Analyzing the technical performance of cotton farmers in developing countries: The case of Burkina Faso

(1)

Uppsala university

Department of Economics Bachelor thesis

Spring 2008

Analyzing the technical performance of cotton farmers in developing countries:

The case of Burkina Faso

Supervisor: Prof. Yves Surry Author: Mohammad Sepahvand and Roujman Shahbazian

(2)

Abstract

In this study the technical efficiency of 142 cotton farmers in Burkina Faso has been estimated. The aim has been to detect the technical efficiency and to explain it by socioeconomic factors. For this, a model has been created where the technical efficiency is regressed against certain socioeconomic variables.

The database for this study, which has been collected from the main agricultural zones in Burkina Faso - West, North and East - has been provided by INERA. The database has been complemented through field studies to cotton producing villages and systemized by the authors. The software package Frontier has been used to conduct the necessary statistical regressions for a sample of 142 cotton farmers. The technical efficiency scores have been obtained by using a stochastic production function approach in a two stage estimation process.

The result from the Frontier estimation outputs shows that there is not much room for improvement of technical efficiency. Thereby an increase in technical efficiency of Burkina Faso’s cotton producers is not the way to go in order to increase the cotton production.

However, from the chosen socioeconomic variables, it can be detected that the producers that have received agricultural training are more productive than those who have not. Also, specialization on cotton production yields higher productivity. However, due to the fact that the cotton farmers can already be considered efficient given the present state of technology and inputs, domestic efforts are not enough rather the efforts has to be put on reducing the international trade barriers in order to increase the development of Burkina Faso.

(3)

This work was performed in Burkina Faso during the period of June and July 2007 as a Minor Field Study (MFS). The authors would like to thank SIDA and Uppsala University for the opportunity that was given to them in the form of two MFS- scholarships, which made it possible to conduct this study. Furthermore, the headquarters of INERA in Bobo-Dioulasso and its personnel have been an enormous help in their assistance in the field and during the complementary study that was carried out. Many thanks go to Sigrun Helmfrid at Stockholm University for preparing us for the field trip and providing us with valuable contacts in Burkina Faso.

Our supervisor in field, Dr. Souleymane Ouedraogo has to be acknowledged as well. A special thanks goes to Prof. Yves Surry for his guidance and encouragement throughout the whole study, which without this study would not have been materialized. The authors would like to take the opportunity to express their appreciations to their families for their support during the study.

Errors and omissions are the responsibility of the authors. Opinions that may be expressed in this report belong to the authors and do not necessarily reflect those of any persons or institutions.

(6)

1. Introduction

In our intermediate economic courses, we came in contact with the devastating impact that agricultural trade distorting policies, imposed by developed countries, have on developing countries. A relevant example is Burkina Faso. In the course Agriculture policy and international trade, taken at SLU, we wrote a paper about how a free trade scenario in the international cotton market would affect Burkina Faso. Using the ATPSM-model¹, which is a global partial equilibrium trade model, the result that the world price of cotton would increase significantly after free trade was obtained. Furthermore it could be concluded that this increase in world prices would have a significant benefit on the total welfare of Burkina Faso.

This conclusion raises the question about the productivity of cotton farmers in this country. If free trade was to be introduced, but the farmers are not productive enough, they would not be able to achieve their full potential level in terms of welfare. So it is essential for both the farmers and the country to seek for the highest level of welfare if free trade would be combined with an appropriate level of productivity.

In order to determine the appropriate level of productivity the technical performance of Burkina Faso’s cotton farmers would be studied. For this purpose this paper will use a stochastic efficiency production frontier model applied to a sample of cotton producers. The reason for measuring technical efficiency (TE) is to see first and foremost if cotton farmers are using their inputs in an efficient way. If they do not use their existing technology and human capital in an efficient way, any improvement toward introducing new production methods would be more costly than simply improving the current technical efficiency (Bravo- Ureta & Evenson; 1994).

In the literature on production efficiency (Bravo-Ureta & Evenson; 1994), the major determinants of TE are socioeconomic variables such as age, education, level of literacy, experience from cotton cultivation, specialization of arable cotton land and agricultural training. The degree of socioeconomic variables could have an impact on the farmers TE.

Therefore the cotton farmers’ TE would be examined against certain socioeconomic variables in order to try to distinguish any patterns behind the expected lack of TE in the sample. The TE itself a difference between estimated output and given output in a regression for a sample

1 http://192.91.247.38/tab/ATPSMabout.asp

(7)

1.1 Objectives and method

The first objective of this study is to estimate the technical efficiency of cotton farmers in Burkina Faso. The second objective is to detect whether the TE scores of these farmers can be explained by socioeconomic factors. With the help of these results it would be possible to point out ineffective areas affecting the production of cotton in Burkina Faso.

To be able to estimate the technical efficiency of cotton farmers in Burkina Faso, INERA’s microeconomic data containing agricultural inputs and socioeconomic variables of 142 farmers that have been collected all across the country of Burkina Faso have been used. The software package Frontier is used to detect the TE scores of the farmers by using a stochastic production frontier approach in a two stage estimation process where the TE-model is obtained.

2. Burkina Faso’s economy: An overview

Burkina Faso is located in West Africa and its neighboring countries are Benin, Ghana, Mali, Ivory Coast, Niger and Togo. With 12 million inhabitants and a size of about half Sweden, it is a representative country for the region.

2.1 Macroeconomic setting

Burkina Faso gained its full independence in 1960 after being a colony to France. Not unlike many of its neighbors, several military “coup d’état” followed the independence. But after the current President Blaise Compaoré came to power in 1987, the political situation has become more stabile. This in return has had an influence on the country’s macroeconomic performance.

(8)

Figure 1 Real GDP growth

2 0 0 5 2 0 0 0

1 9 9 5 1 9 9 0

1 9 8 5 1 5

1 0

5

0

- 5

Y e a r

Percentage growth

Source IMF²

Figure 1 gives a hint about the economic situation that Burkina Faso has experienced during the last 20 years. The average GDP growth rate in period 1986 to 1993 was -0.03 in comparison to 4,09 in the period 1994-2006. Economic development is concentrated in the capital Ouagadougou and there are large regional inequalities in terms of infrastructure and income between urban and rural areas.³

Burkina Faso is a developing country characterized by low income which gives it the second last place amongst United Nations Development Programs (UNDP) human development index in the most recent published index.⁴ The country is struggling with many difficult issues, such as illiteracy, healthcare problems, environmental and democracy issues. For example only 21.8 percent of the population can read and write. In order for Burkina Faso to tackle these problems it is vital that there are sufficient labour opportunities for the population. The agriculture sector which employs over 80 percent of the population, is not profitable enough to sustain an acceptable economic well being. One of Burkina Faso’s largest export commodities and main income sources is cotton which accounts for over one- third of the country’s exports. (New Agriculturist)⁵

2 http://www.imf.org/external/pubs/ft/weo/2008/01/weodata/weoselgr.aspx (2008-06-09)

3 Utrikespolitiska institutet

4 http://hdr.undp.org/en/statistics/ (2008-06-09)

5 New Agriculturist 2007, se website: http://www.new-agri.co.uk/07/01/countryp.php (2007-12-27)

(9)

During the last 10 years the macroeconomic performance of Burkina Faso has been somewhat stronger. Regardless of this economic performance, the country’s is highly vulnerable due to the lack of economic diversification. Structural and institutional reforms are needed to diversify the economy. The agricultural sector is also characterized by low productivity.

