income Distribution in Nigeria . 7
1.4. Current Policies and Data Requirements . 9

2. CHARACTERISTICS OF THE STUDY AREA AND SURVEY METHODS 14

2.1. Climate and Soils of the Study Area ...... .15
2.2. Characteristics of the Study Villages . .. 17
2.3. Sampling and Survey Methods . .... .19
2.4. Characteristics of the Farming Households .. 20

3. LEVEL, DISTRIBUTION, AND COMPOSITION OF INCOME .... 23

3.1. Definition of Farm Family Income . ... .23
3.2. Man-Equivalent Consumer Units . ... .25
3.3. Mean Income Levels by Village and Household
Sector . . . . 27
3.4. The Size Distribution of Incomes . ... .31
3.5. Intervillage Comparisons . . ... .35
3.6. Female Earnings in Trading and Commerical Food
Processing . . . ... .41
3.7. Sources of Earnings by Income Stratum . .. 44
3.8. Gift Transfers By Income Stratum . ... .47
3.9. Monetization of Households by Income Stratum 49
3.10. Available Calories by Income Stratum ...... 53

Page

4. HOUSEHOLD DEMOGRAPHIC CHARACTERISTICS. . ..

4.1. Family Structure and The Life Cycle . .
4.2. The Distribution of Modern Education . .

5. FACTOR USE AND PRODUCTIVITY . . . .. .

5.1.
5.2.
5.3.
5.4.
5.5.

Land Use . . . .
Land Tenure and Type . .
Ownership of Non-Land Capital .
Labor Use . . . .
Farm Productivity . . .

6. ENTERPRISE SELECTION ACROSS INCOME CLASSES . .

6.1. Subsistence vs. Cash Crop Emphasis . .
6.2. Crop Enterprise Balance by Village and Income
Stratum . . . .

6.3.
6.4.
6.5.

. 82

Crop Mix Variation and Land Productivity .
Choice of Enterprise in Off-Farm Employment
Average Labor Earnings in Non-Agricultural
Occupations . . . .

7. CONCLUSIONS AND POLICY IMPLICATIONS

7.1. Summary of Major Findings .
7.2. Policy Conclusions . .

BIBLIOGRAPHY . . . .

APPENDICES

1 APPENDIX A . . .

2 APPENDIX B . . .

3 APPENDIX C . . . .

. 97
. 101

LIST OF ILLUSTRATIONS

DETAILED MAP OF THE THREE STUDY VILLAGES IN KANO
STATE, WITH INSET OF NIGERIA . . .

THE PERCENTAGE DISTRIBUTION OF RESIDENTS WITHIN
INCOME PER CAPITAL STRATA . . . .

Figure

2.1.

3.1.

Page

LIST OF TABLES

Table Page

1.1 THE SIZE DISTRIBUTION OF INCOME IN 13 AFRICAN
COUNTRIES . . . . .... 6

2.1 TWO-TIER SAMPLING PROCEDURE: DATA TYPES, INTERVIEW
FREQUENCY, AND SAMPLE SIZES . . .. 21

3.1 COMPONENTS OF NET HOUSEHOLD INCOME . .. 24

3.2 COEFFICIENTS APPLIED TO ESTIMATE THE NUMBER OF
MAN-EQUIVALENT CONSUMER UNITS PER HOUSEHOLD .. 27

3.3 MEAN NET INCOMES BY VILLAGE AND HOUSEHOLD SECTOR
(IN NAIRA) . . . . .. 28

3.4 PERCENT OF NET HOUSEHOLD INCOMES IN CASH OR IN-KIND
BY SECTOR . . . . 30

3.5 AVERAGE AND CUMULATIVE INCOME, NUMBER OF RESIDENTS
AND CONSUMER UNITS BY DECILE. . . 33

3.6 AVERAGE AND CUMULATIVE INCOMES, NUMBER OF RESIDENTS
AND CONSUMER UNITS BY VILLAGE DECILES . .. 36

3.7 THREE SUMMARY MEASURES OF THE SIZE DISTRIBUTION OF
PERSONAL INCOME BY HOUSEHOLD SECTOR AND VILLAGE 39

3.8 ESTIMATED FEMALE EARNINGS GENERATED IN TRADING AND
COMMERCIAL FOOD PROCESSING BY INCOME STRATUM . 42

3.9 PERCENT OF HOUSEHOLD INCOME EARNED IN OFF-FARM
EMPLOYMENT BY VILLAGE AND INCOME STRATUM . .. 44

3.10 NET CASH AND IN-KIND GIFTS PER HOUSEHOLD REPORTED
BY VILLAGE AND STRATUM (IN NAIRA) . . .. 48

3.11 CASH INCOME AS A PROPORTION OF TOTAL HOUSEHOLD
INCOME AND THE SOURCES OF CASH EARNINGS BY SECTOR 50

3.12 TOTAL ANNUAL SUBSISTENCE GRAINS PURCHASES AND SALES
PER HOUSEHOLD BY INCOME STRATUM . . .. 52

Table Page

4.1 MEAN INCOME PER CONSUMER BY SIZE OF HOUSEHOLD AND
AGE OF HEAD FOR NUCLEAR AND EXTENDED FAMILIES
(IN NAIRA) . . . . ... 60

4.2 FREQUENCY DISTRIBUTION OF POOREST 30 PERCENT OF
HOUSEHOLDS BY SIZE OF HOUSEHOLD AND AGE OF HEAD FOR
NUCLEAR AND EXTENDED FAMILIES . .... .61

4.3 HOUSEHOLD DEMOGRAPHIC CHARACTERISTICS BY VILLAGE
AND INCOME STRATUM . . . .... .63

5.1 CULTIVATED LAND HOLDINGS BY VILLAGE AND INCOME
STRATUM . . . .. .68

5.2 RELATIONSHIP BETWEEN OFF-FARM INCOME, HOUSEHOLD
INCOME STATUS, AND HECTARES PER CONSUMER . .. 70

5.3 MEAN FARM INCOME PER CONSUMER BY INCOME STRATUM
AND CULTIVATED HECTARES PER CONSUMER (IN NAIRA) 71

5.4 AVERAGE VALUE OF LIVESTOCK AND WORKING CAPITAL PER
HOUSEHOLD BY INCOME STRATA (IN NAIRA) ...... 75

5.5 AVERAGE DAILY HOURS WORKED PER ADULT MALE (16+
YEARS) ACCORDING TO HOUSEHOLD SECTOR AND INCOME
CLASS, SMALL SAMPLE . . . ... .76

5.6 AVERAGE COSTS AND RETURNS PER HECTARE FOR UPLAND
FIELDS BY INCOME CLASS, SMALL SAMPLE (IN NAIRA) 80

6.1 THE PERCENT OF TOTAL HARVEST VALUE FOR SUBSISTENCE
GRAINS AND CASH CROPS BY INCOME STRATUM . .. 83

6.2 THE HARVEST VALUE OF 12 MAJOR CROPS EXPRESSED AS A
PERCENT OF THE TOTAL HARVEST VALUE BY INCOME
STRATUM . . . .... 85

6.3 THE EFFECT OF CROP MIX ON GROSS MARGINS PER HECTARE
BY INCOME STRATUM (IN NAIRA) . . .... .88

6.4 THE DISTRIBUTION AND SELECTED CHARACTERISTICS OF
OFF-FARM OCCUPATIONS BY HOUSEHOLD INCOME CLASS 92

6.5 AVERAGE RETURNS PER HOUR REALIZED IN 23 OFF-FARM
OCCUPATIONS DISAGGREGATED BY INCOME BIAS CATEGORY,
SMALL SAMPLE (IN NAIRA) . . .... .96

S. the overwhelming need for data on income
distribution is not so much for better data on
income shares, as for better data on the sectoral
distribution of the poor, their occupational char-
acteristics and educational levels, their owner-
ship of productive assets, and their access to key
production inputs. These characteristics deter-
mine the processes of income generation in 'poverty
groups and the constraints on these processes
[Chenery, 1974].

1. INTRODUCTION
A more equal distribution of the gains from economic growth has

emerged as an increasingly prominent development objective during the

1970s. This is reflected not only within the national plans of most

low income countries, but also in mandates guiding the assistance

programs of major external donors [USAID, 1975; McNamara, 1978]. Interest

in distribution reflects a growing awareness that the income gap

separating the rich and poor has widened substantially in all but

a few developing countries during the past two decades. The continuing

presence of substantial pockets of poverty has aroused both humanitarian

concerns and fears of political instability. But it has also become

increasingly evident that in the absence of strong foreign markets the

domestic intersectoral linkages needed for rapid growth cannot be ex-

ploited by policies which result in a further concentration of income

[Mellor, 1976].

In spite of the commitment towards more broadly based growth, efforts

to operationalize equity as a planning objective have been hindered

by insufficient knowledge of how to design policies which ensure broad

participation, how to implement them, and how to measure their impact.

Underlying these policy questions is a general paucity of information

2

on incomes, on the occupational and demographic characteristics of the

poor, and on how the poor respond to and are affected by alternative

development policies.

Although a disproportionate number of the poorest within developing

countries live in rural areas and derive their incomes primarily from

agriculture [Chenery, op. cit.], the problems of the rural poor have been

especially hard to address. Because most rural populations are highly

dispersed geographically, often working within widely varying ecological,

institutional, and market conditions, it has proven difficult to design.

policy instruments which effectively reach more than a small proportion

of the rural poor. Moreover, information on rural incomes are particu-

larly inadequate in almost all low income countries.

The present study was conceived to partially fill this knowledge

gap through an analysis of income in one area of northern Nigeria. During

1974-75, a comprehensive set of household data was collected in three

villages of southwestern Kano State. This paper summarizes some of the

empirical findings of the survey through a description of the levels,

distribution, and structure of income in that region. The paper is

intended to provide Nigerian planners with a better understanding of

who constitute the rural poor, what are their sources of income, and why

they remain in poverty. In a broader context, the paper serves as a case

study of the distribution and structure of personal income within an

essentially traditional society characterized by low population pressure

and by a production system experiencing the first stages of technolo-

gical change.

1.1. Rural Income, Growth, and Changes in Income Distribution

Before turning to an examination of the survey, it is useful to

briefly place the analysis into a broader framework by relating rural

incomes to patterns of national distribution. Numerous authors have

concluded from cross-country evidence that economic growth is accom-

panied by an initial period of increasing national inequality followed

at some point by a tendency towards a more equal distribution [Kuznets,

1955, 1963; Paukert, 1973; Adelman and Morris, 1973a; Ahluwalia, 1976].

A common model put forward to explain this secular trend relies upon

intersectoral income differentials and changes in the structure of the

economy which occur as part of the growth process. The dynamic of the

model is a more rapid growth of personal incomes within the industrial

sector accompanied by a shift of population out of the rural sector into

industrial employment. It can easily be shown that an expanding high

income population within an initially larger but proportionately dimin-

ishing lower income population automatically produces the U-shaped equality

function [Lydall, 1977]. In short, although national inequality is

amplified if incomes are less equally distributed within industry, the

model suggests that the primary cause of national inequality is the

income gap between the agricultural and industrial sectors, rather than

disparities within either sector.

Results of recent decomposition analyses which separate national

inequality into between-sector and within-sector components, however,

have challenged the general validity of the intersectoral model [van

Ginneken, 1976; Fields and Schultz, 1977; Fishlow, 1972]. Among the

developing countries examined, inequality between sectors has typically

been found to explain well under one third of overall national inequality,

with the greatest proportion attributable to factors related to within

sector disparities. Particularly significant is the finding that in a

number of low income countries representing a range of development stages,

inequality within rural areas explains a greater proportion of overall

inequality than either urban or between-sector disparities [van Ginneken,

1977].

These results reflect the combined effect of two sets of factors:

the "pre-growth" distribution of income among traditional farm producers,

and the emergence of economic dualism within agriculture that is, the

growth of small modern agricultural sub-sectors characterized by the

application of new production techniques, within a larger, less productive,

and lower income traditional sector [Oshima, 1975].

Both factors are, of course, closely interrelated. Experience in

countries which have witnessed the introduction of seed fertilizer

technologies has shown that the pattern of adoption is importantly affected

by the existing distribution of resources and incomes. When such tech-

nologies have been introduced in areas already characterized by wide

inequalities, not only has the productivity impact been weak, but the

pattern of inequality has been reinforced [Ruttan, 1977]. If successful

adoption requires increased use of factors which are positively related

to current income (such as human or physical capital), or if access to

modern inputs or extension assistance is influenced by institutional

factors similarly related to income, a skewed traditional distribution

can both retard modern sector expansion and contribute to greater overall

inequality.

These patterns underline the need for detailed knowledge of the

distribution of resources and incomes at the household level. Such

information combined with an understanding of the factor requirements

implicit in new production packages can assist in predicting adoption

patterns and their distributional impact. More important is the

ex ante contribution micro-level data provides in the design of policy

interventions. Understanding the determinants of incomes among tradi-

tional producers or conversely, an identification of constraints

limiting incomes is clearly necessary for the development of appro-

priate packages. And to the extent that constraints vary across income

strata, such knowledge disaggregated by income class can permit a more

efficient targeting of interventions to specific poverty groups. Despite

these considerations, very little micro data documenting rural incomes

and examining households by income strata are available in most developing

countries.

1.2. Income Distribution in Africa

Among the areas of the developing world perhaps least is known about

the size distribution and structure of personal incomes in Africa. The

available data are highly aggregated and have been used primarily to

estimate national averages and to draw comparisons among regions or

industrial categories [Phillips, 1975]. In very few instances are data
available to examine the interpersonal distribution, or changes in dis-

tribution over time. Moreover, coverage is almost exclusively limited

to the modern urban sector.

Table 1.1 summarizes data describing national and sectoral dis-

tributions for 13 countries in sub-Saharan Africa. Because of differences

Table 1.1 THE SIZE DISTRIBUTION OF INCOME IN 13 AFRICAN COUNTRIESa

Country

Botswana

Chad

Benin

Gabon

Ivory Coast

Kenya

Malawi

Zimbabwe

Senegal

Sierra Leone

Tanzaniae

Uganda

Zambia

Gin i

a. With the exception of rural Botswana, Sierra Leone, and Tanzania as
indicated, all data are from Jain [1976].

b. Republic of Botswana [1976].

c. Farm survey results reported in Eponou [1979].

d. From survey of urban migrants reported in Eponou, op. cit.

e. From van Ginneken [1976].

Gini
Coefficient

.5740

.5200

.3687

.4675

.6439

.5342

.6368

.4790

.4696

.6627

.5874

.6117

.3774

.4224

.3030

.3260

.4007

.3978

.3968

.2662

.2716

.5226

Year

1971-72

1974-75b

1958

1959

1968

1970

1969

1968-69

1969

1968

1960

1968-69

1974-75c

1974-75d

1969

1969

1970

1970

1970

1970

1970

1959

Population

Active Population

Household

All

All

Income Recipient

Income Recipient

Income Recipient

Household

Household

Income Recipient

All

Household

Household

Household

Household

Household

African Male Employees

African Male Employees

African Male Employees

African Male Employees

African Male Employees

Household

Coverage

National

Rural

National

National

National

National

National

National

National

National

National

(Excluding Urban Western Province)

Rural

Urban

Rural

Urban

National

Non-Agricultural

Urban

Agricultural

Rural

National

in income concepts, survey methods, and coverage, it is difficult

to derive meaningful cross-country comparisons. However, it is notable

that estimates of distribution in rural areas exist for only four coun-

tries. Among these four countries, two points merit mention. Within

each, rural incomes were less concentrated than the national or urban

distributions. And with the exception of Botswana, the rural Gini co-

efficients are generally low reflecting consistently more equitable intra-

sectoral patterns compared with the rural distributions in most Latin

American and Asian countries [Jain, 1976].

Two factors help explain these low levels of rural inequality. Im-

portant changes in farm production technology have not been widespread

in most African countries. Because the vast majority of producers still

employ essentially traditional cultural practices, wide disparities

in income attributable to technique based productivity differentials are

uncommon. Second, most areas in Africa continue to enjoy access to

surplus land. Thus problems of land tenure which can become most acute

under conditions of land shortage are similarly uncommon. Existing in-

equalities are believed to reflect interregional variation in soils,

climate, and population pressure, location with respect to markets, and

institutional factors affecting access to and cost of production inputs

[ILO, 1972; Heyer, 1975; Essang, 1970].

1.3. Growth and Income Distribution in Nigeria

During the past decade, the Nigerian economy has experienced extremely

rapid aggregate economic growth. Fueled by the expansion of petroleum

exports, between 1965 and 1974 the Gross National Product (GNP) is esti-

mated to have increased at a real annual rate of 8.5 percent, and GNP

per capital at an annual rate of 6 percent (to an average of $280 in

1974) [World Bank, 1976]. Accompanying this growth, income disparities

within Nigeria are believed to have widened substantially. Although

the relative importance of within sector inequalities is not fully

known, the impact of intersectoral differentials is clearly substantial.

