Effects of resource constraints on the production systems and earnings potential of small farm households in the West Province of Cameroon

Material Information

Effects of resource constraints on the production systems and earnings potential of small farm households in the West Province of Cameroon
Series Title:
International working paper series
Stall, Steven J
Davis, C. G ( Carlton George ), 1936-
University of Florida -- Food and Resource Economics Dept
Place of Publication:
Gainesville Fla
Food and Resource Economics Dept., Institute of Food and Agricultural Sciences, University of Florida
Publication Date:
Physical Description:
54 p. : ; 28 cm.


Subjects / Keywords:
Farms, Small -- Cameroon -- West Province ( lcsh )
Agriculture -- Economic aspects -- Cameroon -- West Province ( lcsh )
bibliography ( marcgt )
non-fiction ( marcgt )


Includes bibliographical references (p. 53-54).
General Note:
Title from cover.
General Note:
"July 1992."
Electronic resources created as part of a prototype UF Institutional Repository and Faculty Papers project by the University of Florida.
Statement of Responsibility:
by Steven J. Stall and Carlton G. Davis.

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
The University of Florida George A. Smathers Libraries respect the intellectual property rights of others and do not claim any copyright interest in this item. This item may be protected by copyright but is made available here under a claim of fair use (17 U.S.C. §107) for non-profit research and educational purposes. Users of this work have responsibility for determining copyright status prior to reusing, publishing or reproducing this item for purposes other than what is allowed by fair use or other copyright exemptions. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder. The Smathers Libraries would like to learn more about this item and invite individuals or organizations to contact Digital Services ( with any additional information they can provide.
Resource Identifier:
001761704 ( ALEPH )
27157308 ( OCLC )
AJH4824 ( NOTIS )

Full Text
The publications in this collection do not reflect current scientific knowledge or recommendations. These texts represent the historic publishing record of the Institute for Food and Agricultural Sciences and should be used only to trace the historic work of the Institute and its staff. Current WFAS research may be found on the Electronic Data Information Source (EDIS)
site maintained by the Florida Cooperative Extension Service.
Copyright 2005, Board of Trustees, University of Florida

Steven J. Stall and Carlton G. Davis
IW92-12 July 1992
Institute of Food and Agricultural Sciences University of Florida
Gainesville, Florida 32611

Steven J. Stall and Carlton G. Davis *
(Department of Food and Resource Economis
University of Florida, Gainesville, Florida 32611, USA) INTRODUCTION
In the years following independence in 1961, Cameroon has experienced an average annual economic growth rate of 3.6 per cent, which is almost three times higher than the Sub-Saharan African nations' average of 1.0 per cent over the same period (World Bank, 1987). The country's agricultural sector grew at an average annual rate of 4.2 per cent over the 1965-80 period. This was considerably better than the average growth rate of 1.9 per cent of Sub-Saharan African countries over the same period. The growth rate of the agricultural sector has slowed substantially
over the 1980-85 period, to about 1.3 per cent per annum. However, even at the reduced rate, the country's agricultural sector growth in the latter period was in excess of the Sub-Saharan African
average of 0.9 per cent (World Bank, 1987). For the most part, Cameroon is self-sufficient in supplying its food needs. In
1984-85 only 2.6 per cent of the food consumed locally originated from net imports (the difference between exports and imports), and these imports were mainly cereals such as rice and wheat flour (Ministry of the Plan and Regional Development, 1986).
* The authors extend thanks without implications to Max
Langham, Francois Kamajou and Bernard Guietang for their assistance with study.

The West Province of Cameroon is the smallest of the country's ten provinces, but is the third largest in population.
The province has the highest population density of any of the provinces. For example, the population densities were 84.2, 95.7
and 107.7 persons per square kilometer in 1981, 1986 and 1991, respectively. In contrast, the population densities for the
country as a whole were 18.9, 22.0 and 25.8 persons per square kilometer in 1981, 1986 and 1991, respectively (Ministry of the Plan and Regional Development, 1986). The Province is also an important producer of agricultural commodities. Agriculture
accounted for over 65 per cent of the province's GDP in 1977, which was more than twice the contribution of agriculture to the
national GDP (Nji, 1982). The province is a primary producer of food crops and the main region for the production of Arabica coffee. In addition, Robusta coffee, cocoa and tobacco are produced in sizeable quantities. Given the land and other
agricultural resource constraints associated with rural population pressures in Cameroon's West Province (and other regions), the purpose of this paper is to report on the results of a study to assess the effects of land and other resource constraints on the agricultural production systems and income generation potentials of small farm households in a sub-region of the West Province.
In assessing the impact of land and other resource
constraints on Cameroon small farm systems, it is inter alia, necessary to develop: (1) an understanding of the
characteristicsof the productive process and (2) identify and

measure the key f actors impacting the process. Economic theory and empirical studies provide the contextual framework for approaching the assessment. In other words, the behavioural assumptions underlying the form and the function of the production systems of
small f farmers are rooted in the tenets of traditional economic theory and the associated body of empirical findings relating to observable economic phenomena. Based on these tenets, we
explicitly assume the following objectives as being typical of the small farmer population: (1) maximization of net returns (2) attainment of a reasonable level of self-sufficiency in household food consumption and (3) minimization of risk in production activities.
The choice of an appropriate empirical model f or assessing the behavioural characteristics of an economic system is subjective. As such, the choice process needs to be informed by a set of criteria. In the case of this study there are two
overriding criteria f or the selection of the model used. These are: (1) adaptability to the compositional theoretical framework
which is providing guidance to the behavioural assumptions and characteristics of the system and (2) "appropriateness" in meeting the general objective of the study. In this case, the objective is to evaluate the impact of land and other resource constraints on small farmer agricultural production systems and their income generation potential. A number of studies of this kind have used a production function approach in which regression analysis is used to determine the coefficients under which small farmers

operate (Wolgin, 1975; De Boer and Chandra, 1978; Dillon and
Anderson, 1971; Yotopoulos, 1967; Hopper, 1965). These types of studies have been primarily directed towards considerations of risk, allocative efficiency and the nature of constraints on farm production systems. Barnett et al (1982) used a linear programming (LP) model to test various objectives of Senegalese farmers. One study that is most closely related to our study in terms of objective is that by Heyer (1971) of small farmers in Kenya. Heyer used a linear programming model to estimate shadow prices for resources, thereby estimating the effective constraint of resources on the production process. A Linear Programming Model: Rationalization and Specification
Based on the model selection criteria discussed above, a linear programming model was selected as being the most suitable empirical procedure for evaluation of the system. One advantage of a linear programming model within the context of traditional agricultural production system such as Cameroon's is that it avoids the question of allocative efficiency (Heyer, 1971).
Although an LP model arrives at a solution by maximizing the model objectives, it avoids the assumption that technical maximization has already occurred (Langham, 1968). Instead, the existing
(although possibly sub-optimal) allocative efficiency rates are determined and used in the model. The model solution then points to activities and levels of allocation that should be pursued if the objectives of the model are desirable.