Therefore in order to improve the economy it is still important to increase the performance of this sector.⁶

2.2 Agricultural sector

Burkina Faso is a relatively flat land-locked savannah country which, despite significant mineral deposits, remains one of the poorest countries in the world. As table 1 indicates the agricultural sector plays a significant role in Burkina Faso’s economy. The share of the agriculture sector in the GDP is almost 30 percent as compared to two percent in the EU. But viewed from a regional perspective, the neighboring countries of Burkina Faso have approximately the same amount of GDP devoted to agriculture.

Table 1 Agriculture in percent of GDP (current prices), 2007⁷

Benin 33.2

Burkina Faso 29.4

Chad 22.2 EU 2 Greece 3.2 Mali 45 Spain 3.8 US 0.9 World 4

Source: EUROSTAT and CIA:s statistical databases

6 http://www.imf.org/external/pubs/ft/survey/so/2008/car022508b.htm(2008-06-09)

7 The share of agriculture in GDP is derived as the gross value added (at basic prices) in percentage of GDP

(10)

The most important agricultural products in Burkina Faso are cotton, sugarcane, peanuts, sesame, sorghum, millet, corn and livestock. Livestock has historically been a major export commodity but has lost its significance. Sugarcane has recently been introduced on a large scale and has become a significant cash crop, while sorghum, millet and corn do not generate any noteworthy income. It is the “white gold”, meaning cotton, that constitutes over a third of Burkina Faso’s exports and the growth of the cotton production during the last decade has been huge. (New Agriculturist⁸) Figure 2 illustrates the increase of production of cottonseed during the period 1992-2006, with the base year as 1993. During the period 1992-2006 the production has increased by almost 600 percent.

Figure 2 Index over production of cottonseed, 1992=100

2 0 0 6 2 0 0 4

2 0 0 2 2 0 0 0

1 9 9 8 1 9 9 6

1 9 9 4 1 9 9 2

7 0 0 6 0 0 5 0 0 4 0 0 3 0 0 2 0 0 1 0 0

Y e a r

Index

Source: FAOSTAT⁹, 2008

2.3 Cotton sector

The agricultural crop that is most cultivated and economically important in Burkina Faso is cotton. But this has not been the case historically. When, Upper Volta (Burkina Faso’s previous name) was colonized by the French in the beginning of the 20^th century, cotton

8 http://www.new-ag.info/07/01/countryp.php (2007-12-27)

9 http://faostat.fao.org/site/567/DesktopDefault.aspx?PageID=567 (2008-06-09)

(11)

production fulfilled two aims: to provide clothing for daily and ceremonial usage and as a means of payment in exchange with neighbouring regions (Hårsmar, 2004:165).

During the beginning of the twentieth century, cotton deliveries to the European textile industry were going through an uncertain period due to trade speculations. In light of this the French government created a new agency that would promote the cotton production in its old as well as new colonies, which also included Burkina Faso. The newly created agency was named I’Association Cotonnière Coloniale (ACC). After the Second World War cotton production in Burkina Faso increased significantly. One of the reasons behind the increased production was the technical development that yields new varieties of cotton with longer fibres than the traditional ones, which was especially attractive for international trade.

(Schwartz, 2003:2-3)

The potential of exporting cotton and selling it on the international marked was the cause behind the creation of SOFITEX in 1979, which was part owned by the Burkina Faso government and part by the French Association for Textile Promotion (CFDT). The aim of SOFITEX was and still is to strengthen and expand cotton as the prime agricultural export product of the country. They try to achieve this aim through offering credits and technical advice to the farmers, distributing inputs of different sorts and acting as market channels to the producers. In order to stimulate and promote the production of cotton amongst farmers, SOFITEX has established a fund that aims to offering the farmers stable prices for their cotton (Hårsmar, 2004:168).

INERA is a research institution for the agricultural sector, which deals amongst other things with cotton production. The institution focuses on socioeconomic factors and technical efficiency of the farmers in Burkina Faso. INERA’s main aim is to increase the quality and quantity of cotton production in Burkina Faso. In other words; to maintain sustainable cotton production for the Burkinabe cotton producers. To be able to conduct this task INERA disposes and processes its own collected microeconomic statistical data concerning agricultural production in the different regions of the country.

(12)

2.4 Current situation

As discussed earlier the macroeconomic performance of Burkina Faso has been encouraging during the last decade. Regardless of this growth, the country’s economy is highly vulnerable due to the lack of economic diversification. Structural and institutional reforms are needed to diversify the economy.¹⁰

Because of the lack of diversification the economy is highly vulnerable to world price changes, especially when it comes to the price of cotton, the main export commodity.

According to the World Bank¹¹ the price of cotton fell by about 15 percent in 2005, in the same time oil prices rose by 43 percent, making the terms of trade impact for the country very destructive.

In 2003 together with three other countries (Benin, Chad, and Mali) Burkina Faso raised The Cotton Initiative in the General Council of the WTO¹², which was developed and submitted in order to eliminate the developed countries’ domestic support for cotton and get a fair compensation for their trade distorting policy. Unfortunately their demand was not fully heard; instead a compromise was reached which was far from the original proposal and therefore not favourable enough for the developing cotton producing countries to enforce slightly.

3. Theoretical framework

The objective of this study is to provide technical efficiency estimates for a sample of cotton farmers. The way to reach this aim is by using the theoretical framework of maximum likelihood estimations for a stochastic production function. This gives rise to mean and individual technical efficiency scores when put in the two stage estimations process that is provided by the software package Frontier.