During 1964/65, the agriculture sector accounted for 58 percent

of Gross Domestic Product (GDP) and approximately 70 percent of the

active work force [Federal Republic of Nigeria, 1975]. By 1970/71

agriculture's share in GDP had fallen to 36 percent; and by 1974/75

to only 23 percent. The proportion of the labor force employed in

agriculture in the latter period remained high, however, at 64 percent.1

Moreover, the rate of decline of relative incomes of the farm population

has been most rapid during the 1970s. From 1970/71 to 1974/75, because

of a range of factors including bad weather, crop disease, declining

agricultural terms of trade, and excellerated rural-urban migration,

total farm output, in fact, showed a slight fall. This is in contrast to

an average annual growth rate of 21 percent in all non-agricultural sec-

tors combined.2

The available data on income are unfortunately inadequate to mea-

sure the impact of these changes on the national size distribution directly.

As is true elsewhere, most income data in Nigeria are limited to the

modern industrial sectors and have been used to describe income differentials

In contrast, during the same period the petroleum and mining sec-
tor increased its share in GDP from 3 percent to 45 percent, while its
proportion of total employment remained at less than 1 percent.

Although the petroleum and mining sector accounted for a large
part of this growth with an annual rate of increase of 27 percent, pro-
duction in all other sectors (excluding agriculture and petroleum) also
grew at an annual rate of 13.5 percent.

among occupational classes and administrative regions [Teriba and

Phillips, 1971; Aboyade, 1973, 1974]. Far less effort has been directed

at the measurement of size distributions nationally or within produc-

tion sectors. However, the rough magnitude of recent changes in the

national distribution has been estimated by Byerlee [1973] using an

input-output model of the Nigerian economy. Dividing the population

into seventeen production sectors and assuming perfectly equal intra-

sectoral distributions, he calculated a base Gini ratio of .49 on income

per capital. Through a simulation approach, he was further able to pro-

ject the distributional impact of the expanding petroleum sector, as

well as the effects of alternative food and export promotion strategies

through the early 1980s. With development policies unchanged, struc-

tural changes within the Nigerian economy would increase the national

Gini ratio to .64 by 1983. Even assuming the most optimistic national

policies balanced food and export promotion combined with lower non-

agricultural wage rates the Gini ratio was still projected to increase

during the period to .51.1

1.4. Current Policies and Data Requirements

Official concern with the rise in income inequality is clearly

present. In the most recent National Development Plan, the Federal

Government assigned high priority to the development of the agricultural

sector. Furthermore, this commitment was framed within the broader

objectives of interregional and interpersonal equity. This statement

1Substantial public and private sector wage increases, most notably
following the Udoji awards in 1974, have made these projections overly
optimistic.

of national purpose places particular emphasis on the development of

farm policies affecting the northern region of the country where incomes

have traditionally remained lowest. Although major agricultural pro-

grams have been introduced on several fronts, results to data have been

mixed and their impact on income distribution within the farm sector is

not yet clear.1 Moreover, efforts to identify policies and projects

which ensure a favorable distributional impact have been hindered by

a lack of data on rural incomes generally, and more particularly by a

lack of information on the characteristics of the rural poor.
No national rural surveys have been undertaken in Nigeria and only

a few sample surveys have examined the structure of incomes at the vil-

lage level. Consequently, only fragmentary evidence is available. From

data collected between 1966 and 1969 in nine villages representing

three areas of the north, Norman and Pryor L1979] have calculated vil-

lage Gini coefficients ranging between .2648 and .5004 on household

incomes. The average village coefficient was .3608 reflecting a rela-

tively equitable within village distribution. Unfortunately, the res-

pective village data sets were not pooled to provide a broader measure

of distribution to include the effect of between village variation in

mean incomes. The purpose of Norman's studies, however, was to develop

These programs include: (1) a reorganization of the marketing board
system and increased producer prices; (2) introduction of the National
Accelerated Food Production Program which involves distribution of higher
yielding crop varieties through a coordinated package approach; (3) estab-
lishment of several large integrated rural development schemes; (4) invest-
ment in a number of state operated large-scale farms and irrigation pro-
jects; and (5) the construction of agro-service centers distributing sub-
sidized inputs to small farmers under the auspices of Operation Feed the
Nation.

a baseline understanding of farm production systems throughout the

north, not to examine the structure and determinants of personal incomes.

Therefore, while the studies provide accurate estimates of farm incomes

derived from crop production, they did not directly measure incomes

generated in off-farm employment or by females. Nor did they examine

the characteristics and production constraints of households stratified

by income class to permit the identification of policies most relevant

to the needs of the rural poor.

A more focused study of economic inequality was conducted by the

anthropologist Polly Hill [1972] in a single village of the former North

Central State during 1967. Although income levels were not estimated,

through the use of informants Hill classified all farming units into

four groups according to their relative ability "to withstand the shock

of an exceptionally poor or late harvest" [p. 58]. This subjective

classification proved to be a useful framework within which to examine

factors associated with relative poverty and, indirectly, to infer

causal relationships. The limitations of the Hill study, however, are

serious. Since she surveyed only a single village, she was unable to

incorporate locational variables, such as market access and population

density, into her analysis. Only crude farm management data were col-

lected and no direct estimates of incomes were obtained. Indeed, Hill

argues that "it is doubtful whether reliable statistics on income and

expenditure could ever be obtained in a Hausa village".1

1Hill pointed to the following problems: (1) the difficulty of
valuing domestic consumption given wide seasonal variation in grain
prices, (2) the fragmentation of extended families into distinct

This brief overview reveals an urgent need for additional micro

level research on the structure of rural incomes in northern Nigeria.

For the design of policies which address the Plan's objective of more

equitable agricultural growth, information is needed to answer the

following questions: (1) What is the degree of relative income in-

equality at the village level? And what are the most important fac-

tors affecting patterns of distribution? (2) Are there indications

pointing toward more or less concentrated incomes in rural areas as

a result of national development? (3) Is there an important incidence

of absolute poverty at the village level? If so, what are the under-

lying causes? (4) Do sources of income, and patterns of resource use

and productivity vary importantly among rural income strata? And what

does this imply for the design of credit, extension and technology

policies?

The present study attempts to provide empirical evidence on each

of these issues. The paper has been divided into seven sections. Sec-

tion 2 reviews the data collection methodology employed in the survey

and general characteristics of the study area. In Section 3 the levels,

production and consumption units during the dry season, (3) the secrecy
of some income generating activities, and (4) limited access to women
due to wife seclusion thereby restricting information on female earnings.
However,considerable experience in the collection of farm level data
in the north has been accumulated, particularly through the work of
the Rural Economy Research Unit at Ahmadu Bello University. The experi-
ence has suggested approaches which importantly reduce each of these
problems in arriving at an accurate and conceptually valid measure of
income. The present survey design has built on the lessons learned from
these earlier efforts. Furthermore, this study employed a highly inten-
sive data collection approach suggested by Hill but which she believed
would prove too costly. In short, with the exception of the last pro-
blem area, female earnings, her caution was unduly pessimistic. For a
discussion of the female earnings problem see Section 3.

13

distribution and sources of household income are examined by income

class. The demographic structure of households are examined in Section

4 to determine the presence of life-cycle income determinants. Sec-

tion 5 examines patterns of resource use and productivity among income

strata. Selected farm and off-farm activities are analyzed in Section

6 to identify differences in choice of enterprise across income strata

and to infer whether enterprise mix may be a determinant of income

variation. Conclusions and policy implications are summarized in

Section 7.

2. CHARACTERISTICS OF THE STUDY AREA AND SURVEY METHODS

Accurate data on income is extremely difficult to obtain in rural

surveys. This is due both to the complexity of the income concept and

because it is usually considered to be a highly sensitive and thus con-

fidential type of datum. For both reasons it was believed necessary to

employ an intensive cost-route approach.1 Because the cost-route tech-

nique employs frequent interviews, it encourages the establishment and

maintenance of rapport with participating households and reduces mea-

surement error due to poor recall. However, it is extremely expensive

which, given a budget constraint, restricts both the sample size and

geographical scope of the study.

Location can be assumed to affect rural incomes through variation

in the quality of natural resources (soil and climate), as well as

through differential access to support services (extension) and markets.

For the purpose of estimating the distribution and structure of incomes,

as well as to identify determinants, it would be desirable to sample

households displaying some diversity with respect to both sets of fac-

tors. Due to limited resources, however, this strategy could not be

followed in the present study. Rather, villages were selected in an

effort to minimize ecological differences while making it possible to

examine the impact of differences in access to support services and

markets. More specifically, three villages in southwestern Kano State

were purposively selected to satisfy the following criteria: (1) that

The cost-route method involves repeated visits to respondents
during an entire production cycle. During each visit data is obtained
on all relevant activities which occurred since the most recent inter-
view [Spencer, 1972].

the villages should differ significantly with respect to proximity to

major roads and thus to the urban marketing centers of Kano City and

Zaria; (2) that at least one of the villages should be the seat of an

agricultural extension campaign effort; and (3) that the three villages

should be sufficiently close together to control for differences in

soils, climate and farming systems, as well as to allow the survey super-

visor to visit each of the study villages on a daily basis. The three

villages chosen Rogo, Zoza, and Barbeji are shown in Map 2.1.

2.1. Climate and Soils of the Study Area

The villages are located in the Guinea-savannah ecological zone

of Nigeria. One of the primary factors limiting agricultural produc-

tion in this semi-arid region is low and highly variable rainfall.

The study area receives an average annual rainfall of approximately 35

inches distributed over a 120 day period extending roughly from May to

September. During the 1974-75 survey year total rainfall was very nearly

equal to the 50 year mean.

The soils of the study area can be divided into upland soils (tudu),

which comprise over 95 percent of the total land area of the region,

and lowland soils (fadama), which are located near river basins and in

valley bottoms. Whereas upland soils cannot be cultivated in the dry

season unless irrigated, the alluvial fadama soils can often support dry

The limitations of the village selection procedure are clear. The
judgement sampling approach restricts the extent to which population
characteristics can be validly infered for either,Kano State or for the
north of Nigeria. In particular, given the range of ecological condi-
tions displayed in the north, it is expected that the income inequality
observed in the present study would understate the actual inequality of
the region as a whole.

MAP 2.1. DETAILED MAP OF THE
STATE, WITH INSET OF

THREE STUDY VILLAGES IN KANO
NIGERIA

Karoye district
headquarters
5 miles /

BARBEJI

Zaria Kano Road
23 miles

LATERITE ROAD
.................. BICYCLE AND FOOT PATH
I 2 3 4 5 6 7
MILES

SKANO STATE
U STUDY AREA

season farming without supplementary irrigation. The upland soils of

the survey area are generally.well drained and;he'avily leached feruginous

tropical soils with chemical properties, which make them poorly suited to agri-

cultural use. Although a farming system which includes frequent bush
fallow and organic manuring can maintain an adequate level of soil fer-

tility, both the frequency of fallow and the amount of organic matter

replacement necessary to: maintain soil nutrient balance greatly exceed

observed practices. While uncqutivated.plots of land were present in

each of the study villages, the practice of incorporating a fallow period

into a regular pattern of crop rotation was not common.1

A soils survey conducted in the three villages concluded that there

were no significant intervillage differences in soils characteristics

which would affect upland productivity. The population density of the

survey area was approximately 130 persons per square mile.

2.2. Characteristics of the Study Villages

In spite of the ecological homogeneity of-the study area, substan-
tial intervillage variation was observed in both the sources and levels
of incomes. To understand the factors underlying these patterns, it is

necessary to review the characteristics of the three villages:

1. Rogo is a relatively large village (population 6405),2 and is
the location of one of the two most important village markets in Karaye

10n 80 percent of all fields cultivated by sample farmers during
the survey period, no fallowing had occurred since the field had been
acquired by the current owner. For the remaining fields on which fal-
lowing had occurred, the mean period since the end of the most recent
fallow was 8.9 years.
population estimates have been taken from the 1963 census.

District. Closely tied to external urban markets by daily lorry traffic

throughout the year, the village was served by a resident agricultural

instructor and several representatives of licensed buying agents pur-

chasing groundnut. Strongly market oriented, Rogo farmers planted

nearly three times the amount of groundnut seed relative to total culti-

vated hectares than did farmers in each of the other two villages. The

largest plantings of sugar cane were also observed in Rogo reflecting

the relatively larger holdings of fadama, 48 percent of the three vil-

lage total. Pressure on the land was high, with a cultivated land per

capital ratio of .24 hectares.

2. Zoza, a smaller village (population 2964) located six miles

north of Rogo, is situated within one mile of the major laterite feeder

road in the district. Lorry connections were infrequent during the sur-

vey year. One licensed buying agent's representative was a resident

of the village. The Rogo agricultural instructor had worked with Zoza

farmers most recently during 1973 when a package of improved groundnut

seed and fertilizer was distributed as a part of a state-wide seed

muliplication program. Cropping patterns were least cash oriented of

the three villages with the highest relative plantings of sorghum and

millet. Population pressure was the lowest, reflected in a land per

capital ratio of .47 hectares.

3. Barbeji is intermediate in size (population 3744), and located

13 miles from the nearest all season road. Connecting trails were motor-

able with great difficulty during the dry season and impassable to any

four-wheel motor traffic during the rains. Lorry contact was consequently

rare with cash crops evacuated by headload, bicycle, and donkey. Although

smaller than that of Rogo, the Barbeji market is considerably larger

than that of Zoza serving several satellite villages and hamlets.

Neither an agricultural instructor nor a licensed buying agent or repre-

sentative had worked in the village in recent years. Population pres-

sure was intermediate with a cultivated land per capital ratio of .45

hectares.

2.3. Sampling and Survey Methods

The sample frame consisted of all household heads included on

recently updated tax lists. Forty-five households were randomly selected

from such lists in each village. The household was defined as those

persons "eating from the same pot" (that is, sharing a common source of

food), a convention commonly used in surveys conducted among the Hausa.

An additional six households were purposively selected on the basis of

elite status they enjoyed in the study villages.1 This latter group

was included in the survey to permit an analysis of how political posi-

tion affects incomes as well as access to government services.

It was assumed that the types of data required vary considerably

both with respect to the rate of memory loss and with respect to the

sample size necessary for different types of analysis. Due to limited

resources a two-tier sampling procedure was employed. From the results

of a situational survey administered to all selected households, the

1The non-random units include the village heads in two of the sur-
vey villages, a hamlet head in one village, and the head farmer (sarkin
noma) in each village. For all subsequent analysis, these elite house-
holds are separated from the random sample and identified as a distinct
sub-set. For a discussion of the positions of village head and sarkin
noma, see Hill [1972, pp. 295 and 316].

general sample in each village was divided into "large sample" (between

33 and 35 households per village) and "small sample" (either 11 or 12

households per village) groups. The interview frequencies employed

for each sample and data type are summarized in Table 2.1.

Harvest weights of all crops as measured in local units were obtained

from the small sample during the twice weekly interviews. Threshing

percentages and size of land holding were also obtained through direct

measurements made during supplemental farm visits. Seasonal retail

prices of all crops grown in the area were obtained in monthly surveys

conducted in each village market.

2.4. Characteristics of the Farming Households

The sampled farming units were generally representative of house-

holds throughout the northern region of Nigeria. The average household

consisted of 6.7 persons holding usufructurary rights over 2.5 hectares

of cultivated land.2 Although nearly 40 different crops were grown in

the area, the basic food staples, millet and sorghum, together with

the dominant cash crop, groundnut, represented 75 percent of the total

harvest value.

The small sample households were chosen based on a four-cell strati-
fication matrix: (1) above and below mean land to worker ratios, and (2)
use or non-use of both chemical fertilizer and seed dressing during the
previous year. The approach was designed to ensure observations in the
small sample with sufficient variation in these key production variables
to increase estimation precision in the agricultural production analysis.
Nine households were chosen for each cell of the stratification matrix.
One small sample farmer was subsequently dropped from the survey reducing
the sample size to 35.

2The average family size found by Norman [1974] in the three village
Zaria study was 6.9 persons cultivating 3.5 hectares. In her Batagarawa
survey Hill [1972] found an average household of 7.2 persons farming 2.6
hectares.

21

Table 2.1 TWO-TIER SAMPLING PROCEDURE: DATA TYPES,
INTERVIEW FREQUENCY, AND SAMPLE SIZES.