Further, given the objectives of the LP model, the shadow prices of the resources are a measure of the constraining effect of those resources. The degree to which the objectives and
constraints of an LP model match those of producers in the farm production system being represented, is a function of the
specification of the actual model. Although an LP model must of necessity maximize or minimize some objective function, the model
readily lends itself to attain other objectives through the use of the constraints (Langham, 1968). Empirical studies by Barnett et al (1982) indicate that the inclusion of other objectives besides profit maximization may not improve the performance of the model. In their study of Senegalese small farmers, a multiple objective
model did not exhibit superiority over a profit maximization model with similarly-structured constraints. Also, a deterministic model requires fewer observations to structure the model, which in the context of the data limitation problems of developing countries, could be an important consideration. Finally, within a resource allocation LP model, the dual provides shadow prices or relative measures of the resource constraints of their associated resources. In this way such a model can show which constraints
are the most seriously binding and which merit the most research.
The general algebraic notion, the primal form of the LP model specification,explicitly assumes the objectives of profit maximization, food self-sufficiency and risk minimization. The
model is:

5 15
-PZ, + E PY (
W,1 k-1
Subject to:
Land constraints.
j X,, g d, (s=1,..4) (2)
Labor constraints.
4 8 5
_B Z. :c b. (n=1,. 4) )
s=1 1=1 1-4
Purchased input identities.
4 8
Z, ,E E A4 = 0 (i=1,..3) (4)
S=1 J-1
Consumption requirement constraints.
4 8
E i c4,Xv Yk a hk (k=1,..15) (5)
s-1 j-1
Xs, Yk Zil, Vk > 0 for all s,k,i,j. Indices:
i) = 1,..3 indexes purchased inputs, while i = 4,5 are
male and female purchased labour
k) = 1,..15 indexes the crops produced by the household.
k = 1,..13 are crops which are considered annually.
k = 14,15 are crops which are considered monthly.
s) = 1,..4 indexes the land types available to household.

j) 1,..8 indexes the activities, or crop mixes, for
which land can be used. j = 1,..5 are activities in the first season, while j = 6,..8 are activities in
the second planting season.
n) 1 1,..4 indexes the types of labour required.
Variables and coefficients:
Pi = market price of the ith input
Zi = quantity of the ith input purchased
Pk = market price per bucket of the kth crop
Yk = quantity sold of the kth crop, in buckets
Xsj = hectares of land of type s devoted to activity j
ds = hectares of land of type s available
Bsjn = hours of nth labour type required for one hectare of
land of type s devoted to activity j
Zin hours of purchased labour of type n
bn = hours of nth labour type available
Asji = quantity of ith input required for one hectare of
land of type s devoted to activity j
Csjk quantity of kth crop produced by one hectare of land
of type s devoted to activity j in buckets
hk minimum quantity of kth crop required for household
consumption, in buckets.
In the model, equation (1) represents the maximization of net returns to resources by the household. Net return is defined as the gross returns from farm products sold less the cost of inputs purchased. The amount of product sold is the amount produced less that reserved for home consumption. Equation (2) is the constraint allocating the household land resources to the various production alternatives or crop mixes. Equation (3) is the constraint

allocating household labour, borrowed labour and hired labour, to
the production alternatives. Equation (4) is condition guiding the allocation of purchased inputs to the production alternatives and equation (5) represents transfer constraint equations allocating
the production of each output to either home consumption or sale.
The linear programming model was separately specified for each of the four household groups. Hyperlindo mathematical
programming computer software was used to solve the model. Two detailed forms of the LP model was specified and these are
referred to as Model 1 and Model 2. Model 1 has as its operational objective the determination of land and other resource constraints on income potential of farmers. Model 2 objective is to assess the relative value of alternative cropping activities of each group of farmers and predict their response to factor and product pricing policies. Each model was solved for a number of variation via a series of runs. Model 1 had 4 runs: run 1 = measured land values, without a binding capital constraint; run 2 = varying land
availability constraint, without a binding capital constraint; run 3 = holding land f ixed, and varying the capital constraint values; run 4 = varying both land availability constraint and capital constraint values. Model 2 had 3 runs: run 1 = measured
coefficients, without a binding capital constraint; run 2 =
varying price received for coffee, without a binding capital
constraint; run 3 = varying price paid for fertilizer, without a

binding capital constraint. The detailed forms of models 1 and 2 are presented below.
Model 1
Model 1 holds the proportion of land allocated to coffee and foodcrops fixed, based on the assumption that in the short run, land cannot be converted from one enterprise to the other. Objective function:
3 7 5 13 7 15
t=1 1-1 1-4 W1 t-1 k-14
pit = price of one hour of ith labour during period t
Zit = hours of ith labour purchased during period t
Pkt = price of crop k during period t
Ykt = quantity of crop k sold during period t
IN = average annual amount of income received from other
agricultural sales, such as fruit, chickens, etc.
FC = average fixed cost of annual general agricultural
purchases, including tools and equipment.
The objective function maximizes the revenue from crop sales less the out of pocket cost of crop production. The time periods represents the months during the period of the second phase survey (November 1988 to April 1989) plus a single period representing the rest of the year. There are therefore 7 time periods. Constraints:
Land allocation constraints, in hectares:

8 (7)
There are four land types: 1 = land available for coffee production; 2 = land available for coffee food crop associated cropping; 3 = land available for food crops in seasons 1 and 2; 4 = land available for food crop in season 2. Constraint (7) limit the amount of land of a specific type to the amount of that land available to the household. The eight cropping activities are: 1 = coffee production with land used throughout the year; 2 = food crops intercropped with coffee (first season); 3 = beans associated with potatoes (first season); 4 = corn associated with food crops (first season); 5 = cabbage (first season); 6 = beans with potatoes (second season); 7 = cabbage (second season); 8 = red cabbage (second season). Labour allocation constraints, in hours: Type 1 labour constraints (female and large child):
4 a 1
E b1,,,Xsjt (F7)l, -#(c17)it c 0 (t=1,7)(8
-1 j=2
Type 2 labour constraints (female and children)
Ej2jt, (F76 o(C12t (C27)t (t=1I..M (9)
S-I J=l

Type 3 labour constraints (all household members)
Ib3,pXj, (F7)3, IO17) 7)3,(10)
8-1 1 2
Type 4 labour constraints (males only)
E jb.jXaJ, (MT)4, 1 0 (t1... (II)
5.1 i-1
(FT)nt = hours of female labour available to type n labour
activities in period t.
(C1T)nt = hours of large child labour available to type n
labour activities in period t
(C2T)nt = hours of small child labour available to type n
labour activities in period t
(MT)nt = hours of male labour available to type n labour
activities in period t
There are four labour types, differentiated by gender, age and task: 1 = tasks assigned to female or large child; 2 = tasks assigned to female, large child or small child; 3 = tasks assigned to general labour, including male, female or child; 4 = tasks assigned to males. A large child was considered to be older than 10 years, while a small child was one of 10 years or less. Based on discussions with local farmers, equivalencies were developed for adult and child labour inputs. One hour of large child labour was equivalent to one-half hour of adult labour, and one hour of small child labour was equivalent to one-third hour of adult labour. Constraints (8) though (11) allocate available labour to the 4 labour types.

Female labour constraints.
E (F7), Z4, f, + fb, (t=1 ,..7) (12)
Large child labour constraints.
E3 (CI 7)4 T l CIS ,.7 (13)
Small child labour constraints.
, (C2)., c2, (t=1,..7) (14)
Male labour constraints.
4 (MT),~ mt +mrb, (=1,..7) (15)
ft = hours of household female labour available in period t
fbt = hours of borrowed female labour available in period t
cit = hours of large child labour available in period t c2t = hours of small child labour available in period t
mt = hours of household male labour available in period t
mbt = hours of borrowed male labour available in period t
Z4t = hours of purchased female labour in period t
Zst = hours of purchased male labour in period t.
All of the labour constraints (8) through (15) limit the labour allocation to the respective labour types to that available either from household members, borrowed, or hired.

Purchased input identities.
4 $ 7
#-1 J-1 -1
Purchased input identity (16) allocate purchased inputs to cropping activities in the production process. The three types of purchased inputs are: 1 = fertilizer purchased from coffee cooperatives, assumed to be allocated to coffee production; 2 = fertilizer purchased on the open market, assumed to be allocated to foodcrops; 3 = other variable inputs, particularly seed and gasoline or irrigation pumps. Household consumption constraints for crops considered annually.
4 8 7
E E E Cjk,Xaj,- Y k hk (k=1,..13) (17)
-I j-1 t-1
Household consumption constraints for crops considered monthly.
4 S
E j CjX.j, Y, hk, (k=14,15) (t=1,..7) (18)
0-1 j-1
Csjkt bucketsI of crop k produced per hectare of activity
j on land type s in period t
Y = buckets sold of crop k
Ykt = amount sold of crop k during time period t
hk = household consumption requirement for crop k
hkt = household consumption requirement for crop k during
time period t.
'Units of 15 liter buckets were used to measure production and sales except in the cases of bananas and plantain, which were measured in bunches, and corn, which was measured in sacks.