10 Utrikespolitiska institutet

11 http://siteresources.worldbank.org/IDA/Resources/IDA-BurkinaFaso.pdf

12 http://www.wto.org/english/tratop_e/agric_e/cotton_subcommittee_e.htm#top (2008-02-17)

(13)

3.1 Theory

The framework to this paper lies in the fact of obtaining optimal estimates of the technical efficiency for the cotton farmers in Burkina Faso. This technical efficiency can be obtained through different ways of measurement. One way that is considered the best in practice and used by the estimation program Frontier (Coelli,1996;3), is to take consideration to the farmers observed output quantities and the output quantities given the production technology available with the given inputs. To be able to use this approach, one has to estimate the given output quantities, which has historically been obtained by setting up a production function.

The production function is not only practical from the fact that it gives an estimation to a regression containing all the inputs but the production function approach is also used by the Frontier program in obtaining TE-estimations for the agricultural sector. Therefore, this approach has been widely used in measuring technical efficiency (Seyoum et al,1998;343).

That is why, in this paper the stochastic frontier production function approach will be used.

Another reason, for using this approach has to do with the fact that it can be implemented as well on cross-section and panel data (Battese and Coelli, 1992; 149). There are three main sub-groups of frontier production functions: the deterministic, stochastic and panel-data frontier models, which are estimated using a maximum likelihood procedure.

The dependent variable in the frontier models is expressed as a product of vectors of a function of inputs and a random stochastic effect (Battese, 1992; 188). The main difference between the deterministic and stochastic model lays in the fact that the technical efficiency (TE) obtained by a stochastic frontier function would be higher than the one given by the deterministic function because the stochastic frontier function does not have any upper limit for outputs. Therefore, the optimal outputs given by the stochastic function can exceed the deterministic function (Battese, 1992; 190f). However, the main difference between a deterministic/stochastic frontier production function and a panel data frontier model has to do with the sample that is used. In the former case the sample consists of cross-sectional data in comparison to the latter where time-series or panel data is used (Battese, 1992; 188ff).

In the stochastic frontier production function the dependent variable and inputs in the production function is calculated through Maximum likelihood, which provides technically

(14)

optimal values given the sample (Coelli,1996; 4)¹³. The production function error term in the stochastic production function consists of a random effect (the same effect used in a simple linear regression) and an inefficiency effect. By using this inefficiency effect, one will obtain the technical efficiency effect from Frontier which is in fact the effect from the usage of inputs to obtain optimal output compared to the actual output obtained from the production process (Battese, 1992; 191). The stochastic production function could usually be expressed in the following way:

) 1 ( ...

, 2 , 1 ),

exp(

)

;

(X V U i n farmer

f

Y= _i β × _i − _i = ^nt

where is the production of the production unit (such as a farm), f is a suitable production function (for example a Cobb-Douglas), represents input quantities of the production unit, β is a vector of estimated parameters, is independently and identically normally distributed random errors with mean 0 and a constant variance, independent of .

is a non negative normal distributed random variable, associated with technical inefficiency in the production.

Yi i^th

Xi i^th

Ui

Vi

Ui

The random error term in equation 1 above, consists of a noise and inefficiency component.

The inefficiency component, , is given in Frontiers simultaneous two stage estimation procedure. This inefficiency term then becomes a dependent variable in a linear technical efficiency model, were the exp(- ) = TE is regressed against relevant explanatory variables that is considered to have an effect on the TE of the production unit (i.e the farm). These explanatory variables have a socioeconomic characteristic. The TE-model can be explained in the following way (Battese, 1992; 192):

Ui

U_i

) 2 ( ,

) exp(

)}

exp(

)

; ( { / )}

exp(

)

; ( {

/ _i^* _i _i _i _i _i _i

i

i Y Y f X V U f X V U

TE = = β × − β × = −

13 For a brief description of how FRONTIER conducts these stochastic frontier production function estimates see Coelli, 1996; 11f.

(15)

Which in turn is regressed against socioeconomic regressors (Tzouvelekas et al, 2002; 5):

) 3 ( ...

, 2 , 1 , )

;

(Z i n farmer

g

TE_i = _i δ +ω_i = ^th

where is the technical efficiency of the ith production unit (such as the farmer), g is a exponential function, represents unit specific socioeconomic characteristics that is assumed to affect the technical efficiency of the ith unit,

TEi

Zi

δ is a vector of estimated parameters, ω_i is independently and identically “truncated” normal distributed random errors with mean 0 and the constant variance. The reason that the regressors explain the TE instead of inefficiency is because the error term in the production function, , which is the dependent variable, is expressed as exp(- ), which makes all the negative regressors in the estimation positive by exp((-) (-)) =exp(+).

Ui

Ui 14

The difference between the estimated outputs¹⁵ from (equation 1) and the observed outputs (meaning from the sample) would give us the rate of TE, which is between 0 and 1. High rate implies efficiency, low rate implies inefficiency. However there are two main definitions of technical efficiency that should be in mind and defined when conducting frontier production function estimates, the Debreu and the Shepard-type of TE. The Debreu type of TE is the output-oriented measure of TE and is the one described above. The Shepard-type is the input- oriented measure of TE and is defined as the difference between observable and optimal input (Tzouvelekas et al, 2002; 4).

14 For the derivation of the TE from the stochastic frontier model see Battese & Coelli (1992).

15 This estimated output is in fact the optimal output derived from the production function, i.e the optimal output obtain by using the input in a optimum way (Battese, p185)

(16)

The underlying understanding of the Debreu- type of TE could be described as in Figure 3 below:

Figure 3: describing the TE for farmers in a production function frontier, the distance A-B indicates the technical inefficiency of the production unit (based on Battese figure, 1992; 187).

Figure 3 illustrates a diminishing frontier production function¹⁶ in the x-y space. The Vertical axis shows the production of each production unit (Y) and the horizontal axis we inputs (X)¹⁷.

The maximum output that the production unit (farmer) can obtain with the inputs available is the output along the production frontier. Therefore, all the points to the right/below this frontier is considered as not optimal, meaning that the production unit is not maximizing its usage of inputs to produce output. This is an indication for inefficiency. So, the points below the production frontier in Figure 3 indicate inefficient use of resources. However, there are different degrees of inefficiency depending on how far you are from the frontier function. For example the inputs used at point A give raise to output y, i.e point (x,y). But with the inputs used at A the production units can produce the optimal amount of output y* at point B, e.g (x,y*). Therefore, point A indicates this unit’s technical efficiency for the inputs available, and the distance between point A and B indicate the technical inefficiency¹⁸.

16 For simplicity we have showed a simple bowed production function, the shape of the stochastic frontier production function may vary from the production function described in Figure 3.