Interview Frequency

Small sample Large sample
Type of Data 2-3 weekly Weekly Monthly Once Monthly Once

A. Agricultural

1. Family labor X
2. Hired labor X X
3. Non-labor inputs X X
4. Harvests X X
5. Non-labor input X X
purchases
6. Crop and livestock X X
purchases (trading)
7. Crop and livestock sales X X
8. Land transfers X X
9. Transport costs X X
10. Assets inventory X X

B. Non-farm occupations

1. Off-farm labor X X
2. Service earnings X X
3. Purchases X X
4. Sales X X
5. Assets inventory X X

C. Other flows

1. Consumer expenditures X X
2. Cash and kind loans X X
given, rec'd, repaid
3. Cash and kind gifts X X
given and received
4. Labor migration X X

Number of households

Village

Rogo
Zoza
Barbeji

Small sample Large sample

11 34
12 37
12 34

The technology of the local farming system was essentially tradi-

tional with only limited use of modern inputs. Chemical fertilizers.

were applied during the survey year by 40 percent of the sampled house-

holds, typically at well below recommended levels. Pre-planting seed

treatment was used by 24 percent. Tractor cultivation was practiced

by only one household in the random sample. None used animal traction.

An improved groundnut variety, highly mixed with traditional varieties,

was sown by nearly all of the sample households. However, the yield

advantage of this improved groundnut variety was minimal, only 10 to

15 percent greater than local varieties on farmers' fields.

Average stocks of farm tools and equipment were valued at less then N9

replacement cost. Average variable costs per farm (all costs, both

cash and in-kind, excluding household labor and land) totalled nearly

N65, of which two-thirds, or N43, was recorded as a cash expense. Aver-

age variable costs per hectare were approximately N26. The largest

single cash expense, accounting for N31, paid for the hiring of non-

family labor. Approximately 60 percent of farm labor was provided by

household members.

1The official foreign exchange rate during 1974 was N1 = US $1.64.

3. LEVEL, DISTRIBUTION, AND COMPOSITION OF INCOME

3.1. Definition of Farm Family Income

Although there is some variation according to family structure,1

the household generally constitutes the primary production and consump-

tion unit throughout rural Hausaland. Moreover, since most major deci-

sions in both production and consumption activities are made by the

household head (mai gida), the farm family was chosen as the most appro-

priate income recipient unit. With one exception discussed below, the

survey obtained information on incomes generated by all family members

in all enterprises. The components of aggregate household income are

presented in Table 3.1.

The pricing procedures applied to evaluate those components which

did not involve cash transactions are discussed in Appendix A. Although

data on cash and in-kind gifts transfers were collected, the value of

such flows were not included as income components. Unrealized capital

gains which arose from the re-evaluation of owned assets during the sur-

vey period were also excluded. The twelve month period over which net

flows were calculated was delimited by the annual agricultural cycle to

capture one complete season.

Households were organized either as nuclear families (iyali) or as
extended units (gandu). Gandu units can be defined as households which
include two or more male adults, often married, with their wives and
children. The gandu unit is typically paternal or fraternal, that is
headed by the father or brother of the other members, though other
arrangements do occur. Understood in the institution of gandu are a
set of rights and obligations between members, primarily regarding the
common production and sharing of a portion of the household's food.
Adult males in gandu, however, have the right to farm their own fields
(called gayaunna) over which these individuals, not the gandu head,
control both planting and disposal decisions. Non-agricultural occu-
pations pursued by other adults in gandu also generally fall outside
the control of the gandu head.

Table 3.1 COMPONENTS OF NET HOUSEHOLD INCOME

Cash payments

Plus: 1. Trees rented out or sold
2. Land rented out or sold

Hired farm labor
Costs transporting crops for sale
Purchases of seed dressing, in-
secticide and fertilizer
Purchases and repairs of farm
tools and storage bins
Rental of work animals
Land rented and purchased
Trees rented and purchased
Interest paid on farm loans re-
paid in cash

1. Off-farm service earnings
2. Sale of purchased crops (trading)
3. Sale of purchased animals and
animal products (trading)
4. Sales of tools and fertilizer
purchased or produced
5. Sale of non-agricultural items
purchased or produced
6. Interest on non-farm loans ex-
tended repaid in cash

1. Purchases of crops for later
resale
2. Purchases of insecticide, seed
dressing, tools, and fertilizer
for later resale
3. Purchases of animals and animal
products for later resale
4. Purchases of non-agricultural
trading items
5. Costs transporting all traded
items
6. Interest paid on non-farm loans
received
7. Depletions in inventories of
Stranded crops, animals, animal
products, and non-agricultural
items

Less: 1.
2.
3.

4.

5.
6.
7.
8.

Imputed value

Farm sector

1. Kind payments for purchases of crops,
animals and animal products

Net household income = Net cash + Net imputed value

Field and tree crops harvested
Kind payments for land rented out or sold
Kind payments for trees rented out or sold

Seeds and cuttings planted
Fertilizers applied
Kind payments to hired labor
Kind payments for purchases and repairs
of tools
Kind payments for rental of work animals
and corralling
Kind payments for land rented and pur-
chased
Kind payments for trees rented and pur-
chased
Interest paid on farm loans repaid in
kind

Kind payments for off-farm service work
Interest on non-farm loans
Increases in inventories of traded
crops, animals, animal products,
and non-agricultural items

Off-farm
sector

Plus:

Less:

Net household income = Net cash

+ Net imputed value

The household "farm-sector" has been defined in Fable 3.1 to include

only those activities related to field and tree crop farming in order

to better identify the reliance of households on their own crop produc-

tion. Activities involving the purchase for resale of livestock and

animal products and purchase for resale of crops (that is, trading acti-

vities) have been assigned to the off-farm sector. Similarly, work as

hired agricultural labor has also been included in the off-farm sector.

The only major source of income not recorded is that earned by

women in trading activities. Due to the Moslem custom of secluding

married women of childbearing age within their compounds, male enumera-

tors were denied access to women engaged in food processing and petty

trading activities. Further, household heads displayed a reluctance to

discuss costs and returns of such female occupations. An accurate esti-

mate of such earnings could only be obtained through an additional team

of female enumerators, an expense which exceeded the project's resources.

Payments received by women working outside the compound as pickers in

the fields of other households are included, however. These data were

generally known to the household head and were easily obtained. The

effect of excluding female incomes generated in trading and commercial

food processing is discussed later in this section.

3.2. Man-Equivalent Consumer Units

In order to make meaningful interpersonal comparisons it is neces-

sary to adjust household income to take account of variation in size

and composition of household membership. Three types of adjustments

are possible. The first simply involves converting each household

income figure to a per capital measure. A second, but rarely applied

adjustment, involves consideration of possible economies of scale in

consumption. To the extent that such economies exist, smaller house-

holds would require greater income per capital to realize any given living

standard. Due to difficulties in estimation [Kleiman, 1966] and recent

evidence which indicates that such economies are probably of relatively

small magnitude among rural African households [King and Byerlee, 1977],

this latter correction for household size has not been made.

The third adjustment is to correct for variation in the age and

sex composition of households. The use of consumer-equivalent scales

has been thoroughly treated in the literature on household budget

studies [Woodbury, 1944; Prais and Houthaker, 1955; Kleiman, 1966].

Several methodological problems are confronted in deriving appropriate

conversion coefficients. Theoretically a unique conversion ratio is

required for each major group of consumption items, income stratum, and

type of consumer group (urban, rural, farm, non-farm, etc.). And in the

absence of highly detailed consumption information, few objective criteria

are available for demarcating appropriate age-sex classes.

Despite these problems, incomes have been converted to a consumer

man-equivalent base in this study. Since the study villages are rela-

tively homogeneous (in spite of their locational differences), all sampled

households were engaged in farming, and the observed range in incomes

was not exceptionally large, the problems cited above are not believed

to be sufficiently important to invalidate the approach in the present

study. Moreover, since food constitutes the largest single component of

consumption across all income strata, tables of caloric needs provide a

first approximation for constructing such a scale.

The coefficients used to calculate the number of consumer man-

equivalents per household are shown in Table 3.2. Derived primarily from

Table 3.2 COEFFICIENTS APPLIED TO ESTIMATE THE NUMBER OF
MAN-EQUIVALENT CONSUMER UNITS PER HOUSEHOLD

Age

0-4 5-9 10-15 16+

Male .2 .5 .75 1

Female .2 .5 .7 .75

the standard calorie requirements for each age and sex group as suggested

by the F.A.O. [1957], additional marginal adjustments were made on the

basis of the author's knowledge of within household sharing patterns for

consumer goods and of work allocation by age and sex.

The resultant income per consumer man-equivalent has been used through-

out the study to stratify households into income classes. In order to

facilitate comparisons with other studies per capital figure are also pre-

sented where relevant.

3.3. Mean Income Levels by Village and Household Sector

Table 3.3 presents average incomes per household, per capital, and

per consumer disaggregated by village and source as calculated for the

random large sample. The average household generated an annual income

of nearly N350, or approximately -52 per capital 1 Household income was

This compares with a mean household income of nearly 4206, and a
per capital income of N31 found by Norman [1972] in his 1966 Zaria area

Table 3.3 MEAN NET INCOMES BY VILLAGE AND HOUSEHOLD SECTORa,b (IN NAIRA)

Per household Per capital Per consumer

Village Farm Off-farmd Total Farm Off-farm Total Farm Off-farm Total

Barbeji 273.64 85.00 358.64 41.46 12.88 54.34 61.71 19.17 80.88
(23.7%)

Zoza 239.50 79.22 318.72 42.05 13.90 55.95 61.41 20.31 81.72
(24.9%)

Rogo 231.69 130.03 361.77 29.66 16.66 46.30 42.20 23.69 65.89
(36.0%)

All 248.95 97.52 346.47 37.21 14.58 51.79 54.17 21.22 75.38

a. Incomes per capital and per consumer have been calculated as weighted averages.

b. The components of each sector's income estimate are presented in Table 3.1 earlier.

c. Consumer man-equivalents were computed by applying consumption weights to each resident on the
basis of the person's age and sex. The weights used, representing approximate caloric require-
ments, are shown in Table 3.2.

d. The percentage of off-farm income in total income is included in parentheses below.

highest in Rogo, the largest village, and lowest in Zoza, the smallest.

For both income per capital and income per consumer measures, however,

these village rankings are reversed due to intervillage differences in

mean household size. In aggregate, off-farm income constituted 28 percent

of net earnings.1 Off-farm earnings were most important in the largest and

most accessible village, where they constituted 36 percent of total

income, and least important in the most remote village, Barbeji, at 24

percent.

A breakdown of income by type (cash or in-kind) is presented in

Table 3.4. To calculate the proportions of cash and in-kind income, sales

of field and tree crops were netted out of the imputed "in-kind" values

of total harvests and assigned to the cash income side. All in-kind

payments earned in off-farm occupations which were subsequently sold

were similarly netted out of in-kind incomes and included as cash earnings.

The relatively high degree of monetization of the surveyed farmers

is reflected in the fact that 53 percent of income was earned or converted

into cash. Moreover, important intervillage differences underlie this

total. Rogo farmers, enjoying the most advantageous market location as

well as the largest proportion of lowland soils, generated 67 percent of

their income in cash. In contrast, farmers in both Barbeji and Zoza

generated less than half in cash, 48 and 42 percent, respectively. The

sale of crops contributed less than half of all income earned in cash.

study. The results of the two studies are nearly indentical given the
annual rate of inflation of 8 percent experienced during the period.

In comparison, in the three village Zaria study Norman [1972] found
the following income composition: farm production 62 percent; off-
farm enterprises (excluding livestock) 20 percent; and livestock 18
percent.

Table 3.4 PERCENT OF NET HOUSEHOLD INCOMES IN CASH OR IN-KIND BY SECTOR

Village
Income by type
and sector All Barbeji Zoza Rogo

Total 100.0 100.0 100,0 100.0

Cash 52.5 100.0 47.7 100.0 41.9 100.0 67.2 100.0

Farm 44.9 49.3 40.2 44.3

Off-farm 55.1 50.7 59.8 55.7

In-kind 47.3 100.0 52.3 100.0 58.1 100.0 32.8 100.0

Farm 95.1 92.5 95.4 98.6

Off-farm 4.9 7.5 4.6 1.4

This underscores the importance of off-farm occupations which supplied

between 50 and 60 percent of household cash earnings among the three

villages.

3.4. The Size Distribution of Incomes

Summary measures of inequality are normally employed either to

rank a set of populations in order of the degree of incomes concentration

or to compare the ex ante and ex post income distributions observed in

a given population following the introduction of a particular policy or

set of policies. There are, however, many attributes of inequality,

attributes which some summary indices reflect better, or are more sensi-

tive to, than others [Champernowne, 1974]. For example, one can distin-

guish among distributions which display either inequality due to extreme

wealth or inequality due to extreme poverty. Conclusions as to whether

one empirical distribution is more or less equitable than another pre-

supposes knowledge of some social welfare function against which the
alternative distributions can be objectively compared. However, because

most summary indices already embody a concept of social welfare in their

mathematical formulation they are biased measurement instruments.

Because of the demonstrated selectivity of various measures to each

type of inequality a combination of approaches which communicate distinct

aspects of the underlying distributions has been used in this paper.

First, the large sample households are disaggregated into deciles. The

average income earned by households in each decile is displayed along

with the cumulative percentage of incomes, residents, and consumer units.

Second, the frequency distribution of residents among discrete income

per capital strata is shown in histograms for each village and for the

combined stratafication. Third, three summary indices are computed for

each measure of income, and for incomes generated in the farm and off-

farm sectors separately.

Table 3.5 presents average income per household, per resident, and

per consumer unit for each decile in the total large sample. Similar

statistics are also shown for the six purposively selected elite house-

holds. Households were distributed among deciles by arraying the large

random sample according to income per consumer, then allocating the poorest

10 percent of the households to the first decile, the second poorest 10

percent of households to the second decile, and so on. Since there are

exactly 100 randomly selected households in the large sample, 10 house-

holds constitute the sample in each decile for the three village total.

Decile assignments within each village were accomplished similarly.

By international standards and compared with the estimated concen-

tration of income in Nigeria as a whole, these figures reflect a decidedly

equal distribution. Examining the tails of the distribution, roughly

the poorest quarter of the population (26.3 percent included in the bottom

two deciles) earned nearly 12 percent of all income, compared with the

richest quarter of the population (included in the top three deciles) which

1Because incomes were not identically distributed within each village
and because mean levels of income varied among villages there is unequal
village representation within each decile of the combined stratification.
Thus, the same household might be assigned to decile two in the combined
stratification but to decile one in its village distribution if the poorer
farmers in that village had higher incomes than similarly ranked farmers
in the other two villages. For this reason the statistics calculated for
the three village aggregate within a particular stratum are neither a simple
nor weighted average of the respective village statistics for that same
strata, and may in fact lie outside the range of the village specific sta-
tistics shown for the corresponding decile. It is also important to note
that due to differences in family size, each decile does not contain exactly
10 percent of the large sample population.

Table 3.5 AVERAGE AND CUMULATIVE INCOME, NUMBER OF RESIDENTS
AND CONSUMER UNITS BY DECILE

Average
Average Average Average Average number
income Cumula- income income number Cumula- of con- Cumula-
per tive % per per of resi- tive % sumers tive % Number of
household of capital consumer dents per of resi- per of con- observations
Decile (N) income (N) (N) household dents household sumers Barbeji Zoza Rogo All

1 177.73 5.1 19.12 27.78 9.3 13.9 6.4 13.9 3 2 5 10

2 234.21 11.9 28.22 39.77 8.3 26.3 5.9 26.7 4 2 4 10

3 231.27 18.6 34.01 49.21 6.8 36.5 4.7 36.9 3 4 3 10

4 328.93 28.1 42.72 61.25 7.7 48.0 5.4 48.6 3 1 6 10

5 247.33 35.2 52.62 71.69 4.7 55.0 3.5 56.1 2 6 2 10

6 385.80 46.3 55.91 82.26 6.9 65.3 4.7 66.3 4 6 10

7 404.78 58.0 63.25 89.95 6.4 74.9 4.5 76.1 5 3 2 10

8 433.96 70.5 72.33 105.59 6.0 83.9 4.1 85.0 3 2 5 10

9 394.01 81.9 87.56 125.48 4.5 90.6 3.1 91.8 5 2 3 10

10 626.59 100.0 99.46 168.46 6.3 100.0 3.7 100.0 3 5 2 10

Elites 2715.59 139.26 208.89 19.5 13.0 2 3 1 6

earned 42 percent of all income. A comparison of the poorest and rich-

est deciles shows that the poorest 13.9 percent of the population re-

ceived 5.1 percent of all incomes, whereas the most wealthy 9.4 percent

earned 18.1 percent. The ratio of average incomes per capital between

extreme deciles is also not wide, only 5:1. Moreover,it is important

to note that because incomes were not highly concentrated and varied

around a low overall mean level,all of the income strata were poor by

national standards. Thus, even among the households included in the

richest decile of the random sample, the mean per capital income (N99)

was less than 60 percent of the national average (4-171).