Constraints (17) are balance equations which move the production of crops to sales and home consumption. Constraints
(18) balance the production of crops for which prices vary monthly to sales and home consumption.
Capital constraint.
3 7 6
iI s*i 1-4
A capital constraint was introduced to test the effect of the availability of capital on the system. The terms of the left-hand side are the expenditure terms from the objective function, and represent the total out-of-pocket costs, in CFA,2 of production. The coefficient e is the amount, in CFA, of capital available to the household.
Model 2
Model 2 allows land to be allocated freely to either coffee or foodcrop production. Further, households are able to purchase foodcrops on the local market to satisfy household consumption needs. This is assumed to free land and labour inputs from necessarily producing for subsistence, so that the highest level of income generation can be attained. Thus, the only difference between Model 1 and Model 2 are the constraints pertaining to land and foodcrop allocation.
2The CFA is the Central African Franc, a currency unit shared by many central and west-African nations, particularly those who were formerly French colonies. The CFA is tied to the French Franc
(FF) at a rate of 50CFA = 1FF. In March 1989 US$1 = 310CFA.

Land constraints.
E X gd (20)
X is hectares of land devoted to activity j.
X 4
Xg EXi (21)
J-6 J-1
Constraint (20) allows the total household land to be available to planting by first-season cropmixes (j=l,..5). Coffee is considered as one of these. Constraint (21) requires that foodcrop land that is replanted during the second season (j=6,..8) cannot exceed a certain proportion (g: (0 Cjj *k + W k (=..3 (22)
-1 J.1 t-i
Wk is quantity purchased of crop k.
Household consumption constraints for crops considered monthly.
4 i CJk,,X.,, Y, + W, k h, (k=14,15), (t=,..7) (23)
S1 J-1
Wkt is quantity purchased of crop k in period t.
The household consumption constraints are similar to Model 1 (constraints (17) and (18)) except that in this case the purchase of food commodities to meet household consumption requirements is permitted. A number of coefficients had to be calculated to 15

operationalize the linear programming models. These include: (1) hourly labour wages, (2) input prices, (3) crop prices, (4) other agricultural income, (5) land availability, (6) labour availability, (7) input application rates, (8) yield rates, (9) labour application rates, (10) household consumption requirements,
(11) household expenses. Details of these calculations are
available from the authors.
Study Area
The data used in the analysis were obtained from the chiefdom of Bafou in Cameroons' West Province. The chiefdom of Bafou is one of the largest and most important of the Bamileke chiefdoms. Bafou is located in the Department of Menoua,3 on the western edge of the West Province, and is at its closest point, about 10
kilometers (km) from the city of Dschang, the departmental center. Dschang is also the location of the University Center of Dschang, the national agricultural university; patterned after the Land Grant University System of the United States. The Bafou chiefdom is roughly rectangular shaped, with dimensions 28 kms. long and 8 kms wide, with a surface area of 162 km2. A range of escarpments divides the terrain into roughly highland and lowland areas. The
3A "Department" is an administrative unit. The West Province has five administrative units (also called districts) Haut-Nkam, Menoua, Mifi, Nde and Noun. The provincial capital of Menoua district is Baffousam, which is the sixth largest city in Cameroon (86,000 inhabitants). The district of Menoua has about 201,409 of the West Province's 968,000 population.

highland area covers the northern 10 kms. length of the chiefdom with elevations ranging from 1600 to 2740 meters. The lowland area occupy the southern 18 kms. length of the chiefdom, with elevations ranging from 1400 to just under 1600 meters (Bergeret et al, 1988). The climate is humid yet moderate, tempered by the relatively high elevation. Annual rainfall averages 1.9 meters, which is spread over an 8.5 month period, beginning in early March and ending around mid-November. The annual variation in rainfall pattern is considered the least varied in Cameroon. In a study of the rainfall pattern over a 30 year period, enough rain had fallen (25mm) to plant by March 31 in 9 years out of 10 (Bergeret et al, 1988).
The Bafou chiefdom is divided into more than 80 "quartiers" (neighbourhoods), each traditionally governed by a sub-chief, whose authority is delegated by the chief. Favourable climate and fertile volcanic soils contributed to the development of a traditional agricultural system of associated cropping of corn, beans, plantain, bananas and rootcrops. During the colonial period, the plantation system of agriculture was introduced, with Arabica and Robusta coffee varieties as the major plantation crops. Coffee was rapidly adopted by small farmers between 1940 and 1960, such that at the present time coffee in association with food crops has become the dominant agricultural practice (Bergeret, et al, 1988). The major food crops grown are corn, beans, cocoyams, yams, potatoes, cassava (manioc), bananas and plantain. Also grown, but in smaller quantities are peanuts,

melons, onion, hot peppers and a variety of greens such as
amaranth. There are two cropping seasons, the first beginning with the rains in March, and the second starting in August or
September. The f irst rainy season is the most important in the production process, since the bulk of the annual plantings occur
during this period. Plantings during the second rainy season is approximately one-quarter of the land planted during the f irst. Although there is wide variation in crop mixes among plots, the following three general cropping mixes are typical of the farming system.
1) Cof fee/Foodcrop Mix: Cof fee is planted at the usual density
of 2500 plants/hectare, or spaced 2 meters apart. Between
these coffee trees are planted any and/or all of the major
crops. However, very little is planted in this type of system during the second season. Responsibility for cropping
decisions may be divided in cases where the cultivators of the coffee and foodcrops are different individuals. Chemical fertilizers are applied to coffee trees and are thus available to the other crops. Fungicides are also applied to
2) Corn Mix:- This is planted exclusively during the f irst
season, and consists of corn and primarily beans, but often with some combination of cocoyams, potatoes, cassava and
yams. Chemical fertilizers are sometimes applied.

3) Bean/Potato Mix:- This can be planted during both seasons,
and occasionally includes small amounts of crops such as
cocoyams. Chemical fertilizers are usually applied.
Agricultural roles within the traditional farm households are generally differentiated along gender lines. The tasks for males
are those associated with coffee, including planting, fertilizing, spraying, pruning, harvesting, depulping and drying. In the
traditional system, among the food crop enterprises, only plantains are cultivated and harvested by males. There are
indications, however, that the traditional gender designated roles are in transition. Some males, particularly younger farmers, do plant and harvest their own food crops, particularly those
produced for market. Also, women own and tend their own coffee plots. Women perform the bulk of agricultural labour, including participating in all of the coffee tasks, except pruning which is
designated exclusively as a male task. For food crops, land
preparation, planting, cultivating and harvesting are all
performed by women and women process and market the food crops. Children in the farm household generally participate in most of
these tasks. Young boys will assist women in their tasks, but with age both boys and girls assume their traditional roles.
Data Elicitation Procedures
Data relating to the production processes of typical small farm households were elicited in a structured format over an eight month period from September 1988 to April 1989 in the Bafou

chiefdom. To determine the characteristics of a typical farm, knowledge of the essential characteristics of each household type is essential. Sample survey procedures were used to acquire this type of information. Such survey procedures need not be random.
A selection choice based on judgement will be at least as good and perhaps better than random choices in determining the typical characteristics. However, Mellor (1969) agree that a beginning knowledge of the population to be studied is essential. In keeping with these considerations a two-phase data elicitation procedure was utilized to generate information required in the linear
programming model. The first phase consisted of a rapid informal survey designed to provide: (1) descriptive analytical parameters of the agricultural system in the study area:(2) an overview of the general production and resource allocation patterns and (3) guidance to the nature of the constraints to be faced in
conducting the formal survey. The second phase consisted of a formal sample survey of small farmer households. This phase was designed to generate data sets to be utilized in the linear programming model specified in the empirical procedures.
Phase I Survey
During the first phase informal survey, 24 farm households were identified through consultation with the area's Agricultural
Extension Posts, as representative of the farming system of the Bafou chiefdom. At each location, both small and large farm households were visited. Each household head was visited in plots
for approximately 90 minutes, and attempts were made to talk to at 20