17 X is considered a vector of inputs used in the production function.

18 The reason that we use the term Technical efficiency or inefficiency, instead of efficiency or inefficiency is because of the fact that we are talking about the output that the farmer produce and can optimally produce with the Inputs or Technology available, without considering the level of input prices.

(17)

The stochastic production function described above is a straightforward way to estimate the TE of cotton growers in a country as Burkina Faso. Because cotton cultivation is the primary agricultural activity that has a direct income effect for the farmers in Burkina Faso, meaning that cotton is considered as a cash crop, a great amount of land has been devoted to cotton cultivation.

This pattern has in turn naturally been followed by investments in land, intermediate consumption¹⁹, labour but also socioeconomic variables such as time put on learning the cultivation process. The usage of the production function is straightforward due to the fact that inputs such as capital and labour are included in the industry. And, the stochastic production function frontier is a modified form of the production frontier that has been used and accepted for the measurement of TE.

3.2 Literature review

There is no direct study related to the measurement of TE in Burkina Faso for cotton farmers.

However, the theoretical framework of this paper has been based on TE approaches implemented by: Coelli (1996), Seyoum et al (1998), Battese and Coelli (1992), Battese (1992) and Tzouvelekas et al (2002).

Empirical work on TE for cotton producers in Burkina Faso is quite limited. Therefore due to the lack of literature on TE in Burkina Faso, it is quite difficult to compare the results of this study with previous research. However there have been studies done on TE of cotton producers in developed countries. A brief overview is provided below.

In a study about the cotton producers in Greece Tzouvelekas et al (2002) show that cotton farmers have on average TE scores of nearly 80 percent. Factors that affected the technical efficiency for the Greek cotton growers were socioeconomic variables such as age and education but also how big share of their land that was put into cotton cultivation, i.e.

specialization in cotton. In a study about the cotton producers in U.S Chakraborty et al (2002) the irrigation system played a role in comparing TE among the Texas cotton farmers. But

19 Such as herbicides and fertilizers, but also the cost for equipments such as tractors and animals but also herbicide machines to be used in the cultivation process.

(18)

unfortunately in Burkina Faso the agricultural sector is not as developed as in US, therefore the irrigation system is not as wide spread.

4. Data description

Our partner institution in Burkina Faso (INERA) suggested that, instead of collecting our own sample we could conduct studies with the INERA microeconomic sample and do complementary studies in villages of choice when needed. However, due to the fact that the sample was made up of raw information, systemizations and decoding was necessary to conduct appropriate interpretation and analysis. As a result the variables in this study are limited to the preferences of INERA, due to the fact that they constructed the questionnaire.

The data was cross sectional and covers the period 2005-06.

The database in INERA’s survey divided Burkina Faso into three different study areas: West, Central and East. The sample consists of 142, observations of which 46 belong to West, 36 to Central and 60 to East. Overall, a total of 20 villages had been chosen for the survey. The survey procedure was conducted according to a random sample process, where farms and villages were chosen in a random fashion. However, consideration has been taken due to the availability of the terrain during the sampling process²⁰.

20 For the collectors of the sample it was imperative that the terrain leading to the village was accessible. The infrastructure of Burkina Faso being underdeveloped as it is, some villages could not be considered in the survey.

(19)

Map 1

The south western part of Burkina Faso, produces most of the cotton output in the country (Hårsmar, 2004:166). Thereby one should expect that more observations from that region should be represented in the sample. However, this is not the case and one explanation could be related to the inaccessibility. Because, the south west part of the region is characterized by heavy rain falls, which during its peak puts a restriction on the poor infrastructure in Burkina Faso.

The database divides the sample into four subgroups according to their level of mechanization: “manual”, small and big “attelage” and motorized.²¹ The group motorized consisted only of ten observations, therefore the two subgroups motorized and big “attelage”

were combined into one. From now on those subgroups would be referred to as small, medium and large.

21 Observations in the group “manual” use only manual labour in the cultivation. Small “attelage” are those observations who use one or two animals in the cultivation and big “attelage” those who uses more than two animals. Motorized are those who have access to some motorized machinery in the cultivation process.

(20)

4.1 Variable description and model specification

The explanatory variables used in this paper are the same that are used in a Cobb-Douglas production function and that are significant in the cotton cultivation process of Burkina Faso.

The socioeconomic variables are created after conducting a literature review for the acceptable variables used in cotton cultivation and receiving advice from INERA about the most common socioeconomic variables in Burkina Faso’s cotton cultivation.

The dependent variable Production (Y) is defined as the total annual production for each cotton farmer and is measured in kilograms. The explanatory variables consist of Capital (C), Adult Labour ( ), Intermediate Consumption ( ) and Land ( ). The relation between the explanatory variables and production is supposed to be positive. The Capital variable is a value variable that is described in Appendix 3 under the section “Capital formation”.

AL IC L

= th

− +

× +

=β β β β β

The Labour variable consists of temporary and permanent labour for each farmer, and no distinction between paid or unpaid labour has been made. This is due to the fact that the variable is constructed according to the number of days that has been invested in each step of the cotton cultivation process. The total number of days has been combined to give an estimate to the total amount of labour that has been used by each household in the cotton cultivation. The process of determining the Labour variable is described in Appendix 3 under the section “Labour formation”.

As far as Intermediate consumption is concerned, it has been created in monetary terms (CFA²²). This variable consists of the total monetary value of different intermediate inputs used in the cotton cultivation process, such as the total value of fertilizer and herbicides. The land variable is the total arable land devoted to cotton and is measured in hectare. The reason is due to the fact that not all cotton farmers solely produce cotton, but also devote some of their arable land to food related crops. The above mentioned variables give rise to the following stochastic production function expressed in logarithmic form²³:

) 4 ( ...

, 2 , 1 ), (

) ( )

( )

( ₂ ₃ ₄

1

0 In C In AL In IC In L V U i n farmer

Y

In _i _i _i _i _i _i _i

22 1 Euro~660 CFA

23 The expected sign of all the explanatory variables in equation (4) are positive.

(21)

The socioeconomic variables that have been used in the optimal technical inefficiency model consist of Agricultural training ( ), Age of the cotton farmer ( and ) and Specialization on the crop of cotton ( ). The relation between the explanatory variables and TE is positive for all the variables except for , due to the fact that there is a diminishing productivity over time. In Appendix 1 a description is given for the variables that have been created but not included into the optimal technical inefficiency model.

Training A.

Spec

Age Age²

Age2

The variable Agricultural training is created as a dummy variable and takes the fallowing values: 1= the cotton farmer that has received agricultural training and 0 = otherwise. The famers that where self-learned or had not received any formal agricultural training, has been included under the category 1.