The elite households represent a clearly atypical subset of the

population. Extremely large, these six units were composed of nearly

twenty residents per household, compared with the random sample average

of less than seven. They were also economically atypical with mean

household income nearly eight times greater than average, and income

per consumer four times larger. Nevertheless, it should be noted that

the mean income per capital of this group of rural elites was still only

four fifths of the national average.

The following comparisons help place the observed distribution
into a broader perspective: In an examination of data from the 1950s
and 1960s, it was found that among the developing countries surveyed the
average income share of the poorest 40 percent of the population was
only 12.5 percent, compared with 16 percent and 25 percent among developed,
non-socialist and socialist countries respectively. Among African coun-
tries the following income shares of the poorest 40 percent were esti-
mated: Kenya (1969) 10.0%; Sierra Leone (1968) 9.6%; Senegal (1960)
10.0%; Ivory Coast (1970) 10.8%; Dahomey (1959) 15.5%; Tanzania (1967)
13.0% Zambia (1959) 14.5%; Chad (1958) 18.0%; Niger (1960) 18.0%;
Uganda (1970) 17.1% [Chenery, 1974 pp. 8-9]. Adelman and Morris [1973b]
estimate a comparable figure for Nigeria of 14.0%, though they do not
indicate the year for which the data are based.

3.5. Intervillage Comparisons

In Table 3.6 a modified tableau is presented disaggregating incomes

by village. Equality comparisons between villages are facilitated through

the addition of an equity index in the last column. The equity index

has been calculated by dividing the income share of each decile by its

share of the population thereby standardizing for intervillage differences

in household size across deciles. A value of one represents perfect

equality. Values tending toward zero represent disproportionately low

shares of income earned by those strata, while values greater than one

reflect shares of income exceeding an equitable allocation. It is ap-

parent from the equity index that income was in general more equally

distributed in Zoza throughout the income range. Barbeji, the most

isolated village, showed greater inequality in the extreme lower income

range, while Rogo, the largest village with the most favorable market

location, was somewhat less equal in the upper income strata.

These relationships can also be seen in Figure 3.1. All villages

display distributions which are positively skewed to the right as would

be expected in a population where mean earnings do not greatly exceed

a minimum subsistence level. The Zoza distribution is more peaked in

the median range, confirming indications from its equity index. In con-

trast, both Barbeji and Rogo show significantly higher proportions of

residents in the under N20 category, 7.8 percent and 18.8 percent, res-

pectively. Considering its low mean income, the Rogo distribution also

has a relatively high proportion of population in its right tail re-

flecting inequality due to disparities in the high income range.

Table 3.6 AVERAGE AND CUMULATIVE INCOMES, NUMBER OF RESIDENTS
AND CONSUMER UNITS BY VILLAGE DECILES

Average Average Average Average
Number income income income number of
of per Cumulative per per residents Cumulative
Village house- household % of capital consumer per % of Equity
decile Village holds (N) income (N) (4) household residents index

Barbeji 4 233.72 7.5 21.84 32.02 10.7 18.6 .40
1 Zoza 4 188.20 7.1 27.68 36.90 6.8 14.4 .49
Rogo 3 105.71 2.7 17.62 23.49 6.0 7.2 .38

Barbeji 3 120.00 10.4 24.00 40.00 5.0 25.1 .43
2 Zoza 3 238.28 13.8 34.03 47.64 7.0 25.5 .60
Rogo 3 346.38 11.7 21.65 31.49 16.0 26.4 .47

Barbeji 3 173.85 14.6 36.99 49.67 4.7 31.2 .69
3 Zoza 3 191.76 19.2 44.60 58.11 4.3 32.4 .78
Rogo 3 238.69 17.9 28.76 39.78 8.3 36.4 .62

Barbeji 4 364.06 26.2 38.32 62.77 9.5 47.6 .71
4 Zoza 4 279.28 29.7 50.78 69.82 5.5 44.1 .90
Rogo 4 266.17 27.1 33.27 49.29 8.0 49.2 .72

Barbeji 3 398.27 35.7 56.90 81.28 7.0 56.7 1.04
5 Zoza 4 344.74 42.7 53.04 78.35 6.5 57.9 .94
Rogo 3 332.33 35.7 43.16 61.54 7.7 58.4 .93

Barbeji 4 305.21 45.4 67.82 84.78 4.5 64.5 1.24
6 Zoza 3 273.97 50.4 54.79 85.62 5.0 65.9 .96
Rogo 3 311.95 43.8 54.73 66.37 5.7 65.2 1.19

Barbeji 4 396.73 58.0 62.97 92.26 6.3 75.3 1.17
7 Zoza 3 307.69 59.1 61.54 90.50 5.0 73.9 1.09
Rogo 3 555.62 58.1 59.74 85.48 9.3 76.4 1.28

Barbeji 3 481.26 69.5 76.39 114.59 6.3 83.5 1.40
8 Zoza 3 404.83 70.4 71.02 115.67 5.7 82.9 1.27
Rogo 4 507.34 75.6 69.50 99.48 7.3 88.0 1.51

Barbeji 4 372.68 81.4 93.17 128.51 4.0 90.4 1.72
9 Zoza 3 474.32 83.9 89.49 139.51 5.3 91.4 1.58
Rogo 3 343.33 84.4 79.84 122.62 4.3 93.0 1.71

Barbeji 3 780.96 100.0 106.98 185.94 7.3 100.0 1.96
10 Zoza 3 532.14 100.0 100.40 166.29 5.3 100.0 1.76
Rogo 3 593.58 100.0 104.14 156.21 5.7 100.0 2.26

a. The Equity Index has been calculated for each decile
in each village to its share of the village sample.

as the ratio of its share of total earnings

FIGURE 3.1.

THE PERCENTAGE DISTRIBUTION OF RESIDENTS
WITHIN INCOME PER CAPITAL STRATA

ROGO

ZOZA

37.8%

3.2%

BARBEJI

ALL VILLAGES

LESS 20- 40- 60- 80- 100- 120-
THAN :9.99 5999 79.99 99.99 119.99 139.99
20
Income per capital (in Naira)

A set of summary measures describing the size distribution of income

is presented in Table 3.7. Three measures have been calculated, the

Gini ratio, the coefficient of variation, and the standard deviation of

the natural log of income. Each has been selected due to its sensiti-

vity to various types of inequality. The coefficient of variation is

particularly effective in discriminating among distributions where

weight is given to differentials in the high income range. In contrast

the log measure gives greater weight to incomes in the lower range and

is thus more appropriate for purposes of ranking where priority is

given to the incidence of extreme relative poverty. The most commonly

used index, the Gini ratio, is more sensitive to differentials in the

broad middle income range. To facilitate comparisons two values are

given for each index. Presented first is the absolute value of each

coefficient. Second, and written in parentheses, each coefficient has

These measures of distribution are defined as follows:

Coefficient of Variation

V
u

Standard Deviation of the Natural Logarithm of Income

o [log (,)]2 f (y) dy

Gini Coefficient

n n
(1/2 n2 u) Z lyi Yj
i=l j=l

where,
v = standard deviation of income,
u* = harmonic mean of income,
y = an income observation,
yi = income of observation i,
y. = income of all other observations j,
n = number of observations.

Table 3.7

THREE SUMMARY MEASURES OF THE SIZE DISTRIBUTION
OF PERSONAL INCOME BY HOUSEHOLD SECTOR AND VILLAGE

Standard deviation
Income Gini Coefficientaof of the natural log
measure Village coefficient variation of income

Total income
per household

Total income
per capital

Total income
per consumer

Farm income b
per capital

Off-farm
come per
capital

Non-agricul-
tural income
per capital

a. In parentheses each measure has been standardized on a scale between
zero and one. Zero represents perfect equality and a value of one
represents perfect inequality.

b. Farm income is the net income obtained from field and tree crop pro-
duction.

c. Non-agricultural income is equal to off-farm income less earnings
obtained through hired farm labor employment.

Barbeji
Zoza
Rogo
All

Barbeji
Zoza
Rogo
All

Barbeji
Zoza
Rogo
All

Barbeji
Zoza
Rogo
All

Barbeji
Zoza
Rogo
All

Barbeji
Zoza
Rogo
All

.3426
.2624
.3176
.3156

.2898
.2251
.3034
.2823

.2899
.2691
.3034
.2947

.3298
.2108
.3504
.3183

.4588
.5562
.5464
.5306

.5574
.6759
.5775
.6097

.6553
.5179
.6381
.6113

.5143
.4142
.5558
.5052

.5432
.4872
.5867
.5490

.5923
.3835
.6475
.5718

.9502
1.0660
1.1717
1.1014

1.1751
1.2948
1.2376
1.2707

(.4584)
(.4084)
(.4535)
(.4450)

(.2098)
(.1464)
(.2360)
(.2033)

(.2278)
(.1918)
(.2561)
(.2316)

(.2604)
(.1282)
(.2954)
(.2464)

(.4745)
(.5319)
(.5786)
(.5481)

(.5800)
(.6265)
(.6050)
(.6176)

.637
.508
.638
.586

.566
.423
.555
.535

.544
.504
.547
.544

.636
.395
.653
.619

1.111
1.616
1.229
1.323

1.208
1.730
1.228
1.406

(.2886)
(.2055)
(.2895)
(.2559)

(.2426)
(.1518)
(.2355)
(.2225)

(.2284)
(.2026)
(.2303)
(.2284)

(.2880)
(.1350)
(.2989)
(.2770)

(.5524)
(.7231)
(.6017)
(.6364)

(.5923)
(.7496)
(.6013)
(.6641)

been standardized such that zero equals perfect equality and a value

of one equals perfect inequality. The Gini coefficient is already s6

standardized.

The Gini coefficient for income per capital computed for the entire

village sample is .2823. Village coefficients range between .3034 in

Rogo and .2251 in Zoza. Overall these are relatively low values re-

flecting somewhat greater equality than the results reported by Norman

for other areas in northern Nigeria. All three indices rank incomes in

Zoza as the most equally distributed whether measured on a household,

per capital, or per consumer base. The changes in village rankings when

applying different measures, however, should be noted. The coefficient

of variation ranks Rogo as less equal compared with Barbeji. These

rankings are reversed when using the standard deviation of the logarithm

of income. The switch in rankings accurately captures the relatively

greater inequality in the extreme high income range in Rogo compared with

the inequality among lower income households found in Barbeji.

Within each village and overall, household incomes were less equally

distributed than income per resident or per consumer. This is to be

expected if household income and family size are positively correlated.

The very minor differences in the degree of inequality between income

Standardized values have been calculated for the other two measures
as follows:

1. Coefficient of Variation: ()/[()2 + 1]
u u
2. Standard Deviation of Ln Income: (V1nY)2/[VlnY)2 + 1]

where

V = standard deviation,
u = mean income,
Y = income.

per resident and income per consumer give a preliminary indication that

variation among income strata with respect to family composition is

probably not great.

Farm and off-farm incomes considered individually were less equally

distributed than their total. This is reflected in Gini ratios of

.5306 and .3183 for off-farm and farm incomes per resident, respectively,

compared with .2823 for their aggregate. This points toward a degree

of household specialization between these two sectors. When off-farm

income earned through hired farm labor is deducted, non-agricultural

earnings display an even greater degree of inequality, reflected in a

Gini coefficient of .6097.

3.6. Female Earnings in Trading and Commercial Food Processing

Although data on female earnings in nonfield work were not directly

obtained in the survey, information on female participation in all such

activities was obtained. By combining these data with information on

returns to women's occupations obtained through secondary sources, a

rough estimate of female incomes can be calculated and the effect of

excluding this income source can be assessed. Given the most reasonable

Twice during the year household heads in the present survey pro-
vided information on which women in the household were active in any
income-earning occupation, the types of occupations each woman pursued,
and during what part of the year each woman was active in each activity.
In an intensive survey of female occupations conducted during 1969/70
in three villages near Zaria, Simmons L1976] estimated that the average
monthly return to all occupations was N2.14. Given a 31 percent period
rate of inflation (derived from the difference in mean food grain prices
observed in the 1969/70 survey villages and the current year prices
observed in the present survey villages), a mean monthly return per
occupation of N2.80 was applied to the reported female employment pat-
terns of the present survey to estimate annual female earnings. For a
more detailed discussion of methodology see Matlon [1978].

assumptions regarding the intensity with which women worked, it is esti-

mated that females contributed an average of *78 to household incomes.

If added to the predominantly male-generated incomes reported above,

this would represent an increment of 23 percent.

Particularly interesting is the distribution of estimated female

earnings among income strata shown in Table 3.8.1 Because females in

Table 3.8 ESTIMATED FEMALE EARNINGS GENERATED IN TRADING
AND COMMERCIAL FOOD PROCESSING BY INCOME STRATUM

Dec

Variable 1

Average number of 37
occupation-months
per household

Average annual 103
female earnings
per household
(in Naira)

Female income as a 58
percent of predominantly
male income

a. Occupation-months represent the
by all females in the household
pation was pursued.

ile

2

29

Quintile

2 3 4

31 27 30

Decile

9 10

21 19

80 87 76 84 59 52

34 31 24 20 15 8

total number of occupations worked
multiplied by the months each occu-

lower income households tended to pursue a larger number of occupations

over a greater part of the year, such earnings reflect an inverse

Following a visual examination of the variation with income of a
large number of variables, it was seen that interesting trends frequently
occurred at both extremes of the income distribution. To capture these
patterns while avoiding repetitiveness in middle income presentation,
the data has been aggregated into the following strata:

--

relationship with household income status. The highest mean female income,

N103 per household, was calculated among households in the poorest decile,

and the lowest, 452, was calculated among the richest decile of households.

In percentage terms the inverse relationship between male and female earn-

ings is particularly strong with the proportion of female to male earnings

falling from 58 percent in the first decile to only 8 percent in the tenth

decile. While these data are highly speculative, they seem to suggest

that female occupations play an important supplemental function among the

poorest households, with lower income families relatively and absolutely

more dependent on female earnings than higher income households.

Because these estimates were not believed to be sufficiently accur-

ate for subsequent analysis, female earnings have not been included as a

component of household incomes in the present study. But it is important

to note that if included, the aggregate level of inequality would be even

lower than that reflected in Tables 3.6 and 3.7. The effect of including

estimated female earnings on the relative ordering of households was exa-

mined to determine the stability of the decile and quintile stratification

set out above. It was found that inclusion would have resulted in only a

Decile 1
Decile 2
Quintile 2 (Decile 3 plus Decile 4)
Quintile 3 (Decile 5 plus Decile 6)
Quintile 4 (Decile 7 plus Decile 8)
Decile 9
Decile 10
This approach best represents the most important patterns in the middle
income groups while permitting a more focused examination of the charac-
teristics of the extreme poor and extreme rich. The cost of retaining
a decile disaggregation in the extreme income ranges is, of course,
reduced sample size and thus reduced statistical precision in the resulting
decile means. The reader should keep in mind the varying sample sizes
for decile and quintile strata when interpreting the following results.

marginal restratification of households, with the effects concentrated

in movements between the lower three deciles.

3.7. Sources of Earnings by income Stratum

In order to determine how the three major household sectors contri-

buted to overall income inequality, the contribution of each sector to

aggregate incomes (both cash and in-kind) is shown by income stratum in

Table 3.9. The percent of off-farm income remains nearly constant in

Table 3.9

PERCENT OF HOUSEHOLD INCOME EARNED IN OFF-FARM
EMPLOYMENT BY VILLAGE AND INCOME STRATUM

Type of
Village Employment

Rogo Total Off-Farm
Hired Farm Labor
Non-Agricultural

Zoza Total Off-Farm
Hired Farm Labor
Non-Agricultural

Barbeji Total Off-Farm
Hired Farm Labor
Non-Agricultural

All Total Off-Farm
Hired Farm Labor
Non-Agricultural

a. Less than .5 percent.

Decile

1 2

36 19
5 5
31 14

9 10
9 10
0 0

19 41
6 9
13 32

20 25
8 4
12 21

Quintile

2 3 4

37 26 25
2 3 a.
35 23 25

28 15 31
6 2 3
22 13 29

22 19 34
8 4 2
14 15 32

23 23 27
4 5 1
19 18 26

the lower four quintiles of the combined three village stratification

varying between only 22 and 27 percent of total income, but rises to

nearly 40 percent in the highest quintile. The proportion contributed

by work on the fields of other households on the other hand decreases

Decile

9 10

64 63
a. 0
64 63

39 32
3 1
36 31

24 16
8 1
16 15

40 37
4 1
36 36

as expected, from 8 percent of all income in the poorest decile to only

1 percent in the richest decile.