least one female member of the household. The survey was conducted in an informal manner, and all questions and responses were
recorded on voice tape. Questions were asked regarding number of women and children in the household, land available to the
household, land sources, amount of land allocated to coffee and other crops. Coffee production and land allocation data were recorded from each farmer's coffee cooperative book. General information was gathered on the production process, the agricultural cycle, labour role in agriculture, availability and
use of inputs, markets f or and sale of agricultural production, and land acquisition patterns.
Analysis of the informal survey data clarif ied a number of important points. It was made clear that much of the agricultural production, particularly rootcrop and banana/plantain, was
harvested and sold on a day-to-day basis. Also, labour use was seasonal, and with the exception of cof fee sales records, no records existed of either labour use or crop yields. Farmers, however, were able to recall the amounts harvested of crops such as corn and beans, which generally were harvested at one point in
time, with the exception of those plots that were close to f arm residences. Furthermore, the land holding size data are
approximate, since few f armers know the actual size of their holdings. Farmers, generally would have a rough idea of how much land was allocated to coffee, because a coffee cooperative
extension worker would have made such a size determination at some point. These farmers generally had little idea, however, of the 21

size of their f ood plots. All in all, one important conclusion derived from the first phase informal survey, was that important elements of the data concerning the production process (and the LP
model) could not be elicited through single interviews of f armers or even over several visits.
Phase 2 Survey
The approach to the second phase formal survey was informed
by the results of the f irst phase informal survey. Thus, the decision was made to select a small number of representative farm
households and conduct repeated visits in order to obtain the daily data necessary for the estimation of labour allocation and
certain crop production levels. Further, a small sample would allow all of the households" land holdings to be visited and measured in order to ascertain an accurate measurement of land and other factor allocations. It was recognized that a small sample survey would carry a much smaller probability of capturing representative households, compared to a statistically large sample. Budget and other constraints precluded the use of a large
sample survey. It was therefore decided to use the large sample data base available for the study area, generated by the project
known as "Operation Bafou",4 to select households most closely matching the typical households described by that data base. The use of this procedure facilitated the selection of farm
4"operation Bafou"l is a data gathering project that has been in operation for several years. The project is administered and
conducted by the Interdisciplinary Farming System Research Team from the University of Dschang.

households, which although not randomly generated, nonetheless permitted a selection of households that were representative, in that they matched known parameters of household size, holding size, and typical agricultural activities.
The second phase or primary survey was implemented via two separate questionnaires. One questionnaire was developed to record the daily activities of the household. The other questionnaire was developed to record data relating to a particular plot of land cultivated by the household. The two types of questionnaires were administered by trained interviewers who were proficient in the local dialect of the Bamileke language. A representative sample of 12 farm households were selected based on the "Operation Bafou" sample frame. This number of households was determined to be the largest that could be operationally accommodated, given the
comprehensiveness of the data requirement and other constraints. However, because of certain data problems, only 10 farm households are included in the empirical analysis. Selected characteristics of these 10 farm households are presented in Table 1.
As shown in Table 1, three quartiers or neighbourhoods
represents the farm production variation found in Bafou, the study area. The three neighbourhoods are Tsinfou, Lefe and Loung. Tsinfou is located in lowland terrain and is considered an average quartier for the study area as a whole, with moderate population density and holding sizes. A total of 3 representative households were selected from Tsinfou neighbourhood, 1 large size household and 2 small size households. Lefe is also located in lowland
terrain, where holding sizes are typically small, and population

Table 1. Selected Characteristics of Representative Sample of
Small Farm Households
Large Size' Lowland Household Group
No. of
No. of No. of Children Total Ha. Av. Ha.
Area Quartier House- Adults in in House- of Coffee of Coffee
(Neighbourhood) holds Households holds land land
Tsinfou 1 4 11 2.62 2.62
Lefe 1 4 11 2.28 2.28
Subtotal 2 8 22 4.90 2.45
- - Small Size* Lowland Household Group - Tsinfou 2 4 5 1.47 0.73
Lefe 2 5 8 1.37 0.69
Subtotal 4 9 13 2.84 0.71
- - Large Size Highland Household Group - Loung 2 8 29 4.49 2.46
Subtotal 2 8 29 4.49 2.46
- - Small Size Highland Household Group - Loung 2 3 7 0.80 0.40
Subtotal 2 3 7 0.80 0.40
GRAND TOTAL 10 28 71 13.03 1.30
Source: Second Phase Formal Survey. aHousehold size is based on the number of adults in the household and the amount of land allocated to coffee production. A small
size household is one with 1 or 2 adults and less than 1.00 hectare of coffee land, while a large size household is one with 4 or more adults and more than 1.5 hectares of coffee land.
density is very high. A total of 3 households were selected from Lefe quarter, 1 large size and 2 small size. Loung is located in highland terrain, and is characterized by low population density, large size holdings and a high incident of market garden
(non-traditional cash crop) cultivation. A total of 4 households were selected from the Loung neighbourhood, 2 large size and 2 24

small size. Summary characteristics of the 10 sample households reveal the following aggregates: 28 adults, 71 children, 13.03 hectares of coffee land and 1.30 average hectare of coffee land
(Table 1). Summary characteristics of average size of crop land allocations and yields are presented in Table 2. These figures are the basis of the L.P. results.
Enumerators visited the sample households in each quarter twice each week on designated days of the week. Each of the adult members of the households were interviewed on each visit, at which time both types of survey questionnaires were administered.
It was established earlier from the first phase survey that the standard unit of measurement for marketing most foodcrops in the
area was a 15 litre bucket. Each of the sample household was provided with a similar bucket and instructed that all information gathered regarding yields, seed and even fertilizer use, would be expressed in bucket units. Other common units of measurement used were the coffee sack and fertilizer sack. Equivalencies in
measurement units were established as 1 coffee sack = a buckets,
and 1 fertilizer sack = 4 buckets. In administering the plot
questionnaire, plot measurements were made with a 15 meter length of rope marked at 1 meter intervals.

Table 2. Summary Data for Representative Small Farm Households Household Group
Small Large Small Large
Crop Land Type Lowland Lowland Highland Highland
--- ----------Average Hectare---- -- -Coffee land 0.498 1.796 0.342 2.246
Coffee food crop
land 0.586 1.580 0.351 1.873
Food crop land
(Season 1) 0.193 0.831 0.401 3.945
Food crop land
(Season 2) 0.073 0.296 0.114 0.917
Average Yield
----- ------------Kg./Hectare--- -- ---Coffee production 296.7 294.0 491.1 726.2
------- ---------Buckets/Hectare---- -- -Food crops
Cof fee mix:
Potato 0.2 1.2 0.0 1.1
Beans 15.2 19.0 1.9 5.9
Comna 186.1 31.5 0.0 38.0
Corn mix
(1st Season):
Potato 51.9 166.3 133.6 294.0
Beans 20.9 21.8 20.7 5.8
Corn 320.1 71.0 53.5 76.4
Beans-potato mix(2nd Season):
Potato 3.6 29.6 126.1 129.5
Beans 16.7 33.4 19.3 19.3
Source: Second Phase Formal Survey aCorn yield measured in sacks/hectare.