The age variable describes the age of each cotton farmer. The variable is divided into four categories. The four categories are defined as follows: 2 = 1-29 years, 3 = 30-39 years, 4 = 40- 59 years, 5 = 60 or older. However, a squared of the age variable has been included as a new variable, to detect the peak age where the farmers reach optimum productivity.

The Specialization variable, explains the level of specialization in cotton by stating how much of the level of the total arable land is devoted to cotton cultivation for each cotton producer.

The above mentioned variables give rise to the following technical efficiency function:²⁴

) 5 ( ...

, 2 , 1 ,

. ₂ ₃ ² ₄

1

0 ATraining Age Age Spec i n farmer

TE_i =δ +δ × _i+δ × _i+δ × _i +δ × _i +ω_i = ^th

24 The expected sign of the explanatory variables in equation (5) should be positive for A.Training, Age and Spec. But for age² it should be negative. Equation (5) takes a exponential form.

(22)

4.2 Descriptive Statistics

In order to analyse the primary data, descriptive statistics of variables were calculated and analysed. The variables used in the stochastic frontier model for estimation of production efficiency of cotton producers in Burkina Faso and their summary statistics are given in table 2. The same table also provides descriptive statistics for variables explaining the technical efficiency score.

Table 2 Descriptive statistics over all variables

Variables Continuous variables Dummy variables

Mean Standard deviation

Median Min. Max

Percentage of farmers with dummy=1

Percentage of farmers with dummy=0 Stochastic frontier model

Production Y (kg)

4 528.6 4 768.9 2 795 596 29 400 C (CFA) 517 188 637 784 320 051 0²⁵ 3982 184

AL 236.81 216.34 175.52 0 1232

IC (CFA) 231 347 187 123 157 237 32 616 868 816

L (ha) 3.5776 2.8072 2.5 0.75 14

Inefficiency effects model

A. training 0.39437 0.60563

Age 3.2464 0.8006 3 2 5

Age² 11.176 5.0967 9 4 21

Specialisation 0.5671 0.1791 0.5371 0.1728 1

25 All values that are zero have been replaced by a small value do to the fact that or models are log linear.

(23)

The average cotton producer has an output of 4528.6 kilograms of cotton, with a standard deviation of 4768.9. This implies the existence of a substantial variability in production among the sampled cotton producers. From the summary statistic in the inefficiency effect model one can detect that almost 40 percent of the producers have received agricultural training. Furthermore, the average age of producers is between 30-39 years old. Moreover, the average degree of specialisation in cotton is 57 percent of the arable land in the sample.

Meaning that over 40 percent of the land is used for cultivation of other crops than cotton.

Descriptive statistics for the sub-samples small, medium and large can be found in Appendix 2.

4.3 Complementary sampling

A complementary sampling was done in three villages in the Western part of Burkina Faso, namely Gombilidougou, Bala and Kendidougou. There were two main reasons behind this complementary study; to evaluate the authenticity of the sample and replace the missing values in the sample.

The evaluation of the authenticity of the survey was due to some questions that had risen during the systemisation procedure. One of the questions concerned the reported amount of intermediate inputs (such as fertilizer, herbicides et cetera), there were no consistency in the usage of intermediate inputs by the cotton farmers. Usually there is a positive trend when it comes to size of arable land and usage of intermediate inputs, i.e. more land leads to a greater use of fertilizers. However, in the survey the consumption of intermediate inputs did not change significantly with the size of the land. During the complementary study we detected that the observation in the smallest group (meaning “Manual”) was recorded in decilitres and not litres. Therefore at first glance it appeared that most observations used roughly the same amount of intermediate inputs.

Some of the observations had missing values in term of no quantity of the cotton production was represented in the survey. During the complementary survey and when in contact with the person in charge of the survey, it was possible to fill in the correct amount of production.

(24)

5. Empirical Results: Whole data sample

Using the computer package Frontier version 4.1, maximum likelihood estimates of the technical inefficiency model and the parameters of the Cobb-Douglas stochastic frontier production function have been obtained. These parameter estimates, along with standard errors and t-ratios of the Maximum likelihood estimators of the Burkina Faso cotton farmers are presented in table 3. The signs of the estimated parameters in the Cobb-Douglas stochastic frontier production model are as expected. Estimated coefficients for all variables in the Cobb-Douglas model confirm the expected positive relationship between Capital, Labour, Intermediate consumption and Land in the cotton production. All of the coefficients are also significant at the one percent level, except Labour which is significant at the five percent level.

Table 3 Estimation for the whole sample

Variables Estimates t-values

Stochastic frontier model

Intercept 0.873 1.10

Ln (C) 0.075* 12.60

Ln (AL) 0.135** 2.52

Ln (IC) 0.458* 6.51

Ln (L) 0.358* 5.17

Intercept 0.610* 2.74

A. training -0.189* -7.81

Age 0.413* 5.01

Age² -0.083* -4.69

Specialisation -2.047* -10.20

Variance parameters

Sigma-squared 0.116* 10.71

Gamma 0.160* 7.12

Log-likelihood -30.78

Notes: * indicates significant at the 1 percent level, ** indicates significant at the 5 percent level

(25)

Not all of the coefficients estimated from the technical inefficiency model are as expected.

The estimated coefficient of the dummy variable agricultural training is negative which indicates the positive effect on technical efficiency. The agricultural training variable is statistically significant at the five percent level. With respect to the level of specialisation on cotton production, the negative and statistically significant coefficient at the five percent level suggests that an increase in the degree of specialisation would contribute to higher technical efficiency level for cotton producers in Burkina Faso. The variable age and Age²are not as expected.

The gamma variable for the stochastic production function in table 3, implies that the TE variables are significant in determining the variability and the level of the technical efficiency for the cotton farmers in Burkina Faso, at the one percent level.

5.1 Specification test and Efficiency results

To be able to find out whether or not the chosen model is the best model, there will be a comparison to another more unrestricted (i.e. more variables) Cobb-Douglas in its TE- modification. The model that will be compared with is the same model as the chosen Cobb- Douglas model for this paper, but with the inclusion of the explanatory variables for the TE model in “8 Appendix 1”. To check which model is the best, a likelihood ratio with a five percentage significant level has been conducted between these models. The outcome of the test is as described in Table 4 below:

(26)

Table 4 Likelihood ratio test of the stochastic production frontier and the TE-model Null Hypothesis Log-Likelihood

ratio

D.f Critical value 5 % Decision

The inefficiency model²⁶ δ⁵ = δ⁶= δ⁷= 0

The inefficiency model 2 δ⁵ = δ⁶ =0

4.530868

3.197832

3

2

7.81

5.99

Not Reject Ho

The outcome from Table 4 shows that the null hypothesis is not rejected of the extended Cobb-Douglas model and its TE model being not significant on the five percentage level. This implies that the chosen Cobb-Douglas model and its corresponding technical efficiency variables are to be preferred. However, to be sure that the chosen model is the best model a second likelihood ratio test on a five percentage level will be carried out, against the same Cobb-Douglas functional form. But, now the technical efficiency variables from “7 Appendix 1” are one less, (i.e. no Experience variable). The reason to take away this variable is linked to the notion of Multicollinearity, as Experience is very close in structure to the Age variables.