It is apparent that an important factor contributing to inequality

of the relative high income type was non-agricultural incomes generated

off the farm. In contrast earnings from hired farm labor tended to

reduce income inequality by partially compensating for low farm earnings

among poorer households. The regular pattern displayed for the entire

sample, however, masks intervillage differences in income profiles. No

consistent association between income and the proportion of off-farm

income was found in Barbeji. In contrast to the aggregate pattern,

off-farm incomes were relatively less important among richer households

in that remote village, falling to less than 16 percent in the tenth

decile. In Zoza the proportion of off-farm incomes and income per con-

sumer were directly related throughout most of the income range. And

in the largest village, Rogo, a strong positive association was evident

with non-agricultural earnings contributing more than 60 percent of total

income in both the ninth and tenth deciles. Earnings from hired farm

labor were of importance in Barbeji throughout its distribution, but of

declining importance in both Zoza and Rogo among the higher income

strata. In Zoza in particular, hired farm labor generated the only off-

farm income realized by households in the lowest two deciles.

Relating these income profiles to the village characteristics pre-

sented earlier, several observations can be made. Of the three villages,

the greatest concentration of income was evident in Rogo. This is the

largest village, characterized by the most advantageous market location,

the highest population density, and the highest proportion of income

derived from off-farm occupations. Inequality in Rogo was marked by

a few extremely high incomes, incomes which were generated primarily

in non-agricultural occupations. The lowest concentration of incomes

on the other hand was observed in Zoza. In contrast to Rogo, Zoza was

the smallest of the study villages with low population density, and

with a substantially lower proportion of income derived from off-farm

employment.

One must be cautious in drawing inferences from only three observa-

tions about the impact of village level factors on the equity of intra-

village distributions. Nevertheless the data suggest that village level

inequality is associated with increased pressure on the land, with the

attendent emergence of even small urban centers, and with an increasing

proportion of income generated off-farm. These results are consistent

with the macro structural change model set out in Section 1.

At the village level, these results may occur for the following

reasons. Given an egalitarian land tenure system and diminishing returns

to labor, as land becomes a scarce factor through population growth,

farm households would be expected to allocate an increasing proportion

of their labor to off-farm employment. However, because of low available

capital, poorer farmers are restricted to labor intensive enterprises

characterized by low returns to labor. If the demand for hired labor

fails to provide a level of employment sufficient to fully occupy the

excess labor, off-farm earnings may not compensate for the low farm

incomes caused by the relative land shortage. In contrast, higher income

households are in a better position to exploit the market advantages of

a more concentrated population by investing revenue earned through sur-

plus farm production in more capital intensive off-farm enterprises.

If in the latter case off-farm incomes more than compensate for their

reduced farm production (due to land scarcity) inequality would increase

in the high income range.

This explanation relies upon a changing composition of off-farm

employment across strata such that both capital intensity and returns

to labor are higher in those activities pursued by rich households. Both

factors are examined in Section 6.

3.8. Gift Transfers By Income Stratum

The exchange of gifts in the form of money, food, cloth, or other

in-kind items is ubiquitous in Hausaland. Contributions of food and

cash (biki) are commonly made in connection with marriage, naming-cere-

monies, and funerals to assist those households incurring large cere-

monial expenditures [Hill, p. 211]. In addition, Islamic custom requires

the giving of grain during prescribed periods to religious leaders, but

also to the poor and disabled (zakka) [Smith, 1962]. Indeed, the trans-

fer of gifts serves to some degree as an informal welfare or insurance

system.

Cash and in-kind gifts data are presented by village and income

stratum in Table 3.10 to determine whether the magnitude and direction

of gift flows importantly altered the distribution of earned income.

The results show that only the extreme deciles and elite households

reflected a clear net flow of gifts down the income spectrum. Moreover,

the net amounts involved were relatively minor compared with the

Table 3.10 NET CASH AND IN-KIND GIFTS PER HOUSEHOLD REPORTED
BY VILLAGE AND STRATUM (IN NAIRA)

Decile Quintile Decile
Variable Village 1 2 2 3 4 9 10 Elites

Value of Net Barbeji -10.23 -6.26 -8.81 -13.63 -17.40 -15.38 -104.93
Cash Gifts Zoza -29.78 -17.01 -5.04 +4.34 -6.39 -7.12 -17.16
Received Rogo -4.00 +13.06 -9.26 -14.12 -20.80 -5.82 -4.74 -6.05
(in Naira) All +.54 -17.98 -12.04 -4.01 -14.32 -9.88 -39.72

Value of Net Barbeji +9.49 -1.69 -6.61 +.93 -7.14 -5.39 -14.54
In-Kind Gifts Zoza -3.59 +1.16 -.79 +1.15 -3.64 -3.61 -10.19
Received Rogo +.69 +9.23 -9.69 -1.30 +2.34 -1.82 -6.65 -63.99
(in Naira) All +3.66 -1.28 -4.70 +.50 -2.21 -4.95 -9.70

Total Barbeji -.74 -7.95 -15.42 -12.70 -24.54 -20.77 -119.47
(in Naira) Zoza -33.37 -15.85 -5.49 +5.49 -10.03 -10.73 -27.35
Rogo 3.31 +22.29 -18.95 -15.42 -18.46 -7.64 -11.39 -70.04
All +4.20 -19.26 -16.74 -3.51 -16.53 -14.83 -49.42

Total Gifts as Barbeji -0.3 -6.6 -5.7 -3.6 -5.6 -5.6 -15.3
a Percent of Zoza -17.7 -6.6 -2.5 +1.8 -2.8 -2.3 -5.1
Generated Income Rogo -3.1 +6.4 -7.5 -1.1 -3.5 -2.2 -1.9 -2.6
All +2.4 -8.2 -6.0 -1.2 -3.9 -3.8 -8.0

differentials in generated earnings. Although it appears that respon-

dents either over-reported gifts given and/or under-reported gifts

received, if it can be assumed that all strata tended to overestimate

net gift outflows in roughly the same magnitude it is clear that the

inclusion of gift transfers would not have significantly decreased the

degree of income inequality.

3.9. Monetization of Households by Income Stratum

Monetization, or the degree of integration into the cash exchange

market, is sometimes used as a measure of the modernization or develop-

ment of a peasant economy. While it may be empirically valid to use the

proportion of cash income as a proxy to compare societies with respect

to the progress made toward Western-style development, it is not clear

that this criterion is equally valid for interhousehold comparisons

within a peasant society at a particular point in time. The motivation

to enter the market economy may differ importantly among income classes.

For example, a high ratio of cash to in-kind income may reflect produc-

tion in excess of household consumption requirements, and thus relative

economic success. Conversely a high ratio may reflect short-term

liquidity problems forcing a high level of crop sales which must be

replenished later through the purchase of food. Differences in mone-

tization are also a reflection of the relative emphasis given food and

cash crops in the farming systems of poor and rich farmers. This

balance is determined by a number of crop characteristics including

relative factor intensity, land type, and differences among crops with

respect to purchased input requirements, as well as price and yield

variance the net effect of which may not necessarily result in a close

association between income and emphasis on cash crops.

The percent of net income represented by cash earnings for each

village and income stratum are shown in Table 3.11. Within each village

Table 3.11 CASH INCOME AS A PROPORTION OF TOTAL HOUSEHOLD
INCOME AND THE SOURCES OF CASH EARNINGS BY SECTOR

Decile Quintile Decile
Village or
Variable Sector 1 2 2 3 4 9 10

Percent of Total Rogo 58 63 66 42 74 75 79
Income Earned Zoza 32 8 37 49 46 47 43
in Cash Barbeji 50 62 29 45 55 49 50
All 60 50 35 50 58 57 55

Percent of Cash Farm 63 48 37 54 52 30 35
Income Earned Hired Farm
by Sector (for Labor 14 7 11 8 1 3 2
three village total) Non-Agric. 23 45 52 38 47 67 63

and for the three-village stratification a U-shaped function is apparent;

that is, relatively high cash orientation is seen in the lower income

strata, falling within the middle strata, and then rising again in the

upper strata. Also shown is the proportion of cash generated within

each household sector for the three-village combined stratification.

Farm cash earnings (crop sales less farm cash expenses) constitute the

highest proportion of cash income in the poorest decile, 63 percent,

but decline with rising incomes to only 35 percent in the tenth decile.

This is mirrored in the cash contribution of the non-agricultural sector,

which increases from 23 percent in the first decile to 63 percent in the

tenth decile. Examining similar data disaggregated by village, the

same reversal pattern was found within both Rogo and Zoza.

Two factors account for the high percentage of cash income among

the poorest 20 percent of households. First, low income farmers in

each village allocated a greater than average proportion of their

resources to the production of the cash crop groundnut. Reasons under-

lying this pattern are discussed later in Section 6. Second, poorer

households also sold an important proportion of their subsistence grains,

with the bulk of these sales occurring somewhat sooner after harvest

than among higher income households. Cash expenditure patterns during

the immediate postharvest period indicate that an important part of the

early sales were incurred to pay taxes, repay debts, and to cover Islamic

holiday expenses.

The occurence of distress sales had important implications for the

welfare of the poorest households, as well as implications for overall

inequality. To meet their consumption objectives, poorer households

matched their early grain sales with even larger purchases of food grains

proceeding the next harvest (Table 3.12). The timing of sales and pur-

chases with respect to seasonal price movements resulted in a reduc-

tion in the real incomes of poorer households and an increased cost of

calories. Moreover, the bulk of the preharvest grains supplied to the

market were supplied by farmers in the ninth and tenth deciles who cap-

tured the benefits of higher grain prices. This not only increased

For a more detailed discussion of marketing and expenditure rela-
tionship see Chapters VII and XI in Matlon [1977].

Table 3.12 TOTAL ANNUAL SUBSISTENCE GRAINS PURCHASES AND
SALES PER HOUSEHOLD BY INCOME STRATUMa

Decile Quintile Decile
Variable Units 1 2 2 3 4 9 10

Observed during data
collection period

Subsistence grains sold as a
percent of production (by weight) % 11.3 13.4 4.4 13.4 9.0 9.8 8.5

Kilograms of.subsistence grains
sold per household kg. 103.0 151.6 60.9 197.4 152.0 149.2 185.1

Kilograms of subsistence grains
purchased per household" kg. 196.5 198.0 51.9 85.1 67.2 111.2 96.8

Ratio of sales to purchases .52 .76 1.17 2.32 2.26 1.35 1.91

Estimated potential minimum

Subsistence grains sold as a
percent of production (by weight) % 11.3 13.4 4.4 22.5 27.2 42.0 48.0
Kilograms of subsistence grains
sold per household kg. 103.0 151.6 60.9 331.9 460.1 639.8 1040.8

Kilograms of subsistence grains
purchased per household kg. 196.5 198.8 51.9 85.1 67.2 111.2 96.8

Ratio of sales to purchases .52 .76 1.17 3.90 6.85 5.75 10.75

a. Subsistence grains include early and late millet and tall and short sorghum.

b. Sales as of early May, 1975.

c. Potential sales were estimated by assuming the sale of all grains held in stock as of May, 1975,
which were in excess of the amount required to meet the average caloric intake per consumer of
the sampled households. See Matlon [1977], Appendix G.

d. Based on actual purchases observed during the 12 month survey period.

overall income inequality, but placed poorer households in a position

of dependence on high income producers with regard to meeting their sub-

sistence requirements.

3.10. Available Calories by Income Stratum

A meaningful appreciation of any given distribution of income re-

quires combining information about the relative inequality among reci-

pients with knowledge of the absolute levels of income attained by reci-

pients in each stratum. An approach which has received increasing atten-

tion to systematize problem identification as well as to guide policy

design has been the application of basic needs standards whereby levels

of economic sufficiency are defined for a range of goods (food, shelter,

clothing, health, education, etc.) [Streeten and Burki, 1977]. The

incidence of shortfalls below each standard can then be measured both

in terms of the number of persons experiencing the shortfall, and in

terms of its absolute magnitude.

While undernutrition is only one reflection of poverty, it is pro-

bably the most pervasive as well as being causally related to other mani-

festations such as morbidity, mortality, and low labor productivity.

Because estimates of minimum calorie requirements exist, undernutrition

is also one of the few basic needs for which reasonably objective stand-

ards can be established.

While the net impact of these transactions was to increase overall
inequality, the magnitude of the impact was found to be relatively small.
An analysis of the seasonal marketing of subsistence grains and of the
cash crops groundnut and pepper, led to the conclusion that differences
in timing resulted in loss of sales revenue amounting to only 2.7 per-
cent of the incomes for households in the poorest decile, and an increase
of only 1.3 percent in the incomes of households in the richest decile
[Matlon, 1977, pp. 250-265].

The data on food production, purchases, sales, and gift transfers

were examined to determine whether caloric needs were being met and

their relation to income.1 Although on average the sample households con-

sumed nearly 11 percent more calories than the required level suggested

by the FAO, there was considerable uneveness across income strata. Among

households in the first and second deciles it was found that domestic food

crop production was approximately 70 percent and 50 percent below require-

ments, respectively. Furthermore, after netting out sales and adding food

purchases and gift transfers, the first and second deciles still experi-

enced calorie deficits of approximately 25 percent and 15 percent. That is,

to meet minimum requirements, purchases and gift transfers well in excess

of observed levels during the previous year's pre-harvest period would

have been required.

It can be concluded that while the income distribution does not

reflect a high degree of relative inequality, because of the generally

low level of income overall the distribution does reflect a serious degree

of absolute impoverishment among the poorest households.

Caloric intake was calculated using the residual method by sub-
tracting annual sales, gifts given, and storage losses from the total
food crops harvested plus annual purchases and gifts of food received.
Caloric requirements were calculated as 2954 per man equivalent. This
figure was derived from a consumption survey conducted among similar
rural households in the Zaria area [Simmons, 1976]. The analysis of
caloric sufficiency is described in detail in Matlon [1977, pp. 277-283].

4. HOUSEHOLD DEMOGRAPHIC CHARACTERISTICS

An accurate identification of poverty group characteristics is of

direct value in the design and delivery of programs assisting low income

households. Research conducted in the United States as well as in other

developed countries has shown that poverty households can be distin-

guished by a fairly common set of structural characteristics [US Govern-

ment Printing Office, 1969]. Attributes found to be associated with

poverty status include: (1) a high dependency ratio; (2) a greater num-

ber of households headed by the elderly, disabled, or by females; (3)

low educational achievement; and (4) membership in ethnic minority groups.

Very few rural income surveys conducted in developing countries have

collected sufficiently detailed household information to construct pro-

files of family characteristics differentiated by income. This section

examines the extent to which a set of socio-economic characteristics of

the sampled households vary with income status and tests a set of hypo-

theses explaining their interaction with income.

4.1. Family Structure and The Life Cycle

The size, composition, and stage of development of the household

are hypothesized to be associated with income through a number of rela-

tionships. On the consumption side, the number of persons to be pro-

vided for importantly determines the level of household income considered

to be adequate. Thus, family size would be expected to directly in-

fluence production objectives. On the production side,family size would

be expected to vary closely with the available work force. The asso-

ciation between household size and income per capital or per consumer,

however, is less clear. Importantly affecting this relationship is

whether or not household composition varies systematically with house-

hold size; in particular, whether the proportion of working age persons

is associated with changes in the number of residents. A second deter-

mining factor is whether labor productivity is associated with the size

of the family work force. This in turn depends upon whether or not

there exist economies of scale in production, whether complementary fac-

tors (especially land) increase in proportion with household size, and

whether worker efficiency and managerial competence are correlated with

family size through variation in the age and experience of the work force.

Several authors have suggested that these relationships are sys-

tematically interrelated with the demographic cycle of family formation,

growth, and decline. Hedges [1963] has distinguished three stages in

the growth of farm firms in developed economies: learning, maturity and

optimum performance, and postmaturity during which the manager's effec-

tiveness declines. Chayanov [1966] has presented a framework for peasant

farming systems within which variation in income per consumer is explained

as a function of household size and composition, both of which are in

turn associated with a family's development. Formulated for application

to a land surplus environment, Chayanov's life-cycle model is based upon

changes in the ratio of consumers-to-workers which accompany household

growth. Assuming normal fertility behavior the consumer-to-worker ratio

has an inverted U-shape when plotted against the number of years since

the family's inception. Controlling for variation in work intensity,

production per consumer declines during that stage of household develop-

ment when the consumer-to-worker ratio is high.