Ef f ects of Resource Constraints on Production Systems and Earnings Potential
Results f rom run 1 of Model I indicate that the predicted optimum allocation of land to alternative crop mix generally
matches the actual household land allocation as observed during the survey. In all cases, all of the available land was allocated by the model to cultivation. This result, however, does not especially underscore the constraints imposed by land
availability, since in the model specification, land is the only effective constraint, since labour, and all other inputs can be purchased and capital was not binding. Table 3 compares the run 1
allocation of land to actual allocation. The primary difference between the model allocation and observed values is the amount of
land left fallow. For example, in the case of the first season corn mix planting, large highland farmers only planted 0.78 hectares in actuality, while the model allocates 3.9 hectares to the mix. The model naturally allocated all land to production as
long as net returns to labour and other inputs were positive. This difference could be a reflection of capital constraint which was assumed to be non-binding in the model. It could also, reflect a recognition on the part of the farmers to the limitations of land to sustained annual droppings. Those farmers who still choose to plant all of their land may simply be experiencing greater land
constraint with respect to their household consumption needs.

Recall that Model 1 does not allow for the purchase of foodcrops and that household consumption requirements were calculated based on observed production from existing crop mix allocation.
Table 3. LP Model Land Allocation and Actual Land Allocation, All
Household Groups
Household Group
Small Large Small Large
Lowland Lowland Highland Highland
Land Type Model Actual Model TActual Model Actual Model Actual
--- --------- Hectare------Coffee Production Land
Coffee 0.498 0.498 1.796 1.796 0.342 0.342 2.246 2.246
Coffee/Foodcrop Land
Foodcrop intercropped
w/coffee 0.586 0.586 1.580 1.580 0.351 0.337 1.873 1.873
Foodcrop Land, Season 1
Corn mix 0.193 0.188 0.831 0.780 0.391 0.340 3.904 0.778
Bean & potato mix NA* NA 0.000 0.048 0.000 0.047 0.000 0.154
Cabbage NA NA NA NA 0.010 0.014 0.041 0.339
Fallow 0.000 0.005 0.000 0.003 0.000 0.000 0.000 2.674
Foodcrop Land, Season 2
Bean & potato mix 0.070 0.070 0.296 0.287 0.104 0.092 0.597 0.523 Cabbage NA NA NA NA 0.010 0.022 0.253 0.264
Red cabbage NA NA NA NA NA NA 0.067 0.069
Fallow 0.000 0.061
*Not applicable.
The allocation of all available land in the model does
indicate clearly that there are positive returns to agricultural inputs in general, and that in the context of the existing constraints, larger land holdings are desirable for all groups. The model allocation could also reflect a response to yield
uncertainty, whereby a large area is planted to ensure consumption needs. This is true even at relatively high levels of land
availability (as shown in run 2), where the proportion of hired to

household labour is high, indicating positive returns to hired labour. The model indicate that the cabbage crop is not a preferred crop mix in terms of land allocation to each particular
crop mix, given the model constraints. As indicated in Table 3 only the two household groups in the highlands grow cabbage. In each of these household groups the model allocated only enough land to cabbage in the first season to meet household consumption requirement for that crop. Only in the large highland group does the model's land allocation solution include cabbage (red)
produced for sale in the second season. The foodcrop mix most f avoured by all of the groups was the f irst season corn mix. In all household groups, more of the corn mix was in the model solution than the actual plantings. The bean and potato crop mix for the first season was not selected by the model for any of the
household groups, yet 3 of the 4 household groups actually planted this crop mix. In the cases of coffee production land and
cof fee/ foodcrop land (the same land) only one mix is possible for each of these land categories. In all household groups, the model elected to plant all of the land available to these crop mixes, indicating positive net returns to coffee. observation of actual foodcrop land plantings for the first and second seasons indicate significant reduction in plantings during the second seasons. On the average, the two highland groups planted about 26 percent of the first season foodcrop land, while the two lowland groups
planted about 36 percent. This would suggest inter alia, that

highland farmers are less pressured by land and are able to leave more land in fallow during the second season.
An interesting and informative dimension of the land constraint question has to do with the relative scarcity of land
among household groups. As indicated earlier, the dual of the linear programming model produces the dual of the resource constraints and reflect the rate of change in the objective function, with respect to each associated constraints. These dual
values are referred to as "shadow prices" or "implied values" for the limited resources that are causing the constraint. A resource
whose availability poses no constraint to the model would carry an implied value of zero. The implied values of land can be taken to reflect the scarcity of land (and the value of the productive capacity), given the available resources and technology. Thus, the generation of implied land values for the four household groups would provide some indication of relative land scarcity among the
groups. In general, it would be expected that households with smaller land holdings would have higher implied land values, under similar technology and farming practices, since they would have
the available labour to make use of more land. This condition, however, would depend on the relative access to other resources,
particularly labour. Survey data indicate that in the highland area, where land is generally considered to be less scarce, that
the labour/land ratios are smaller than in the lowland area,

reflecting a greater availability of land to labour.5 It would be expected then that lowland households would reflect higher implied values for land.
Table 4 shows the calculated implied land values for each household type under run 1 of Model 1. Note that two types of implied land values are computed in the table. One set of implied land values in Table 4 are for coffee land, coffee/foodcrop land, first season foodcrop land and second season foodcrop land. A comparison of these values provide non-conclusive results, since there is no correlation of land value to household size. The relative values of coffee land, however, match closely the relative yield rates (Table 2) that were computed for coffee for each household groups. The coffee yield for both lowland groups is approximately 295 kg/hectare, while small highland farms produce 491 kg/hectare, and large highland farms 726 kg/hectare. The
implied values of the land used for that purpose rise with increased yield. As a general rule, implied land values were higher for first season foodcrop land than for coffee land. This finding is consistent with the model results reported in Table 3, where the corn mix was favoured and the relatively high revenues available from the sale of corn, potatoes and beans. The second set of implied land values in Table 4 are estimates of total weighted average implied land values for coffee land and food crop land.
5The ratios of total annual household labour to total land holding for all households are: small lowland = 8,482 hours/ hectare; large lowland = 5,011 hours/hectare; small highland = 4,213 hours/hectare; large lowland = 3,195 hours/hectare.

Table 4. Implied and Weighted Average Implied Land Values, all
Household Group
Lowland Highland
Land Type Small Large Small Large
- - Implied Land Values - (CFA/Hectare)
Coffee land 105,200 128,400 213,100 333,200
Foodcrop land 262,200 77,400 162,400 67,400
Foodcrop land
ist season 370,100 305,400 200,800 379,800
Foodcrop land
2nd season 27,200 72,700 420,100 207,400
- - Total Weighted Average - Implied Land Values
Coffee land 367,400 205,790 376,510 400,570
Foodcrop land 379,990 331,270 320,140 427,990
Weighted average 370,520 245,490 345,990 418,050
These weighted average implied land values were calculated using a modified version of the formula used to compute the implied land values for the upper portion of the table. The total weighted average figures are the average value of all land to the farmer based on coffee and foodcrop holdings. The weighted average
implied land values indicate as expected, that land values are higher for small farmers, and lower for large land holders, with the noted exception of large highland farmers.

Closer examination of the pattern of implied land values was
conducted through the relaxation of the land constraints, allowing increasing land availability to each household group, and
calculating implied land values (Model 1, run 2). The land constraints for each land type were relaxed by equal percentages. In each household group implied land values were assessed under conditions of an increase in land availability to a maximum of 300 per cent of existing land holdings. Land holdings beyond that point were considered unrealistic in light of the requirement for
high proportions of hired labour. Table 5 shows total weighted average implied land values for all household groups under these
conditions. Note that in Table 5 only land values for which the basis of the model is changed are presented, so that the
incremental increases shown are not all equal. The implied land values for the small and large lowland farmers and the small highland farmers follow roughly the same trend. The results for the large highland farmers are significantly different. The
patterns exhibited in the total weighted average implied land values are also reflected in the calculated implied land values for coffee land and foodcrop land. These results are interpreted as reflecting a fundamental difference in productivity of inputs, including land, labour and other factors, and could also reflect
some economies of size. The large highland farmers are probably operating at a level of technology which is superior to the other