The outcome of the test is as described in Table 4 above. Even here one can see, that the null hypothesis is not rejected on the five percentage level, and therefore the chosen Cobb- Douglas model can be considered the best model.

In Table 5 the frequency distribution of technical efficiency for the whole sample is presented together with the sample size, mean value and max and min values. The computed average technical efficiency is 90.4 percent, ranging from a minimum of 61.3 percent to a maximum of 98.4 percent. Given the present stat of technology and input levels, this result suggests that farmers in the sample are producing on average at 90.5 percent of their potential.

26 From the Extended Cobb-Douglas model. The variables represented for the coefficients here are δ^{5 =}

Analphabetic, δ^{6 =}Experience of cotton cultivation and δ^{7 =}Child-Total labour ratio variable.

(27)

Table 5 Frequency Distribution Technical Efficiency

Efficiency (%) TE of whole sample

<60

60-70 9

70-80 12

80-90 16

>90 105 N 142

Mean 90.5

Median 91.1

Minimum 61.3

Maximum 94.4

5.2 Empirical results: Sub-sample groups

In order to enrich the study, sub samples from the whole sample have been estimated. There are three groups: small, medium and large. The descriptive statistics of these three groups can be found in Appendix 2.

In table 6 the standard errors, t-ratios of the Maximum likelihood estimated are presented for all the inputs and explanatory variables that are presented in equation 4 and 5 for the Burkina Faso cotton farmers. The sign of the estimated parameters in the Cobb-Douglas stochastic frontier model are not as expected for all variables. The labour variable (AL) exhibits reverse signs in the subsamples Medium and Large when it comes to its correlation with cotton production. Concerning the significant level, the Capital variable (C) is not significant for the Large subsample. This is also the case for the Land variable (L) in the subsample Small.

For the TE-model the Age variables do not show the expected diminishing marginal return on cotton productivity. The other variables, Agricultural training (A.training) show the expected signs but only Specialization is significant on the five percent level among the TE-models explanatory variables. The gamma variable for the production function model is significant on

(28)

the 1 percent level for all the subsamples. This implies that the TE-models explanatory variables are significant in determining the variability these variables have toward the cotton production in Burkina Faso. But if one would like to interpret the values of the gamma for the subsamples, the interpretation would be that 99 percent of the variation between the observed output and the frontier is caused by technical inefficiency. However for the other two subsamples the corresponding values are 8 and 4.8 percent.

Table 6 Estimation for the sub samples

Variables Small Medium Large

Stochastic frontier model

Intercept 0.0231 (0.03) 0.1386 (0.15) 2.5270 (0.97)

Ln (C) 0.0621* (11.1) 0.3494* (17.7) 0.0730 (0.60)

Ln (AL) 0.0404* (2.04) -0.0770* (-2.40) -0.1838* (-3.03)

Ln (IC) 0.6269* (8.96) 0.3016* (4.67) 0.4614** (1.91)

Ln (L) 0.0340 (0.25) 0.3670* (5.60) 0.6224* (2.04)

Inefficiency effect model

Intercept 2.5852* (3.06) 0.1633 (0.25) 2.5891* (3.81) A. training -0.2429 (-1.50) -0.0664 (-0.75) -0.1449 (-0.83)

Age -0.7725 (-1.54) 0.3455 (0.91) -0.4462 (-0.59)

Age² 0.1061 (1.41) -0.0681 (-1.06) 0.0754 (0.65)

Specialisation -1.0030* (-3.01) -0.7584* (-6.84) -2.6438* (-3.30) Variance parameters

Sigma-squared 0.0825* (5.10) 0.0278* (5.44) 0.1404* (6.39)

Gamma 0.0825* (4.44) 0.0485* (5.44) 0.9900* (4.51)

Log-likelihood -6.116 29.78 5.257

N 46 64 32

Notes: * indicates significance at the 5% level, ** indicates significance at the 10 % level

(29)

6. Discussion and Conclusion

The objective of this paper has been to estimate technical efficiency scores and determine whether these scores can be explained by socioeconomic factors. This has been done for a sample of 142 cotton producers in Burkina Faso through the use of a stochastic production frontier function. The data sample that has been used in this paper was collected by INERA and is cross-sectional in nature. The variability of the variables was large and has been presented in the descriptive statistics. With this fact in mind the statistical inference that is done from the econometric results is not as solid and straight forward as it could have been.

The variables presented for the subsamples in Table 6 differ from the ones shown for the whole sample in table 3. The subsample variables for the Cobb-Douglas model do not have the right sign, compared to the whole sample. This difference becomes apparent in a comparison between the influence of Labour variable (AL) in the different samples. The labour variable shows a positive relationship toward the production for the whole sample, which in the subsamples shift when going from a less advance farming system (the Small sample farmers) towards a more advanced one (meaning the Medium and Large sample famers). The reason to this pattern could be several, such as the fact that more advance farmers use less labour in the production.

As it is shown from the estimation results of the TE-model Table 3, technical efficiency increases with the level of agricultural training. This relationship highlights the fact that the implemented policies from the government have had the desire outcome. It can also be detected that the level of technical efficiency increases as the farmers become more specialized in cotton cultivation. Meaning, the higher level of specialization, by devoting more arable land to cotton, the higher the level of technical efficiency. However, there is a danger for the farmers to implement a total specialization, especially the small farmers.

Because specialization involves risk taking, such as not being able to sell the harvest in time or risking a drop in the price. Therefore the small farm must become more self sufficient, due to these risks. In the descriptive statistics in Appendix 2 this pattern can also be detected, meaning the bigger the farm the more specialized are the farmers. Therefore there is a trade off between specializing and diversifying that the farmers are exposed to.

(30)

In theory the age variable is explained to have a pattern of diminishing return, meaning the older the farmers get the less productive they become. However, the empirical results from Table 3 indicates the opposite pattern, meaning that the level of productivity increase with age, due to the fact that Age have a negative and Age² a positive impact on the level of technical efficiency. An explanation to this pattern could be related to specification problems.