The life-cycle hypothesis of income variation has also been explored

by Kuznets [1976] in an examination of aggregate U.S. data. Finding strong

evidence of a close non-linear correlation between age and personal

income, Kuznets concluded that valid normative judgments regarding the

personal distribution of income must take into account the earnings

life-cycle.

To determine the presence of a life-cycle earnings pattern one would

ideally trace the characteristics and incomes of actual cohorts through

time series data. Unfortunately such data are not available. As a

second best alternative, households have been jointly stratified by size

of household and by age of household head. The stages of family develop-

ment can be roughly inferred by tracing patterns across these two dimen-

sions. To control for differences in family organization, nuclear and

extended (gandu) household units have been separated. Due to limited

sample size, the number of observations per cell is in most cases too

small to draw valid statistical inferences regarding the strength of

these relationships. Rather, the purpose of this discussion is to deter-

mine whether general patterns indicate that life-cycle factors contribute

to the observed distribution of income.

The variation in consumers per worker was examined using this frame-

work (see Appendix B, Table B. 1).1 Among nuclear households, it was

1The number of "workers" in each household is equal to the number of
persons who engaged in weeding (the primary task during the agricultural
labor bottleneck period) weighted by a productivity coefficient. The
following worker productivity weights were employed:

Worker Equivalent Weights by Age and Sex
Sex 5-9 years 10-15 years 16+ years

Male .25 .8 1.0
Female .25 .5 .6

found that the ratio of consumers-to-workers was directly related to both

the size of family and the age of head, reflecting both additional wives

and children. Furthermore, compared with nuclear families, consumer-to-

worker ratios were generally more favorable among extended (gandu) units.

Among smaller gandu units headed by men in their twenties and among

units headed by men in their forties, households were composed of a

greater proportion of workers. The first group was composed predominantly

of small households united in fraternal gandu, while the latter included

primarily paternal gandu in which the sons of the household head had

joined the adult work force. This compositional advantage was lost,

however, for gandu heads in their fifties as their sons established fam-

ilies thereby increasing the dependency burden. Among the most elderly

gandu heads the consumer-to-worker ratio increased even more rapidly

as sons broke away from the extended unit and the gandu unit began to

fragment.

The variation in farmed hectares per consumer was also examined

within the life-cycle framework revealing a well defined pattern of

accumulation then loss of land for nuclear households (Appendix B,

Table B.2). Cultivated area per consumer was found to increase until

the head was in his thirties, then decline, most rapidly after age fifty.

A similar but less well defined trend is evident for extended families.

Furthermore, the reduction in holdings occurred at a somewhat later

stage in the development of the extended units. From the earlier dis-

cussion it is likely that this was the result of a more favorable con-

sumer-to-worker ratio in larger extended families reflecting the

availability of sons in paternal gandu units.

The effect of these factors on mean incomes per consumer is displayed

in Table 4.1. In view of both the consumer per worker and land per con-

sumer patterns, it is not surprising that among nuclear units the highest

incomes were realized by small families in relatively early stages of

development. As nuclear families expand, a fairly consistent inverse

relation with income is evident with a particularly rapid decline in

incomes for large nuclear families with heads 50 years and older. An im-

portant exception is among families with very young household heads, aged

24 or less, for whom incomes were also relatively low. This latter group

may have been characterized by inexperience and thus below average man-

agement skills.

The decline in incomes for extended families occurred later with

respect to the age of head. Incomes were fairly uniform through 50

years of age, though they decline noticeably for heads aged 60 or greater.

The sharp reduction in gandu size associated with the low incomes of

this age group again points toward the disintegration of the extended

unit.

In Table 4.2 the distribution of the poorest 30 percent of households

is shown as a proportion of the total number of observations per cell.

Three sets of households are disproportionately represented in this

poverty group: (1) households headed by persons aged 60 years or older,

(2) households headed by persons less than 25 years of age, and (3)

nuclear households consisting of seven or more residents (the average

household size). As a group these households constitute only 18 percent

of the sample but include 47 percent of those households included in the

poorest three deciles.

Table 4.1 MEAN INCOME PER CONSUMER BY SIZE OF HOUSEHOLD AND
AGE OF HEAD FOR NUCLEAR AND EXTENDED FAMILIESa
(IN NAIRA)

Number of Resi- Age of Household Head
Number of Resi-
dents Per Nuclear Households Extended Households
Household
-24 25-29 30-39 40-49 50-59 60+ Total -24 25-29 30-39 40-49 50-59 60+ Total

1-2 115.00 141.00 79.50 95.67

3-4 38.50 83.00 119.60 99.33 94.50 54.00 92.50 98.00 95.67 96.60

5-6 46.00 108.00 75.17 56.60 102.50 51.00 73.65 90.67 86.38 25.00 82.33

7-8 165.00 92.50 32.00 82.80 34.00 231.00 66.50 101.00 100.33

9-10 52.00 52.00 59.50 86.00 37.00 71.43

11-12 24.00 40.00 32.00 99.00 52.00 86.50

13-14 50.00 58.00 54.00

15-16 61.00 71.50

17-18 38.00 38.00

19-20 87.00 87.00

21+ 27.00 27.00

Total 59.50 106.75 84.88 81.84 76.33 52.50 82.85 34.00 231.00 72.42 88.81 89.58 41.50 82.52

a. Calculated as simple means.

Table 4.2 FREQUENCY DISTRIBUTION OF POOREST 30 PERCENT OF HOUSEHOLDS BY SIZE OF HOUSEHOLD
AND AGE OF HEAD FOR NUCLEAR AND EXTENDED FAMILIES

Number of Resi- Age of Household Head
dents Per
Household Nuclear Households Extended Households
-24 25-29 30-39 40-49 50-59 60+ Total -24 25-29 30-39 40-49 50-59 60+ Total

1-2 0/1 0/1 0/4 0/6
3-4 2/2 1/4 0/5 1/6 0/4 1/1 5/22 0/2 0/3 0/5

5-6 1/1 0/2 1/7 3/5 0/2 1/1 6/18 1/3 3/8 1/1 5/12

7-8 0/1 0/2 2/2 2/5 1/1 0/1 1/2 0/5 2/9
9-10 1/1 1/1 1/2 2/4 1/1 4/7
11-12 1/1 1/1 2/2 0/2 0/2 1/2 1/6

13-14 1/1 0/1 1/2
15-16 0/1 0/1 0/2
17-18 1/1 1/1
19-20 0/1 0/1
21+ 0/1
Total 3/4 1/8 3/18 4/13 3/9 2/2 16/54 1/1 0/1 5/12 5/16 1/1 3/4 15/46

a. Within each cell, the numerator represents the number of households
denominator represents the total number of households in the cell.

included in the poorest 30 percent of family units, and the

Although a larger sample would have facilitated a more rigorous

test of the life-cycle hypothesis, it can be concluded from the avail-

able evidence that systematic changes in demographic factors and access

to land, both of which are associated with household growth and develop-

ment, contribute to a life-cycle income pattern. Moreover, it is clear

that the form of household structure importantly affects both the

sequence and rate in which households experience these general income

stages. Households which maintain or adopt a gandu structure as the

household develops enjoy consistently higher incomes than did advanced

nuclear units. However, the number of exceptions to these patterns

suggest that life-cycle factors account for only a limited proportion

of incomes variation.

To summarize the association between demographic factors and income

per consumer, average household characteristics have been calculated

for each income stratum overall and by village in Table 4.3. Regardless

of the measure employed, household size was inversely related to income

per consumer. It is important to note the exception to this pattern

posed by the village elites among whom household size by each standard

was nearly three times the random sample average.

No association is apparent between the number of consumers per

worker and household income status. The hypothesized inverse relation-

ship was not supported because workers faced with a high dependency

ratio tend to increase work levels through farming larger areas per

worker, as well as through increased off-farm employment in an effort

to supplement farm earnings (see Appendix B, Table B.3).

63

Table 4.3 HOUSEHOLD DEMOGRAPHIC CHARACTERISTICS BY
VILLAGE'AND INCOME STRATUM

Decile Quintile Decile
Village
Variable Village Mean 1 2 2 3 4 9 10 Elites

Size

Residents
(number)

Consumer map-
equivalents
(number)

Workers
(number)

Composition

Consumer to
worker ratio

Number of
wives (three
village total)

Age of Household
Head

Frequency in
extreme age
groups (three
village total)

Barbeji
Zoza
Rogo
All

Barbeji
Zoza
Rogo
All

Barbeji
Zoza
Rogo
All

Barbeji
Zoza
Rogo
All

1.40

Barbeji
Zoza
Rogo
All

-24
60+
Total
extreme

5.0
7.0
16.0
8.3

6.3 4.0
5.3 5.3
8.1 4.3
6.2 4.5

7.3 3.0 4.9 4.2 4.3 2.9
5.1 5.0 3.7 3.9 3.5 3.4
4.5 11.0 5.6 5.0 5.7 2.8
6.4 5.9 5.0 4.1 4.3 3.1

1.3 1.7 2.2 1.9
2.1 1.8 1.9 1.7
5.3 2.0 2.3 2.4
2.8 2.1 1.9 1.9

1.30 1.30 1.70 1.35 1.35 1.40 1.20 2.50

40.2 40.0 45.3 40.7 41.3 37.3
36.3 32.0 36.7 40.4 35.7 37.8
42.3 48.3 45.0 41.0 41.3 43.3
39.6 39.5 43.9 39.0 39.9 39.2

43.0
31.7
34.7
29.9

- 1

- 1

a. Consumer man-equivalents have been determined by weighting each member of the household by a con-
sumption coefficient on the basis of the person's age and sex.

b. The number of "workers"
activities (the primary
tivity coefficient (see

in each household is equal to the number of persons who engaged in weeding
task during the agricultural labor bottleneck period) weighted by a produc-
text).

13.0

5.9

34.3
35.0
43.3
36.3

45.6

When aggregated into income strata, no association is evident bet-

ween the age of the household head and income per consumer. This is

because low income households were disproportionately represented by

both very young and elderly heads, and because peak incomes among nuclear

and extended households tended to occur at different stages in their

family development.

The demographic characteristics of the small set of households

selected to represent village elites are of particular interest. Un-

usually large paternal gandu households, they provide examples of what

has traditionally been considered the ideal Hausa family unit [Hill, pp.

165-167]. Each of the six elite heads had two or more wives, compared

with only 36 percent of the random sample with greater than one. More-

over, they represent a select group of particularly strong extended

units in which still active fathers are supported by a work force of

several adult sons. It is important to recognize, however, that these

elites were a clearly distinct and atypical subset of the most affluent.

4.2. The Distribution of Modern Education

Due to historical circumstances which limited the establishment of

mission schools in the predominantly Moslem north, modern formal and in-

formal education in this region of Nigeria is relatively recent and sub-

stantially below levels achieved elsewhere in the country. This was

In 1975 the Federal Government of Nigeria committed itself to pro-
viding universal primary education, a program which is expected to impor-
tantly reduce regional inequalities by the early 1980s. These data re-
flect conditions preceding the initiation of that program.

clearly evident within the study villages. Only one percent of house-

hold heads among the random sample and only six percent of school aged

children had attended primary school. Eight percent of the random

household heads had attended adult literacy class, and only 15 percent

had met with an extension agent during the previous five years. Fur-

thermore, literacy in either Hausa or Arabic was limited to only seven

percent of the random heads. While these levels are too low to derive

conclusive inferences, it should be noted that none of these measures

of modern education reflected a consistent positive correlation with

income.

The elite households present a minor exception. Although none of

the village elites had gone to primary school, three of the six village

leaders had attended adult education classes and two of the six were

literate in at least one language. Similarly, 27 percent of school aged

children in elite households were currently attending primary school.

As expected in light of village institutions, the elites also enjoyed

privileged access to the agricultural extension system with five of six
2
having had contact with the extension agent during the previous five years.

Among the three study villages, two primary schools were operating
in Rogo, one in Zoza, but none in more remote Barbeji. Adult literacy
classes had also been offered in both Rogo and Zoza in recent years.
Similarly, the Rogo extension agent had worked in these two more acces-
sible village areas.

2The majority of contacts with the agent were for the purpose of
obtaining fertilizer and groundnut seed at subsidized prices. The vil-
lage elites played a central role in the allocation of inputs received
from government sources. It is important to note that in several in-
stances they were observed to use this role to divert disproportionate
shares of government supplied inputs to their personal use. For a fur-
ther discussion of these activities and the resulting perceptions of
villagers, see Matlon [1977,'pp. 389-400].

66

It can be concluded that with the exception of a numerically small

group of village leaders, the data do not suggest that current patterns

of education within the study villages contribute either to a widening

of income differences or to a transmission of income differentials

across generations.

5. FACTOR USE AND PRODUCTIVITY

5.1. Land Use

Many rural income studies conducted in developing countries have

found that access to land is the single most important factor explaining

income differences. Indeed, in the absence of income data, land use is

commonly employed as a proxy variable to stratify households into income

or welfare classes [Mellor, 1975; King, 1976]. But while the land proxy

has considerable intuitive appeal in a land shortage environment, or

where land tenure institutions result in restricted access to land, its

relevance to a more land abundant environment, such as northern Nigeria,

is questionable. Indirect evidence of an association between the amount

of cultivated land and income was seen earlier in the discussion of life-

cycle income patterns. This relationship will now be examined more di-

rectly.

Land use patterns across income strata are shown in Table 5.1 It

is clear that while higher income households farmed somewhat larger land

areas, with the exception of the elite households the relationship was

not strong. The simple correlation coefficient between income per con-

sumer and cultivated area per household for the random sample is only .2045.

As would be expected, a higher correlation was evident between income

per consumer and cultivated area per consumer, reflected in a coefficient

of .5428.

However, the size of this coefficient as well as the magnitudes of

the hectare per consumer figures in Table 5.1 indicate that land use

alone accounts for less than half of the variation in incomes. For

example, in both Zoza and Barbeji the most land short income class was

Table 5,1 CULTIVATED LAND HOLDINGS BY VILLAGE AND INCOME STRATUM

Decile Quintile Decile
Village
Variable Village mean 1 2 2 3 4 9 10 Elites

Cultivated hectares Barbeji 3.0 3.1 1.1 2.7 2.1 4.0 3.6 4.2
per household Zoza 2.7 3.5 2.5 2.1 2.7 2.3 3.6 2.8
Rogo 1.9 1.0 2.7 1.5 1.9 2.3 1.5 1.6
All 2.5 2.2 2.4 2.2 2.4 2.9 2.7 3.2 11.4
Cultivated hectares per Barbeji 0.45 0.29 0.22 0.36 0.38 0.63 0.90 0.58
capital Zoza 0.47 0.51 0.36 0.42 0.46 0.43 0.68 0.53
Rogo 0.24 0.17 0.17 0.19 0.28 0.28 0.35 0.28
All 0.37 0.24 0.29 0.30 0.41 0.47 0.60 0.51 .58
Cultivated hectares per Barbeji 0.68 0.42 0.37 0.55 0.50 0.93 1.24 1.00
consumer Zoza 0.69 0.69 0.50 0.57 0.69 0.66 1.06 0.88
Rogo 0.35 0.22 0.25 0.27 0.38 0.40 0.54 0.42
All 0.54 0.34 0.41 0.44 0.59 0.67 0.87 0.86 .87

not the poorest decile. Indeed, in Zoza the land area farmed per con-

sumer by the poorest decile was greater than or equal to all other

strata with the exception of the ninth and tenth deciles.

This conclusion is amplified by comparing land use and incomes be-

tween the extreme deciles. The ratios between land per consumer levels
/
observed in the richest and poorest strata are as follows: Rogo 2.1:1;

Zoza 1.3:1; and Barbeji 2.4:1. In contrast, the corresponding

income per consumer ratios between extreme deciles are: Rogo 6.6:1;

Zoza 4.5:1; and Barbeji 5.9:1. Thus the income ratios in Zoza and

Rogo are more than triple the corresponding land ratios, and in Barbeji

more than double.