Table 5. Total Weighted Average Implied Land Values, all
Household Group Land
Total Large Small Large Small
Land Lowland Lowland Highland Highland
Hectares 1000 of CFA/Hectare
0.8 370.5 INFa 346.0 INF
0.9 363.7 INF 346.0 INF
1.0 363.7 INF 327.0 INF
1.1 363.7 INF 327.0 INF
1.2 325.9 INF 327.0 INF
1.3 325.9 INF 327.0 INF
1.4 284.4 INF 293.9 INF
1.5 284.4 INF 263.7 INF
1.6 259.3 INF 263.7 INF
1.7 255.7 INF 263.4 INF
1.8 255.7 INF 260.8 INF
1.9 233.0 INF 260.8 INF
2.0 220.0 INF 260.4 INF
2.1 220.0 INF 259.5 INF
2.4 NAb INF 259.5 INF
2.5 NA INF 259.5 INF
2.6 NA 245.5 199.6 INF
2.9 NA 245.5 199.6 INF
3.2 NA 220.7 NAb INF
3.4 NA 195.1 NA INF
4.0 NA 195.1 NA INF
5.0 NA 195.1 NA INF
5.8 NA 149.4 NA INF
6.1 NA 142.0 NA 418.1
6.8 NA 140.8 NA 418.1
7.1 NA 138.0 NA 418.1
7.4 NA 138.0 NA 415.3
8.0 NA NA NA 403.7
8.6 NA NA NA 384.9
9.0 NA NA NA 384.9
10.0 NA NA NA 384.9
11.0 NA NA NA 362.0
alndicates infeasible solution due to constraints of household consumption requirements.
blndicates solutions are no longer applicable because of unrealistic land availability.

groups. The lowland as well as the small highland f armers f ace generally the same land constraints, since their resource base, factor efficiency, activities productivity and available alternatives are not markedly different.
Although not presented here, it was observed that implied land value f or cof fee the traditional "cash crop", is not markedly dif ferent f rom that of f oodcrop land. Indeed, there is strong evidence suggesting that the implied value of the cof fee land, particularly for the two small farm household groups (lowland and highland) comes from the value of foodcrops planted in association with coffee. Examination of data collating
increasing land availability with implied land value, shows that in 3 of the 4 household groups, f oodcrop land exhibits a higher implied value. In all f our household groups implied land value declines more slowly when available land is increased. Labour
Results f rom run 1 of Model 1 indicate that labour is not generally a constraint at existing levels of land availability. Seasonally, however, some labour constraints occur in two of the
household groups. Although the model allows f or the hiring of labour, labour was only hired in two cases: (1) by the small lowland group during the month of March (94.3 female hours) and
(2) by the large highland group during March (499.9 female hours) These household hired labour quantities represents 3.6 per cent and 4.3 per cent of total annual labour for small lowland and
large highland groups, respectively. According to the results, as 35

land is made available to households (run 2, Model 1) labour
becomes increasingly more constraining. Table 6 shows the
relationship of hired labour to total labour under increasing land availability. In the case of both large highland and small lowland farmers, the proportion of hired labour increases to approximately 11 percent with a 50 percent increase in available land. All of
these increases come from f emale labour. 6 In the case of large highland households, although only a small percentage of labour is hired at existing levels of land availability, that percentage represents a large quantity of labour demanded from the local labour market. Local labour supply might not be available to meet that demand, or capital constraint might not permit the hiring of that quantity of labour. In either case, the net effect could be
less than optimal allocation of land. This finding adds further insights into the earlier discussion regarding the relatively lower proportion of cropland planted to foodstuffs by the highland household group (26 percent) during the second season, compared to lowland household groups (36 percent). Capital
Capital constraints may have significant impacts on the production systems. In all of the model results discussed so far,
a solution was feasible as long as revenues exceed the out-ofpocket costs of production. In order to assess the impact of
6Model solutions include hired male labour after a 90 percent increase in land availability in the case of large highland farm
households, and after a 120 percent increase for small lowland households.

Table 6. Hired Labour as Percentage of Total Labour with
Percentage Increase in Land Available, all Households Household Group
Increase in
Total Land Small Large Small Large
Available Lowland Lowland Highland Highland
- - - Percent - -
00 3.6 0.0 0.0 4.3
10 5.2 0.0 0.0 5.8
20 6.5 0.7 0.0 6.6
30 7.9 2.0 0.5 6.9
40 9.1 4.1 1.4 8.5
50 10.9 5.9 2.2 11.0
60 13.0 7.6 2.3 13.3
70 14.9 9.1 3.5 15.5
80 16.6 10.5 4.3 17.8
90 19.0 11.7 5.6 21.1
100 21.8 12.9 7.7 24.3
150 36.1 22.9 21.3 38.0
200 51.5 34.0 32.4 46.5
capital limitations the model was run over a range of capital constraint values (run 3, Model 1), and the calculated implied value of capital plotted against capital availability. Two data sets are generated, one reflecting actual land constraints and the other reflecting available land equal to 150 per cent of actual land holdings. The results begin at the minimum amount of capital sufficient to arrive at a feasible solution, and continue until the implied value of capital goes to zero, in which case capital
is no longer a binding constraint. In all cases, the implied value of capital is very large at the point of minimum feasibility. The largest implied value occurs for large highland households, where at a minimum feasible amount of capital (227,000 CFA per
household), the implied value is more than 14 times the actual 37

capital. This reflect the relatively high productive potential of these farmers. In all four household groups, however, the implied value of capital is initially large and remain positive for some
time, as capital availability increases. with increased capital availability, at 150 percent of actual land holdings, the implied
value is initially generally greater, and remains positive over an even wider range. capital is therefore highly valued by the production system at levels close to minimum amount feasible. As such, a lack of capital availability would severely constrain the income generating potential of the production system.
Capital constraint effects at different levels of land
availability are examined for a range of land holding levels (run
4, Model 1) Results indicate that the minimum capital requirement is the same f or any level of land holding for each household group, reflecting the minimum requirement of meeting household consumption demands. With the exception of the large highland household group, the results show that at the smallest level of land holding (actual) the capital requirement range is quite small. We further computed the ratio of capital requirement to land used at the point where the implied value of capital goes to
zero. Results indicate no markedly different ratios, with the least amount of capital required per hectare found among the large highland farmers. Similarities of these capital/hectare ratios, combined with the proportion of land left fallow by the large highland farmers, suggest differences between these farmers and

the other household groups in terms of adequacy of land holdings size, with respect to household consumption requirements. Farm Earning~s
Since of f-f arm employment activity was negligible in all f our household groups, crop sales were the only means of income
generation specified in the model. Coffee sales for each household group generated in the solution (run 1, Model 1) matches observed sales. In the case of foodcrop sales, the model solution had much higher sales than were actually observed, particularly with
respect to corn, beans and potato. This was due to the allocation
of land to those crops in the solution. Table 7 shows a comparison of the model's prediction of foodcrop sales to actual sales in terms of percentage of total production. As indicated in the
table, not only are sales of corn, beans and potato higher in the model solution, but sale of cabbage is generally lower.
For the small highland household group the model produce
only enough cabbage for household consumption. These results are supported by comparative data of model solution and actual gross revenues from all crop sales, as a percentage of total revenue (data not presented here). These data indicate that while coffee sales in actuality make up the majority of gross revenue in all household groups, in percent terms, gross revenue from coffee sales were smaller than those from foodcrop sales in model solutions.

Table 7. Percent of Crop Production Sold, Model Results and
Observed Values, all Households
Household Group
Small Lowland Large Lowland
Crop Actual Model Actual Model
--- ---------------Percent--------Banana 22.2 22.8 15.6 18.1
Beans 0.0 9.2 12.4 47.8
Cabbage NAa NA NA NA
Cocoyam 0.0 0.5 0.4 2.3
Corn 3.6 89.5 29.6 49.3
Manioc 0.0 0.0 62.9 62.2
Plantain 38.9 39.1 8.3 12.0
Potato 0.0 49.0 9.2 68.7
Yam 2.4 3.1 0.0 3.3
Household Group
Small Highland Large Highland
Actual Model Actual Model
--- ---------------Percent--------Banana 0.0 0.0 2.8 12.6
Beans 0.0 77.9 3.7 45.4
Cabbage 70.7 0.0 90.7 85.3
Cocoyam 0.0 0.3 5.1 23.2
Corn 0.0 28.2 31.7 83.2
Manioc 0.0 0.0 0.0 2.0
Plantain 0.0 0.0 3.0 21.2
Potato 27.4 66.6 27.5 86.3
Red cabbage NA NA 100.0 100.0
Yam 0.0 0.0 0.0 39.2
aNot applicable.