Because in this paper Age has been given a quadratic form, instead of using a non quadratic one. Another explanation could do with the creation of the variable itself. Meaning if the variable was continuous, instead of in categorizes, the result could have been different.

Some of the econometric results that have been obtained in this work are counterintuitive and therefore the results should be interpreted with some caution. One example are the gamma results that were obtained from the subsample in section 5.2. The high gamma value of 99 percent in subsample large and the low values in the other two subsamples indicate that there could be some problems with the database. But one has to have in mind the current situation of Burkina Faso. Being a developing country facing problems such as lack of accessibility in infrastructure and wide spread poverty, it is not an easy task to conduct a survey. Therefore during the collecting process problems with measurement could arises, which is difficult to tackle once the econometric process begins.

6.1 Final remarks

As discussed in the previous part some of the econometric results are counterintuitive, which may be related to the database, and should be interpreted with some caution. Nevertheless the result implies that the cotton farmers in Burkina Faso are quite efficient given the existing technology and inputs. Further policy program targeting the TE of the cotton farmers from the government is not likely to increase the production with much more.

However there exist other initiatives that the government can take in order to support the cotton sector and thereby increase the total welfare of the country. One of them is to continue pushing to lower the trade barriers imposed by first and foremost US and EU through the cotton initiative in the WTO. If the subsidies and trade barriers would decrease, the world

(31)

price of cotton would increase. This would provide incentives for cotton farmers in Burkina Faso to invest and thereby increase the production. Because the price that cotton farmers receive have during this past decade followed the same trend as the world market price. This would in turn be an important step toward increasing Burkina Faso’s competitiveness in the world market and improving the total welfare of the country.

However, the price of inputs, such as inputs used for intermediate consumption, and transportation cost have increased during the last period. This fact has a negative impact on the balance of trade for Burkina Faso, which in turn affects the welfare of the country in a unconstructive way.

(32)

7. References

Battese, G.E and Coelli, T.J. (1992). “Frontier Production Functions, Technical Efficiency and Panel Data: With Application to Paddy Farmers in India”, The Journal of Productivity Analysis, No. 3, pp. 153-169

Battese, G.E. (1992). “Frontier production functions and technical efficiency: a survey of empirical applications in agricultural economics”, Agricultural Economics, No. 7, pp. 185- 208

Bravo-Ureta, B. & Evenson, R. (1994). “Efficiency in agricultural production: the case of peasant farmers in eastern Paraguay”, Agricultural Economics, No. 10, pp. 27-37

Chakrabortv K, Misra S and Johnson P, (2002). “Cotton farmers' technical efficiency:

Stochastic and nonstochastic production function approaches“, Agricultural and Resource Economics Review, October 2002

Coelli, T. (1996). “A Guide to Frontier Version 4.1: A Computer Program for Stochastic Frontier Production and Cost Function Estimation, Center for Efficiency and Productivity Analysis (CEPA)”, Working Papers, No. 7, pp. 1-33

Haji, J, (2007). ”Production Efficiency of Smallholders’ Vegetable-dominated Mixed Farming System in Eastern Ethiopia: A Non-Parametric Approach”, Journal of African economies, 16 (1), pp. 1-27

Hårsmar, M. (2004). “Heavy Cloud But No Rain; Agricultural Growth Theories and Peasant Strategies on the Mossi Plateau, Burkina Faso”, Department of Rural Development Studies Uppsala, pp. 1-239

Jorgenson D.W. and Landau, R. (1989). “Technology and Capital Formation”, MIT Press Boston, p. 1-505

(33)

Schwartz, A, (1993). “L’Adhésion des paysans à la culture du coton au Burkina Faso. Des comportements contrastés”, ORSTOM, Bondy, France

Seyouma, E. Battese, G.E. Fleming, E.M (1998). “Technical efficiency and productivity of maize producers in eastern Ethiopia: a study of farmers within and outside the Sasakawa- Global 2000 project”, Agricultural Economics, No. 19, pp. 341-48

Tzouvelekas, V. Rozakis, S. and Pantzios, C. (2002). “Assesing the Perspectives of EU Cotton Farming: Technical and Scale Efficiencies of Greek Cotton Growers”, European Association of Agricultural Economists International Congress, Zaragoza, Spain, August 28- 31, pp. 1-21

”Länder i fickformat: Ghana och Burkina Faso” (2004) Utrikespolitiska institutet

Data sources

International Monetary Fund

http://www.imf.org/external/pubs/ft/weo/2008/01/weodata/weoselgr.aspx http://www.imf.org/external/pubs/ft/survey/so/2008/car022508b.htm Food and Agriculture Organization Corporate Statistical Database http://faostat.fao.org/site/567/DesktopDefault.aspx?PageID=567 New Agriculturist 2007

http://www.new-agri.co.uk/07/01/countryp.php

Swedish international development Cooperation Agency http://www.sida.se/sida/jsp/sida.jsp?d=274&a=924

Trade Analysis Branch, United Nations Conference on trade and Development http://192.91.247.38/tab/ATPSMabout.asp

World trade organisation

http://www.wto.org/english/tratop_e/agric_e/cotton_subcommittee_e.htm#top World Bank

http://siteresources.worldbank.org/IDA/Resources/IDA-BurkinaFaso.pdf

(34)

8. Appendix 1: Variable description

The socioeconomic variables that have not been included in the optimal inefficiency model are Analphabetic, Experience of the cotton cultivation and Child-Total labour ratio variable.

Analphabetic looks at the level of literacy among the cotton producing farmers. This variable is similarly constructed as the Agricultural training variable, meaning that it is a dummy variable with 1=not being analphabetic and 0=analphabetic.

Experience of the cotton cultivation variable is similar to the age variables, from that point of view that in the age variable it is possible to in a general way detect the amount of years in the cotton cultivation. However, this Experience variable states more directly the amount of years the cotton producing farmer have been involved in the cotton cultivation process. The variable is divided into 5 categories; 1=1-5 years, 2=6-10 years, 3=11-20 years, 4=21-30 years, 5=31 and over.

Another ratio variable, as the specialization variable above, is the Child-Total Labour ratio variable. This variable states the share of Child Labour in relation to Total Labour.