Factors other than land use clearly account for the major proportion

of income variation. At the most general level, these factors must

include either income generated in off-farm activities and/or interhouse-

hold differences in land productivity. Table 5.2 presents the mean off-

farm income per consumer and the average proportion of income generated

in off-farm employment for households stratified by hectares per consumer

and income. After controlling for differences in cultivated land it is

clear that higher income households consistently earned greater off-farm

incomes than did poor households. Higher income households also made

more efficient use of their land resources. This can be seen in Table

5.3. Greater land productivity among richer households is most evident

in the higher range of hectares per consumer, while among the most land

short strata, higher income households gave considerably greater emphasis

to their off-farm activities with a consequent decline in the value of

crops production per unit of land. It is concluded that while incomes

70
Table 5.2 RELATIONSHIP BETWEEN OFF-FARM INCOME, HOUSEHOLD
INCOME STATUS, AND HECTARES PER CONSUMERa

Hectares Income Quintile
Per
Variable Consumer 1 2 3 4 5 Total
<.29 9.37 17.22 140.71 23.10
(6) (6) (1) (13)
Off-Farm .3-.49 9.79 8.92 26.87 47.16 103.30 31.68
Income Per (10) (6) (5) (6) (4) (31)
Consumer .5-.69 10.25 7.73 12.19 28.69 33.10 17.69
(in Naira) (1) (6) (7) (4) (4) (22)
.7-.89 16.42 16.27 23.60 54.45 25.61
(2) (6) (6) (3) (17)
.9+ 1.56 2.81 26.70 34.31 23.03
(3) (2) (4) (8) (17)
Total 8.45 11.80 16.35 32.31 56.21 25.02
(20) (20) (20) (20) (20) (100)
<.29 .313 .325 .874 .362
(6) (6) (1) (13)
Off-Farm Income .3-.49 .270 .168 .360 .510 .745 .373
as a Proportion of (10) (6) (5) (6) (4) (31)
Income From All Sources .5-.69 .277 .129 .165 .278 .245 .195
(percent) (1) (6) (7) (4) (4) (22)
.7-.89 .257 .217 .224 .334 .245
(2) (6) (6) (3) (17)
.9+ .048 .037 .287 .229 .188
(3) (2) (4) (8) (17)
Total .249 .213 .217 .333 .384 .279
(20) (20) (20) (20) (20) (100)
a. The number of observations is in parenthesis.

Table 5.3

MEAN FARM INCOME PER CONSUMER BY INCOME STRATUM
AND CULTIVATED HECTARES PER CONSUMERa (IN NAIRA)

Hectares Income Quintile
per
Consumer 1 2 3 4 5 Total
<.29 20.30 35.28 20.29 27.21
(6) (6) (1) (13)
.3-.49 25.71 44.91 47.93 45.84 36.20 38.26
(10) (6) (5) (6) (4) (31)
.5-.69 26.75 51.76 69.15 73.31 102.40 69.28
(1) (6) (7) (4) (4) (22)
.7-.89 39.09 59.56 79.73 102.55 71.86
(2) (6) (6) (3) (17)
.9+ 33.43 74.20 66.80 120.32 86.97
(3) (2) (4) (8) (17)
Total 25.30 43.49 61.47 65.69 92.25 57.64
(20) (20) (20) (20) (20) (100)
a. The number of observations is in parenthesis.

do vary directly with farmed area, due to differences in off-farm earnings

and in land productivity, land use alone is only a very rough proxy for

income. It is further clear that for policy purposes, the stratification

of households by size of land holding is an inappropriate tool for the

identification of poverty households.

5.2. Land Tenure and Type

Within the study villages, as within Hausaland more generally, all

lands under cultivation are retained through use rights held by family

units and vested in the head of household. Permanent transfer of usu-

fructuary rights between households or the expansion of farming onto bush

lands must be done subject to approval of the village head. Variation

in the proportion of land held under different types of tenure could

influence incomes both through income transfers contained in rental pay-

ments and, due to differences in the security of tenure, through willing-

ness to invest in land improvements thereby resulting in variation in

land quality.

Five tenurial arrangements were observed. Fifty-eight percent of

farmed areas consisted of fields inherited (gado) by the current operator.

Purchased (saye) fields constituted 20 percent of farmed area. Rented

(aro) fields constituted 16 percent and pledged (jingina) fields1

1Jingina lands are those fields for which rights have been tempor-
arily transferred from one who has borrowed cash to the household from
which the cash loan was extended. The use rights remain with the loaner
until repayment is completed. While only a small proportion of all cash
loans involve the pledging of land, pledging is not uncommon in cases
where the amount of the cash loan is relatively high and the borrower is
a poorer farmer for whom the risk of default is high. Many such transfers
become equivalent to purchases over time.

represented only 4 percent of farmed area. An even smaller proportion

of land, 3 percent, had been initially cleared out of bush by the cur-

rent operator.

Only the percentage of land held as pledged fields showed a con-

sistent, and positive, association with income status reflecting the

presence of creditor households among the upper income strata. But even

this variation was relatively minor. The proportion of pledged fields

varied from zero in the lowest decile to only 10 percent among house-

holds in the richest decile. No consistent patterns were evident re-

lating the percentages of inherited, purchased, or rented holdings with

income.1

The data also showed that there was little association between the

distribution of high value lowland (fadama) soils and income. Only in

Barbeji, where fadama fields constituted 5 percent of total cultivated

area did the proportion of fadama soil increase with income. Among

Barbeji's richest one-third of households, 9 percent of cultivated land

was fadama, compared with only 3 and 2 percent, respectively, for the

middle and lower income groups. Fadama land was most abundant in Rogo,

representing 11 percent of cultivated area. Although among the richest

third of its small sample 11 percent of farmed area was fadama, this

was offset by the poorest third, whose much smaller land base was com-

posed of 14 percent lowland soils. Thus the data suggest that neither

access to high quality lowland soils nor tenurial arrangements were sig-

nificant factors in explaining the observed income distribution.

The data are presented in Matlon [1977, p. 111].

5.3. Ownership of Non-Land Capital

Like land, the value of farm and non-farm capital equipment is a

measure of production scale. Stocks of working capital and livestock are

also a measure of accumulated wealth and represent a source of immediate

cash in the event of a production shortfall or other household financial

emergency.

The average values of livestock and working capital disaggregated

by income class are presented in Table 5.4. Although both livestock and

production capital were in general positively associated with income,

comparing the value of these stocks with household income (Table 3.5)

it is clear that capital was considerably more equally distributed than

income among the income strata. Because all households were hand tool

cultivators, the relatively minor variation in the value of farm tools

per household and per worker reflect differences in the size of inven-

tories and age of tools rather than in the types of capital employed.

Since higher income households often supplied tools to hired farm laborers,

their inventories were somewhat larger. Once again the subset of politi-

cally elite households stand out as atypical with the value of all capi-

tal nearly 20 times greater than the random sample average, and more than

10 times greater than households in the tenth decile.

5.4. Labor Use

Two aspects of labor use are briefly considered in this section.

First, we examine how the levels of employment varied annually and

by period of the year in order to determine the extent to which the

supply of household labor may have been a factor constraining incomes

Table 5.4 AVERAGE VALUE OF LIVESTOCK AND WORKING CAPITAL
PER HOUSEHOLD BY INCOME STRATA (IN NAIRA)

Decile Quintile Decile

Assets category 1 2 2 3 4 9 10 Elites

1. Livestock-Totalb 50.40 106.60 70.75 73.00 169.10 136.80 126.90 1579.67
Cattle 30.00 94.90 36.00 20.00 1301.67
(0.40) (0.80) (0.40) (0.30) (8.30)
Donkey 9.30 12.10 11.90 3.05 7.80 6.80 5.10 17.00
(0.80) (0.70) (0.90) (0.30) (0.40) (0.80) (0.50) (.88)

Sheep and Goats 11.40 12.60 16.15 15.65 17.70 32.30 32.20 277.50
(1.50) (1.70) (2.25) )1.65) (1.75) (3.40) (2.10) (24.80)

Chicken 1.50 2.00 2.30 3.85 3.40 5.00 3.70 7.33
(4.00) (4.20) (7.40) (8.45) (8.40) (14.60) (8.90) (16)
Other poultry 1.90 4.10 0.60 0.80 0.9b 7.40 3.70 15.18
Other Livestock 6.00 1.20 0.15 27.77

2. Farm tools 6.20 9.50 8.55 5.40 10.15 8.60 13.00 40.41

3. Value of farm tools
per worker 2.21 3.39 4.07 2.84 5.34 4.78 8.13 6.85

4. Non-farm capitalc 4.30 12.30 23.95 6.60 36.00 2.60 19.60 141.13

a. Values are estimated as current sale value.

b. The average number of animals per household is included in parenthesis.

c. Included are all tools and other fixed assets (e.g., shop structures) used in off-farm occupa-
tions. Inventory stocks of non-farm trading items are not included.

within each income stratum. Second, we examine the allocation of labor

to farm and off-farm enterprises to determine whether variation in the

composition of employment is related to income. Since labor data was

not obtained from the large sample, the data to address these issues are

taken from the small sample of 35 households. Because of the smaller

sample size, only three income classes are distinguished.

Average hours of employment for adult males during the entire year

as well as during the three months of peak farm work are shown in Table

5.5. The overall employment levels were low, varying between only 1.8

Table 5.5 AVERAGE DAILY HOURS WORKED PER ADULT MALE
(16+ YEARS) ACCORDING TO HOUSEHOLD
SECTOR AND INCOME CLASS, SMALL SAMPLEa

Income Class

Period Sector Low Middle High

All Year Farm 1.17 1.77 1.48
Hired Farm Labor .18 .06 .02
Off-Farm Non-Agric. .43 .24 .25

Total Hours 1.78 2.07 1.75

May-July Farm 2.29 3.50 2.78
Hired Farm Labor .31 .04 .05
Off-Farm Non-Agric. .33 .05 .09

Total Hours 2.93 3.59 2.92

Number of persons observed within
each income category. 20 24 24

a. Travel time to and from places of employment as well as work within
the family compound are not included in these figures. In addition,
these figures represent the mean daily work levels observed for each
period, not the mean hours of work only for those days during which
work was observed.

and 2.1 hours per day annually, and between 2.9 and 3.6 hours during the

peak farming period. Within both time frames, the highest relative

work rates were recorded among middle income farmers. It is particularly

important to note that when only labor on own fields is considered, low

income farmers worked the least hours. Moreover, this was true both for

the entire year and for the peak period during which the poorest farmers

worked an average of only 2.3 hours per available man day.2

The low hours worked on the farms of the poorest households reflect

at least four interrelated factors. First, as seen earlier, poorer

households farmed somewhat smaller holdings. Second, although poor

farmers expended least hours per unit area (see Section 5.5) the marginal

value product of labor was lowest among low income producers NO.055

per hour, compared to t.096 and 4.139 for middle and high income farmers,

respectively [Matlon, 1977, pp. 216-224].3 Third, the calorie shortage

experienced by the poorest households may have importantly limited the

potential energy expenditure of low income workers. And fourth, in order

1This range is well below the annual mean of 3.3 hours estimated by
Norman [1968] in his three village Zaria study. Norman, however, did not
actually collect data on the number of hours devoted to work other than on
the family farm. Rather, that study obtained information only on the
number of days during which off-farm activities were pursued and assumed
that farmers worked as long at off-farm occupations during each day worked
as they did on family farm work. That procedure almost certainly over-
estimated the off-farm labor component of his total estimate.

2The magnitude of these on-farm employment levels are consistent with
the average daily hours worked reported by Norman [1968] for the entire
year of his survey (1.64 hours per man day) as well as being in general
agreement with his peak period estimates.

3The average hourly wage rate for hired farm labor was NO.10 for
adult males.

to generate an immediate cash inflow low income males allocated a sub-

stantial proportion of their labor time to off-farm activities. On an

annual basis, low income males spent 34 percent of their total work

time in off-farm activities, compared to only 14 percent among males in

each higher income stratum. And during the peak farming months, when
their cash and food reserves were at a minimum, low income males allo-

cated 22 percent of their work time off the farm. This compared to less

than 5 percent among adult males in higher income households. Neverthe-

less, in view of the low overall employment levels of poor adult males,

it is clear that work off-farm was at best only partially responsible

for low levels of on-farm work.

With one qualification, it can be concluded that labor time was not

a significant constraint limiting the incomes of poor farmers generated

in either farm or off-farm occupations. That qualification is the possi-

bility that the time expended in job search activities and in travel to

and from off-farm employment (time not accounted for in the survey) may

have been substantial. It is clear that if important, such activities

would have disproportionately reduced the available labor supply of low

income farmers. Unfortunately data is not available to examine that

issue directly.

5.5. Farm Productivity
In the examination of land use patterns above, it was seen that

higher income households generated substantially greater farm incomes

than poorer farmers after controlling for differences in size of holding

(Table 5.3). Variation in land productivity can be caused by several

factors including: (1) differences in factor quality, especially soils;

(2) variation in the combination of crops grown; and/or (3) variation

in production technique in particular, the intensity with which the

land is farmed. Although data on soil quality was not available on a

per-field basis, it was mentioned earlier that the distribution of high

quality lowland soils was not generally associated with the household's

income status. Moreover, a soil survey carried out in the study area

concluded that there were no important differences in the physical and

chemical properties of the upland soils tested which would result in

significant productivity differentials. The possible effect of varia-

tion in cropping emphasis is examined in the next section. At this point

it is useful to briefly examine in somewhat more detail how farming

intensity and factor costs and returns in crop production varied by

income strata. Because labor data was obtained only for the small sam-

ple, the analysis is again limited to those households. To control for

general soil type differences only upland fields are examined.

Data summarizing average costs and returns per hectare for house-

holds in the low, middle, and high income classes are shown in Table 5.6.

Three measures of productivity the value of output per hectare, gross

margins per hectare, and returns to household labor, management, and

capital all indicate a strong direct relationship between production

efficiency and income. It is also clear that higher income households

farmed their upland fields more intensively with respect to both ferti-

lizer and labor. Although fertilizer use was generally low overall, high

income farmers on average applied 27 percent more fertilizer per hectare

than low income households. They also expended 21 percent more labor,

Table 5.6 AVERAGE COSTS AND RETURNS PER HECTARE FOR UPLAND
FIELDS BY INCOME CLASS, SMALL SAMPLE (IN NAIRA)

Income Class
Budget Item Low Middle High

Value of Output 99.73 120.44 148.97

Variable Costs (total) 29.78 28.68 33.88
Seed 7.64 7.89 5.57
Fertilizer (total) 1.76 2.04 2.25
Organica b 1.57 1.89 1.99
Inorganic .19 .15 .26
Hired Labor 20.38 18.75 26.06

Gross Margins 69.95 91.76 115.09

Opportunity Cost of Landc 5.01 4.36 4.52

Labor Use (hours)d 587 694 712
Family 406 430 349
Hired 181 264 363

Returns to Household Labor,
Management, and Capital per Hour 0.16 0.20 0.32

No. of Field Observations 49 56 68

a. Organic fertilizers were valued at the mean purchase price for each
type of manure applied. The average cost was NO.08 for an equiva-
lent of 160 liters of compound sweepings or manure.

b. Chemical fertilizer was valued at the current subsidy price of
41.60 per cwt. for superphosphate and N2.00 per cwt. for ammonium
sulfate.

c. All land, regardless of tenure, was valued at the average rental
rates observed in each village.

d. Hours of labor are measured in terms of man-equivalent work hours.

primarily from hired workers. In comparison the differential in value of

production between extreme income classes was 49 percent.

These relative differences indicate that unless there existed in-

creasing returns to fertilizer and labor, variation in the amounts of

conventional inputs applied does not alone explain the substantial pro-

duction gradient. Production function analysis confirmed that both

fertilizer and labor were subject to diminishing returns within the

range of observed use levels [Matlon, 1977, Chapter VI]. Moreover, a

Chow test applied to detect structural differences among the set of pro-

duction functions fitted to each income class concluded that the null

hypothesis of structural similarity across income classes could be re-

jected at the 2 percent level. The nature of these structural diffe-

rences has not yet been identified.

One explanation is that management practices were systematically

related to income. Familiarity with the farming systems of the area

suggests that differences in cultural practices could include variation

in the timing of operations, the selection of complementary crop mix-

tures, rotation practices, and the allocation of crops among fields

showing micro-area soil variation. Although it is clear that an iden-

tification of such differences in essentially traditional production

techniques would add importantly to an understanding of income distri-

bution, these questions lie outside the scope of the present paper.

An alternative explanation is that high income farmers gave greater

emphasis to upland crops with more favorable returns characteristics

than did low income farmers. This last hypothesis is examined in the

following section.

6. ENTERPRISE SELECTION ACROSS INCOME CLASSES

This section examines the extent to which selection of cropping enter-

prises varied across income classes and how cropping emphasis affected

returns to land. The relative emphasis given a set of off-farm activities,

and implications for returns to labor in off-farm employment are also exa-

mined.

6.1. Subsistence vs. Cash Crop Emphasis

It is sometimes assumed that poorer farmers tend to be more oriented

to the production of subsistence crops. Guided by a food first objective,

it follows that land and labor would be allocated to cash crops production

only after their domestic consumption objectives are met. It is of direct

interest to know whether this pattern applies to the present sample. For

this purpose, the major crops in the area have been grouped into three

categories: (1) cash crops, (2) subsistence grains and (3) intermediate

crops. Although nearly 40 crops were grown by sample households, to

simplify the analysis and presentation, only the 12 most important are

examined. These major crops include over 95 percent of the total har-

vest value of each income stratum, and nearly 96 percent overall. Four

crops stand out in terms of percentage sales: onion, pepper, groundnut,

and sugar cane, each with over 70 percent marketed. Grown primarily for

the market and only secondarily for domestic consumption, these crops have

been grouped into the category of cash crops. The sorghums and millets,

the most important food staples in the diet of rural northern Nigerians,

have been similarly grouped to comprise the subsistence grains category.