Table 8 shows a comparison between the objective function value (earnings) result and estimated average annual family expenses. Of the four household groups, only in the case of the large highland group was the net returns from farming expressed
in the objective function, enough to meet the estimated living expenses of the household. Although it is likely that some income was unreported, the consistency and magnitudes of these outcomes
are consistent with those of Fouda (1988) in her study of farm households in the Western Province. She found that the average household"s net revenue of 1,106,014 CFA did not cover family living expenses (1,222,447 CFA). The implication being that the available resource base, particularly land, is not large enough to support the households. In the case of the large highland
households, it would appear that the level of land holding is sufficiently large to provide for the household's need. However, in light of the relatively large amount of land actually devoted
to fallow by this household group, it is expected that their income will not match the income level in the objective function
of the model solution. In fact, as shown in Table 8 their reported gross revenue of 1,133,201 CFA does not meet their estimated
annual living expenses (1,450,523 CFA).

Table 8. Comparison of Objective Function Values and Gross Annual
Revenues and Annual Living Expenses
Objective Average Annual Average Annual Function Gross Revenues Living Expenses Value
Household Group (Earnings)
- - - - CFA - - - -
Small Lowland 220,299 73,182 414,915
Large Lowland 245,969 94,395 1,231,250
Small Highland 151,M 111,171 219,375
Large Highland 2,154,097 1,133,201 1,450,523
Crop Land Production Mix and Earnings Potential
Under specifications of Model 2, complete flexibility of land use is permitted, as well as purchase of food for household
consumption. Run 1 of Model 2 was used to examine the long run optimality in land allocation to alternative crop mix. Table 9 shows the crop land allocation results as predicted by the model.
Of some note is the finding that during the first season, the only period when non-irrigated cultivation of cabbage is possible, no
cabbage crop was planted by either of the highland household groups. This result contrasts strongly with the tendency of most highland farmers to allocate land to cabbage during this season, as indicated earlier in Table 7. During the survey, farmers
complained of lower prices received for cabbage. The model results indicate that at least for the first season, prices are such that cabbage production might be unattractive. The location of Douala
and Yaounde (the two largest and rapidly growing urban areas), adjacent to the West Province, might be creating transitional trends in the demand for cabbage. There is evidence that the

supply of cabbage has increased significantly as farmers seek alternatives to the perceived low returns to coffee. Nevertheless, the rate of growth in this demand might not be significant. All the results indicate, however, that of the various food crops, corn and to some extent potatoes, generate large positive returns. Under Model 2, the solution for both of the large household groups included exclusively, the first season corn cropping mix.
Table 9. Optimal Land Allocation, Model 2, all Households Household Group
Highland Lowland
Cropmix Large Small Large Small
- - - - Hectares-------Coffee 0.00 0.75 0.00 0.58
Season 1
Foodcrops intercropped
with w/coffee 0.00 0.75 0.00 0.58
Cornmix 6.19 0.00 2.63 0.20
Bean and potato mix 0.00 0.00 0.00 0.00
Cabbage 0.00 0.00 NA NAa
Season 2
Bean and potato mix 1.54 0.00 0.78 0.07
Cabbage 0.25 0.00 NA NA
Red cabbage 0.07 NA NA NA
aNot applicable.
The optimality of land allocation to coffee production shows mixed results from Model 2. Large household groups chose not to plant coffee, even though there are indications that the rate of yield is higher among the large highland farms. This tendency can be understood in terms of the weighted average implied values for 43

coffee land and foodcrop land shown earlier in Table 4. Recall that in that table, the weighted average value of foodcrop land is greater than that of coffee land in both of the large household groups. Also, the value of both cropland types are virtually the same for the two lowland household groups, suggesting that there might be a high degree of sensitivity to coffee prices. As such, reduction in the price of coffee could quickly shift the basis of the solution to reflect greater allocation of land to foodcrop production.
Results of run 2 of Model 2 show the effects of changes in coffee price on optimal land allocation. Table 10 shows the results for the small and large highland household groups and Table 11 shows the results for the small and large lowland groups. Results are shown for only those prices at which a change in the basis of the model occur. Since at the existing prices, neither of the large household models chose to plant coffee, the results were determined for rising coffee prices. In the large highland group, a moderate 16 per cent increase in coffee price (550 CFA/kg)
resulted in the model allocating land to coffee. However, a 68 per cent increase in coffee prices (800 CFA/kg) was required to induce the model to abandon food entirely and produce only coffee. This is further indication of the attractiveness of food crop in the production system. It has been estimated that the implicit tax on Arabica coffee production was about 71 percent at the time of the study. The 68 percentage required increase in coffee prices is thus less than the implicit tax, so even if farmers received the

Table 10. Optimal Land Allocation Under Changing Coffee Prices,
Highland Household Groups
Crop Mix Price of Coffee
Small Highland Household Group CFA/Kg
325 375 400 475a
--------- Hectares-------Coffee 0.00 0.34 0.61 0.75
First Season
with coffee 0.00 0.34 0.61 0.75
Corn mix 0.75 0.41 0.15 0.00
Cabbage 0.00 0.00 0.00 0.00
Second Season
Beans & potatoes 0.23 0.12 0.05 0.00
Cabbage 0.00 0.00 0.00 0.00
Large Highland Household Group CFA/Kg
475 550 600 625 650 800
- - - -Hectares-------Coffee 0.0 2.9 3.0 3.1 4.3 6.2
First Season
with coffee 0.0 2.9 3.0 3.1 4.3 0.0
Corn mix 6.2 3.3 3.2 3.1 1.9 0.0
Cabbage 0.0 0.0 0.0 0.0 0.0 0.0
Second Season
Beans & potatoes 1.5 0.8 0.9 0.9 0.5 0.0
Cabbage 0.3 0.1 0.0 0.0 0.0 0.0
Red cabbage 0.1 0.1 0.1 0.1 0.1 0.0
aPrice of coffee at time of study.

world market price for coffee, foodcrop production would still be an attractive alternative.
In the large lowland household group the optimal model solution also did not include cof fee planting at the existing price level. It took an 89 per cent increase in the price of coffee (900 CFA/kg) to induce the model solution to allocate land to coffee production (Table 11). However, even at that price level, the model solution included some corn cropping mix. Since
food purchasing is permitted in the solution, the corn cropping mix is most likely for sale, rather than household consumption. In the case of both small household groups the optimal model solution included coffee production at the existing price level (475
CFA/kg). However, in the small highland household group, a 21 per cent decrease in coffee price to 375 CFA/kg, leads to a decrease of more than one-half of the land allocated to coffee production, and 31 per cent decrease leads to a complete shift away from
coffee. In the case of the small lowland household group, a 16 per
cent decrease in coffee price to 400 CFA/kg leads to a decrease in coffee land allocation by more than one-half. The general results of the optimal model solution indicates that under even moderate
decreases in coffee prices, the tendency is to shift away from coffee production to foodcrops. Thus, at prices existing at the time of the study, coffee production appears to be at best marginally attractive to some of the household groups.