(35)

8.1 Appendix 2: Descriptive statistics

Table 7 Descriptive statistics over the sub-sample Small

Median Min. Max

Production Y (Kg)

1621.4 892.79 1437.5 596 3700 C (CFA) 171 703 126 457 144 604 0 517 422

AL 122.35 79.406 107 0 294.99

IC (CFA) 96 741 41 169 86 263 32 616 189 857 L (ha) 1.5023 0.67983 1.249 0.75 4 Inefficiency effects model

A. training 0.368 0.632

Age 2.8947 0.7983 3 2 5

Age² 9 4.9972 9 4 25

Specialisation 0.4610 0.1547 0.4397 0.1728 0.8571

(36)

Table 8 Descriptive statistics over the sub-sample Medium

Median Min. Max

Production Y

(Kg) 3 394.1 1 990.9 2 710.5 720 8 800 C (CFA) 379 837 274 687 310 405 4420 ^{1221 837}

AL 239.04 155.12 211 38 620

IC (CFA) 211 568 174 660 152 303 32 616 868 816

L (ha) 3.2937 2.6957 2 0.75 14

A. training 0.3125 0.6875

Age 3.2112 0.8062 3 2 5

Age² 10.958 5.1182 9 4 25

Specialisation 0.5628 0.1754 0.5342 0.1727 1

Table 9 Descriptive statistics the sub-sample Large

Mean Standard

deviation Median Min. Max

Production Y ^{9 609.7} ^{7 085.9} ^{7 660} ^{1 400} ^{29 400} C (CFA) 1159 884 1003 956 842 257 0 3982 184

AL 410.05 318.23 293 41 1232

IC (CFA) ^{425 957} ^{220 347} 447 094 84 546 1363 062

L (ha) 6.5309 3.3557 6.5 1 14

A. training 0.5625 0.4375

Age 3.8125 0.6445 4 2 5

Age² 14.937 4.4354 16 4 25

Specialisation 0.6182 0.1612 0.5924 0.3363 0.9113

(37)

8.2 Appendix 3: Labour formation

Preliminary estimation of the Frontier production function were not interpretable concerning the variable Labour. The Labour coefficient was not significant on the 5 % level and the sign of it was the reverse from what a Labour coefficient should have in a Cobb-Douglas functional form, meaning the sign was negative. So an increase in Labour would reduce the production of cotton.

After studying the article by Haji Jema (Haji, 2006;10) two new explanatory variables were created, Child and Adult Labour . Thereafter Frontier output estimation were conducted with different productivity levels of Child Labour as compared to Adult Labour (such as regarding Child labour as 0,1..0,25 as productive as 1 Adult Labour). Here the estimation results improved significantly when a child was regarded to be as one-fourth productive as an adult, this result was also identified by Haji Jema. The Labour variable had the right sign, e.g positive but only Adult Labour was significant on a 5 % level. Therefore, the Child Labour variable was not included in the stochastic production frontier.

A new TE variable was introduced, the Child/Total Labour (C/TotL) variable, as we saw improved result from the estimation with the two Labour variables. The C/TotL variable is a ratio variable between the amounts of Child Labour divided by total Labour and describes if the level of child Labour has any impact on the level of TE for each cotton producing farmer.

This introduction provided a better model by having better results for the coefficients. But, the C/TotL variable was still not significant on the 5 % level. Therefore, it was decided not to include the C/TotL variable in the TE-model. A distinction between Male and Female Labour was also made in the frontier output estimations to see if there is any effect imposed by gender. However, the coefficient values were not realistic because of the high effect.

Therefore these were not included in the model.

(38)

8.2.1 Capital formation

Due to a lack of a proper capital variable in the sample, it was essential to conduct preliminary studies of the production function in order to improve the function. So, useful measurement of capital inputs for the cotton producers in the sample had to be created.

Therefore a Capital stock rental prices of Capital services was created (Jorgenson and Landau, 1989;21), i.e for equipment, Intermediate inputs and hired labour. The rental price formula that has been created here, include rental prices for capital input assets which is used to calculate a rental cost, i.e a price (Jorgenson and Landau, 1989;332). These input services can includes the price for equipment (annual and rental), intermediate inputs (such as the price for herbicides, pesticides, fertilizers et cetera) and the monthly average labour cost (e.g salary paid to hired workers) for each capital stock. Thereafter each rental cost is combined and withdrawn from the total value of production (in monetary terms), which leaves us with a value for the Capital services.

However we start of by showing the relative efficiency of each sub-sample on each other by creating dummy variables for each group and see if these dummy variables in a regression are significant and affect the intercept, from here we can decide whether the Capital variable has any significant effect on cotton production. The construction of the dummy variables is in the following way:

D¹= were 1=if the farmer is Small and 0=otherwise

D²= were 1=if the farmer is Medium and 0=otherwise

We will include the sub-sample Large as the reference group. After regressing these dummy variables with the explanatory variables, the Maximum Likelihood estimates for the dummy variables are significant on the 5 % level. This is an indication that the different levels of Capital have an impact on the production level by the different subgroups in the sample having a greater impact on the production level. Such as if farmer is in the group Medium instead of Large, there is a less negative (more positive) impact on the level of production.

(39)

The conclusion that can be reached from including the Dummy variables makes it possible to include the Capital variable in the production stochastic function frontier. This give rise to the following equation:

farmer n

i U V L In IC

In AL

In C

In Y

In _i =β₀+β₁× ( _i)+β₂× ( _i)+β₃× ( _i)+β₄× ( _i)+( _i − _i), =1,2,... ^th

Analyzing the technical performance of cotton farmers in developing countries: The case of Burkina Faso

Uppsala university

Analyzing the technical performance of cotton farmers in developing countries:

The case of Burkina Faso

Abstract

Table of contents:

List of Abbreviations...4

1. Introduction ...6

2. Burkina Faso’s economy: An overview ...7

3. Theoretical framework ...12

4. Data description...18

5. Empirical Results: Whole data sample...24

6. Discussion and Conclusion...29

7. References ...32

8. Appendix 1: Variable description...34

List of Abbreviations

Preface

1. Introduction

1.1 Objectives and method

2. Burkina Faso’s economy: An overview

2.1 Macroeconomic setting

2.2 Agricultural sector

2.3 Cotton sector

2.4 Current situation

3. Theoretical framework

3.1 Theory

3.2 Literature review

4. Data description

4.1 Variable description and model specification

4.2 Descriptive Statistics

4.3 Complementary sampling

5. Empirical Results: Whole data sample

5.1 Specification test and Efficiency results

5.2 Empirical results: Sub-sample groups

6. Discussion and Conclusion

6.1 Final remarks

7. References

Data sources

8. Appendix 1: Variable description

8.1 Appendix 2: Descriptive statistics

8.2 Appendix 3: Labour formation

8.2.1 Capital formation