All remaining crops, including minor crops have been categorized as inter-

mediate crops.

In Table 6.1 the percentage of total harvest value represented by

the subsistence grains and cash crops is shown for each income stratum.

Table 6.1 THE PERCENT OF TOTAL HARVEST VALUE FOR SUBSISTENCE
GRAINS AND CASH CROPS BY INCOME STRATUMa

Decile Quintile Decile
Crop
Village Category 1 2 2 3 4 9 10

Rogo Subsistence Grains 41 31 27 27 25 31 26
Cash Crops 58 61 54 58 65 68 64

Zoza Subsistence Grains 58 49 60 38 46 54 46
Cash Crops 37 42 29 39 38 37 31

Barbeji Subsistence Grains 62 56 54 54 45 50 37
Cash Crops 31 25 34 31 46 35 55

All Subsistence Grains 42 46 45 47 36 44 42
Cash Crops 52 39 45 36 55 45 47

a. The percentage of the residual crop category, intermediate crops,
is not shown.

The emphasis given the major crop groups within the cropping systems

of each income stratum are surprisingly uniform. From a 52 percent

emphasis on cash crops in the first decile, the proportion falls to 36

percent in the third quintile, rises to the highest proportion, 55 per-

cent, in the fourth quintile, then plateaus at approximately 45 per-

cent in the top two deciles.

The three-village total, however, masks underlying patterns pecu-

liar to each village. Reflecting the same village rankings with respect

to inequality and degree of monetization, Rogo farmers on average gave

the greatest emphasis to cash crops (61 percent of their total production),

followed by Barbeji (40 percent), and Zoza (37 percent). The cash crop

share of total production in both Rogo and Barbeji showed a weak posi-

tive association with income levels, while little trend was evident in

Zoza.

Of particular interest in the three-village aggregate pattern is

the relative cash crop emphasis of the poorest households. It was seen

earlier that the lowest two deciles experienced a calorie deficiency of

approximately 20 percent of estimated requirements. Nevertheless, roughly

52 percent and 39 percent of the harvest value of the poorest two deciles,

respectively, represented the production of cash crops. This is clearly

inconsistent with the hypothesis that the primary objective of the poorest

farmers is to produce a food supply sufficient to meet the domestic consump-

tion needs. Several factors explaining the importance of cash crops among

the poorest households are discussed later.

6.2. Crop Enterprise Balance by Village and Income Stratum

A more detailed breakdown of crop mix among income classes is dis-

played in Table 6.2. An index of crop emphasis has been computed by

dividing the crop percent for each stratum by the overall mean percent

for the entire sample, thereby standardizing at one. Values greater than

one represent disproportionate emphasis given to that crop with values

less than one reflecting lower than average emphasis.

The basic similarities in crop allocation among income strata are

striking. With the exception of rice, sugar cane, and root crops, each

of the 12 major crops was produced by households in each class of the

overall sample, and in roughly similar proportions. The absolute size

Table 6.2 THE

HARVEST VALUE OF 12 MAJOR CROPS EXPRESSED AS A PERCENT
OF THE TOTAL HARVEST VALUE BY INCOME STRATUMa

Tall Short
Income Early Late sor- sor- Ground- Sugar Root
strata millet millet ghum ghum Maize Rice Cowpea nut Onion Pepper cane crops Total

Decile 1 7.8 2.1 18.5 13.4 1.5 2.4 50.0 0.5 1.8 -- 98.0
S 2 7.7 2.3 18.7 17.4 1.7 3.1 3.1 31.1 4.9 3.0 2.3 95.3
Quintile 2 6.4 1.1 30.2 7.1 1.4 1.1 3.5 31.2 4.3 5.1 4.2 0.6 96.2
S 3 7.2 1.9 27.5 10.3 1.7 0.9 3.0 24.9 3.7 5.7 1.6 6.9 95.3
S 4 6.5 1.5 24.1 3.7 0.6 2.7 2.7 37.2 5.6 5.0 6.7 2.4 98.7
Decile 9 5.2 1.1 26.4 11.1 1.2 1.1 5.4 33.0 5.6 6.1 0.8 97.0
S 10 7.2 2.2 24.5 7.9 1.2 2.2 3.9 26.6 4.8 9.4 6.1 0.7 96.7

All 6.8 1.7 25.3 8.2 1.2 1.7 3.4 32.3 4.6 5.6 3.7 1.3 95.8

Relative Cropping Emphasis Indexb
Decile 1 1.15 1.25 .73 1.63 1.25 .71 1.55 .11 .32 -
S 2 1.13 1.35 .74 2.12 1.42 1.82 .91 .96 1.07 .54 1.77
Quintile 2 .94 .65 1.19 .87 1.17 .65 1.03 .97 .93 .91 1.14 .46
S 3 1.06 1.12 1.09 1.26 1.42 .53 .88 .77 .80 1.02 .43 5.31
4 .96 .88 .95 .45 .50 1.58 .79 1.15 1.22 .89 1.81 1.84
Decile 9 .76 .65 1.04 1.35 1.00 .65 1.59 1.02 1.22 1.09 .62
1 0 1.05 1.29 .97 .96 1.00 1.29 1.14 .82 1.04 1.68 1.65 .54

a. Percentages have been calculated as weighted means.

b. The relative cropping emphasis index has been calculated as the ratio of the percentage harvest value of each
class to the overall percentage harvest value for the respective crop.

crop in each income

of interstrata differences in production shares are particularly small

for the millets, maize, rice, cowpea, onion, pepper, sugar cane, and

root crops. The widest range is evident for groundnut.

Crops which comprised a greater than average share of harvest value

among the lowest income classes with generally decreasing shares as

incomes increase, include early millet, short sorghum, and maize. Crops

which show the opposite pattern, that is lower than average share in

total harvest value among households in the lowest income classes and

a generally increasing share in the upper income strata, include cowpea,

onion, pepper, and sugar cane. A U-shaped relationship with income des-

cribes the emphasis given late millet, short sorghum, and groundnut.

The importance of groundnut in the cropping system of the lowest decile-

representing 55 percent greater share than average-should be noted in

particular. Given these patterns it is necessary to determine whether

the relatively minor variations in interstrata crop mix reflect an under-

lying shift among high income households towards crops with more favor-

able returns characteristics.

6.3. Crop Mix Variation and Land Productivity

Farmers in the study area plant their crop in mixtures with sole-

cropped fields representing only a minor proportion of sown area. More-

over, the heterogeneity of intercropped fields is high. The variety

A total of 225 distinct mixtures were recorded on 205 separate
fields. Of these, only 20, or less than 9 percent, were sole-cropped.
The three most frequently observed combinations (tall sorghum, early
millet, and cowpea; tall sorghum and cowpea; and sole cropped sorghum)
occurred, respectively, on only 30, 27 and 18 plots, out of a total of
484 plots (plots were defined as contiguous pieces of land, not less
than 100 square meters in area, on which a single crop or crop mixture
was present).

of intercropped mixtures presented considerable problems to estimate

the cost and return characteristics of individual crops. Except for

actual planting and harvest activities, few labor inputs could be
assigned to a specific crop. Similarly, the amount and value of fer-
tilizers applied to intercropped plots could only be crudely disaggre-

gated by crop. Finally, lacking plant stand counts with which it would

have been possible to estimate adjusted crop areas, crop-specific costs

and harvest values could not be directly related to a crop-specific
hectare base.

These data problems prevented the use of standard farm management

techniques, such as potential gross margins analysis, to measure the

effect of enterprise choice on factor returns. Instead, a two step

analytical procedure was used. First, analysis of variance was applied

to derive estimates of average gross margins per hectare for each major

crop enterprise.1 Second, a weighted sum of gross margins was calculated
for each income class to reflect the expected returns to land given the
observed crop mix but controlling for interstrata variation in land pro-
ductivity attributable to differences in technique or resource quality.

The analysis is described in detail in Appendix C.

The results are shown in Table 6.3. These figures can be inter-
preted as representing the approximate change in the value of gross mar-
gins per hectare attributable to shifting from production of all minor

1Gross margin per hectare was defined as the value of harvest less
the imputed value of all seeds, cuttings, organic manures, chemical fer-
tilizers, and seed dressing applied, and less the value of all cash and
in-kind payments to hired labor, divided by field area. Observations
for fields on which sugar cane and cassava were grown were excluded due
to lack of full data sets. Both crops have growth periods which fell
outside of the duration of the survey.

Table 6.3 THE EFFECT OF CROP MIX ON GROSS MARGINS
PER HECTARE BY INCOME STRATUM
(IN NAIRA)

Decile Quintile Decile

1 2 2 3 4 9 10

Excluding Sugar -3.31 -11.24 -8.75 -12.54 -0.17 -6.57 -5.25
Cane and Root
Crops

Including Sugar -3.31 -10.48 -3.82 -8.30 +7.12 -6.35 +1.57
Cane and Root
Cropsa

a. See Appendix C.

crops (the reference crops category) to the mix of major crops represen-

tative of each income class. In an effort to take into account the

presence of sugar cane and cassava, weighted sums were also calculated

using assumed coefficients. It is clear that the effect of variation

in crop mix on interstrata differences in returns to land is relatively
minor. Moreover, there is no consistent trend across strata. In short,
the data suggest that choice of crop enterprise is not an important

factor in explaining the strong direct relationship between farm pro-

ductivity and household income status.

The results of the analysis also suggest a likely explanation for
the high degree of cash cropping among low income producers. It was

seen earlier that relatively greater emphasis on groundnut was seen

among households in the lower and upper income extremes with a decline

in emphasis evident among middle income households. The high level of

groundnut production among the poorest households is initially puzzling

in view of their calorie deficit position. A possible rationale fol-

lows.

Two general strategies can be pursued in supplying domestic calorie
needs through crop production. Calories can either be directly produced
for household consumption through the cultivation of food crops, or

calories can be provided through the production and sale of cash crops

with subsequent purchases of food in the market. Only three crops

ranked higher than groundnut in terms of gross margins per hectare:

onion, pepper, and maize. It is significant that each of these crops

The analysis of variance procedure resulted in the following
ordering of crops in terms of gross margins per hectare (in descending

was characterized by high requirements of purchased inputs, especially

seed and fertilizer. Moreover, these expenditures occur during a period

of cash shortage which is most acute among low income households. In

contrast to the cash crops, the staple food grains ranked lowest in

returns per hectare.

Given a limited land base, relative to their consumption objectives,

such that meeting household food requirements was unattainable regard-

less of cropping emphasis, it is likely that the lowest income house-

holds allocated greater land and labor to the production of the most pro-

fitable crop compatible with their low capital position, groundnut.

Revenues received from the sale of groundnut thus allowed a higher level

of consumption of food through purchases of grains than if the entire

land base had been allocated to less profitable food crops alone. Ground-

nut was made even more attractive to low income producers since it was

the only crop for which there was an assured demand and an established

price determined by marketing board purchases, thereby reducing the un-

certainty of price variation.

Reasons for the declining share of groundnut as one moves above the

poorest decile are less clear, but probably reflect a change in produc-

tion objectives. While there is no direct social prohibition among the

Hausa which limits a household's purchases of grain in the market-indeed

grain purchases were observed among all strata -dependence on the market

to meet household requirements is clearly associated with a social stigma.

order): onion, pepper, maize, groundnut, cowpea, early millet, late
millet, rice, tall sorghum, short sorghum.

The largest production shares of the major staple, tall sorghum, occurred

in the second and third quintile, 30 percent and 28 percent of each

stratum's total production, respectively. Given a more ample land base,

middle income households were consequently able to meet a self-sufficiency

objective, thereby reducing their dependence on the market, but only by

decreasing their groundnut plantings. Thus self-sufficiency was attained

at the cost of shifting to the less profitable food crop mix. That is,

with a sacrifice in aggregate income.

6.4. Choice of Enterprise in Off-Farm Employment

Variation in the types of non-agricultural activities pursued by

household members across income classes is shown in Table 6.4. Forty-

eight off-farm occupations have been grouped according to the distribu-

tion of each occupation's market share among income classes.1

It is evident in Table 6.4 that characteristics of non-agricultural

occupations shift systematically with household income status. All occu-

pations classified as "only low income" are service occupations employing

little or no working capital, while the number of occupations requiring

substantial levels of working capital increases directly with the income

category. An annual cash expenditure of only N2.10 per household was

If all gross sales of an occupation's products or services came from
the lowest (highest) two income quintiles, the occupation has been included
in the "Only Low (High) Income" category. If 75 percent or more, but less
than 100 percent, of total gross sales occurred in the lowest (highest)
two quintiles, the occupation was categorized as "Low (High) Income Biased."
An occupation was categorized as "Intermediate" if it did not qualify in
these other classes; that is, if less than 75 percent of total sales occur-
red in households falling within either the lower or upper two income
quintiles.

Table 6.4 THE DISTRIBUTION AND SELECTED CHARACTERISTICS OF OFF-FARM OCCUPATIONS BY HOUSEHOLD INCOME CLASS

No. of house-
hold observations
Barbeji Zoza Rogo

Total annual cash
expenditure
per household
(in Naira)

Share of total market
"sales" by income
quintilea
1 2 3 4 5

Share of net household
income derived from
this occupation
1 2 3 4 5

Only low income Begging
Shoe repair
Calabash cutting
Head transportation
Total

Low income bias

Intermediate

High income bias

Selling grass
Hauling water
Tailoring
Spinning and weaving
Lorry mate
Provisions trading
Medicine trading
Bicycle repair
Total

Hired farm labor
Selling firewood
Cap making
Decorticating groundnut
Trans. crops from field
Mat making
Rope making
Bed making
Well digging
Donkey transportation
Cloth trading
Kola nut trading
Local medicine
Washing clothes
Butcher
Total

Labor processing sugar cane
Selling roasted meat
Mud block builder
Processed food trading
Cigarettes trading
Koranic teacher
Praise singer/musician

Income bias

Occupation

1.00 -
- 1.00
1.00 -
1.00 -

.010 -

6.25
.05

2.10c

37.89
4.13

41.88
3.69
31.54c
23.24

8.47

12.80

141.64
206.14

463.97_
35.06c

228.95

252.96

2.46

.15
.82
.15
.75
.29
.75
1.00

.25

.13

.17
.09

.49

.01

.14

.017

.005
.001

.001
.001
.013
.007
.023
.018
.001

.062

- O~

.001

.012
.004
.008
.003

.032

- O~
.001

.011

.08 .37
.41 .12
.39 .15
- 1.00
.63 .13
- 1.00
- 1.00
- 1.00
.71
.43 .07
.65 -
.36 -
.91
- 1.00
.01

.006

.010

.046

.01

.040
.013
.011

.002

.029
.009
.005

.002
.001
.001
.006
.001
.001
.001

.001

.002
.002

.001

.012
.001
.008
- O~
.001

.007
.006
.002
.004

.020

.006
.010
.012
.018
.002
.028

.001
.001

.001

.045
.004
.004
.001
.001
.001
.001
- O~

.001
.026
.004

.017
.001
.005

.014
.009

.002
.008
.004

.004
.001

.001

.026
.010

.010
.001
.006

.001

.002
.058
.009
.060
.001

.017

MILLISECOND	CLASS.METHOD	MESSAGE
0	sobekcm_page_globals.constructor
0	sobekcm_page_globals.constructor	Application State validated or built
0	sobekcm_database.verify_item_lookup_object
0	sobekcm_page_globals.constructor	Navigation Object created from URI query string
0	sobekcm_database.verify_item_lookup_object
0	sobekcm_page_globals.display_item	Retrieving item or group information
0	sobekcm_page_globals.get_entire_collection_hierarchy	Retrieving hierarchy information
0	sobekcm_assistant.get_entire_collection_hierarchy
0	cached_data_manager.retrieve_item_aggregation
0	cached_data_manager.retrieve_item_aggregation	Found item aggregation on local cache
0	item_aggregation_builder.get_item_aggregation	Found 'all' item aggregation in cache
0	system.web.ui.page.page_load (ufdc.page_load)
0	sobekcm_page_globals.constructor.on_page_load
0	html_echo_mainwriter.add_style_references	Adding style references to HTML
0	html_echo_mainwriter.add_text_to_page	Reading the text from the file and echoing back to the output stream
40	html_echo_mainwriter.add_text_to_page	Finished reading and writing the file