Table 11. Optimal Land Allocation Under Changing Coffee Prices,
Lowland Household Groups
Crop Mix Price of Coffee
Small Lowland Household Group CFA/Kg
100 250 400 450 500 550 800
---------Hectares- -------Coffee 0.0 0.07 0.23 0.53 0.58 0.68 0.78
First Season
with coffee 0.07 0.07 0.23 0.53 0.58 0.68 0.78
Corn mix 0.71 0.71 0.55 0.25 0.20 0.09 0.00
Second Season
Beans &
potatoes 0.26 0.26 0.20 0.09 0.07 0.04 0.00
Large Lowland Household Group CFA/Kg
475a 900 1000
-------- Hectares------Coffee 0.00 2.21 2.63
First Season
with coffee 0.00 2.21 2.63
Corn mix 2.63 0.41 0.00
Cabbage 0.00 0.00 0.00
Second Season
Beans & potatoes 0.79 0.12 0.00
price of coffee received by farmer at time of study.

Model 2 also generated an optimal solution for cropping mix, over a range of fertilizer prices (run 3). The results indicate some resistance in optimal land allocation to changes in the price of fertilizer. In the small highland and large lowland groups, an increase in fertilizer price by more than 100 per cent effected no change in the model's basis. In the large highland household
group, the rise in fertilizer price led to a shift away from
cabbage in the second season, but this reallocation represented a small portion of total land. In the small lowland group, a rise in fertilizer price7 to 4500 CFA/sack for both coffee and foodcrops led to a significant shift away from coffee production and towards the corn cropping mix. The general implication is that, assuming that the removal of subsidies increases the price of fertilizer to the farmer, the implementation of such a policy would have minimal differential effect on cropping patterns. The exception might be cabbage production, which like other non-traditional cash crops, heavy fertilizer use would put them at a disadvantage when fertilizer prices rise. The small lowland household group model gave results suggesting that an increase in the price of fertilizer could have an additional negative impact on coffee production. The model indicate that in the other three household groups, the effect of capital is neutral with respect to
fertilizer price increases. Thus, in terms of the small lowland households, fertilizer price increase in combination with capital
7Fertilizer prices at the time of the study were measured at an average of 3500 CFA/sack when purchased on the open market, and 2500 CFA/sack when obtained from the coffee cooperative.

constraint, could make the outlook for coffee production even less attractive.
The farmers of the Bafou study area can be categorized into
four representative farm household groups: (1) small lowland households, (2) large lowland households, (3) small highland households and (4) large highland households. The households can
be further categorized into two groups with respect to the ef f ects of land constraints. Group 1 consists of both lowland household groups and the small highland household group. The implied values of land to the households in Group 1, show that smaller land
holders generally face higher land values than do larger land holders. That is, other things being equal, additional land is of more value to those with less to begin with. As such, the effects
of land constraints on income generation are more severe for those farmers with smaller holdings. However, all members of Group 1 share similar levels of productivity and efficiency, in terms of
rates of yield of labour and other factor use. Also, results indicate that for all members of this group, returns from
agricultural production are not enough to meet household living expenses. The conclusion is that for this group of householders, their resource base is too small to support the household size.
Group 2, the large highland farmers, display greater resource productivity and allocative efficiency than Group I farmers,
resulting in slightly higher implied land values. Thus, due to 49

simply technical differences, land scarcity has a greater impact
on the income generation of Group 2 households, inspite of the relatively large size of their holdings. Further, Group 2 farmers
have enough land to be able to generate enough income to meet household living expenses, and they do not allocate all of their land to productive activities. All households have a strong need for capital, and the lack of access to capital might be as large factor in their behaviour as are land constraints. Under existing levels of farm technology, most households, even most large
households, do not have enough land to generate sufficient income to meet household needs. As population increase, so will population densities. Unless off-farm employment opportunities become available, holding sizes will become even less adequate to sustain households.
In terms of optimal crop land allocation, the results indicate that the future for the continued widespread cultivation
of Arabica coffee in the West Province is at risk. At existing prices, two of the four household optimization models selected coffee over foodcrop production. However, each of the four
household group models showed sensitivity to coffee prices. As such, even moderate reduction in coffee prices caused the profit
maximization model solution to indicate a near complete shift from coffee to foodcrop production. Any shift away from coffee
production may not necessarily take the form of clear-cutting coffee plants. Farmers were observed to be increasing the spacing

between plants when replanting their coffee, and thereby
increasing the area available f or alley-cropping with cabbage, beans and other foodcrops in association with the coffee.
Cabbage (including red cabbage) is the most widely planted non-traditional cash crop. Recent years have witnessed its increasing cultivation, as farmers search for alternatives to
cof fee production. Results indicate, however, that cabbage as a non-traditional (or market garden) cash crop might not be as financially attractive alternative as is believed. of the two household groups who cultivated cabbage, only one household group
selected the crop in the optimal allocation solution, and the selection occurred only during the second season. During the second (dry) season cabbage must be irrigated, and the resulting
small supply assures higher prices. In the aggregate, however, the demand side of the market is small and prices have declined in recent years as supply increased. The implied land value results of the model indicate that in general, a higher land value exists f or land allocated to f oodcrops than f or cof fee production. It was f ound that even on land allocated to cof fee, much of the land value comes from foodcrops planted in association with coffee. In
the optimal land allocation model, two of the four household
groups chose to plant only the corn crop mix during the first planting season, consisting of corn, beans and potatoes. In one of these household groups (the large highland farmers), the corn crop mix was selected inspite of the highest coffee yield rate. In this household group,nearly all land is reallocated to the corn, bean 51

and potatoes crop mix when there is a moderate decline in coffee prices.

Barnett,Douglas, Brian Blake and Bruce McCarl, 1982. "Goal
Programming via Multidimensional Scaling Applied to Senegalese Subsistence Farms." American Journal of Agricultural
Economics, 64(4): 720-727.
Bergeret, Pascal, Guy Ducret, Isabell Grangeret, M.N. Roux and J.L.
Schafer, 1988. "Le Systeme Agraire de Bafou (Ouest Cameroun) Chefferie du Pays Bamilike." Dschang: University Center of
Dschang, Operation Bafou Report.
DeBoer, John A. and Satish Chandra, 1978. "A Model of Crop
Selection in Semi-subsistence Agriculture and an Application to Mixed Agriculture in Fiji." American Journal of
Agricultural Economics, 60(3): 436-444.
Dillon, John L. and Jock R. Anderson, 1971. "Allocative Efficiency,
Traditional Agriculture, and Risk." American Journal of
Agricultural Economics, 53(1): 26-32.
Fouda, Therese M., 1988. "Assessing Small Farmers' Credit Needs: A
Case Study of the West Province of Cameroon." Gainesville: M.Sc. thesis, Food and Resource Economics Department,
University of Florida.
Hayer, Judith, 1971. "A Linear Programming Analysis of Constraints
on Peasant Farms in Kenya." Food Research Institute Studies,
10(1): 55-67.
Hooper, W. David, 1965. "Allocative Efficiency in Traditional
Indian Agriculture." Journal of Farm Economics, 47(5): 221239.
Langham, Max R., 1968. "A Dynamic Linear Programming Model for
Development Planning." In: W.W. McPherson (Editor), Economic Development in Tropical Agriculture. Gainesville: University
of Florida Press.
Mellor, John, 1969. "The Subsistence Farmer in Traditional
Economies." In: Clifton R. Wharton (Editor), Subsistence Agriculture and Economic Development. Chicago: Aldine
Publishing Company.
Ministry of the Plan and Regional Development, 1986.Sixth FiveYear Economic, Social and Cultural Development Plan, 19861991. Yaounde: Republic of Cameroon.
Nji, Ajaga, 1982. "Socioeconomic Factors Related to the Adoption of
Agricultural Technology in the Western Province of Cameroon."
Dschang: University Center of Dschang, Mimeograph Report.

Wolgin, J.M., 1975. "Resource Allocation and Risk: A Case Study of
Smallholder Agriculture in Kenya." American Journal of
Agricultural Economics, 57(4): 622-630.
World Bank, 1987. World Development Report. New York: Oxford
University Press.
Yotopoulos, P.A., 1967. Allocative Efficiency in Economic
Development. Athens: Center of Planning and Economic Research.