Analysis of socio-economic factors affecting technical efficiency of small-holder coffee farming in the Krong Ana Watershed, Dak Lak Province, Vietnam

Analysis of socio-economic factors affecting technical efficiency of small-holder coffee farming in the Krong Ana Watershed, Dak Lak Province, Vietnam

ANALYSIS OF SOCIO-ECONOMIC FACTORS AFFECTING TECHNICAL

EFFICIENCY OF SMALL-HOLDER COFFEE FARMING IN THE KRONG ANA

WATERSHED, DAK LAK PROVINCE, VIETNAM

UNIVERSITY OF HAWAI῾I AT MANOA

ANALYSIS OF SOCIO-ECONOMIC FACTORS AFFECTING TECHNICAL

EFFICIENCY OF SMALL-HOLDER COFFEE FARMING IN THE KRONG ANA

WATERSHED, DAK LAK PROVINCE, VIETNAM

A THESIS SUBMITTED TO THE GRADUATE DIVISION OF THE

UNIVERSITY OF HAWAI῾I AT IN PARTIAL FULFILLMENT OF THE

REQUIREMENTS FOR THE DEGREE OF

ANALYSIS OF SOCIO-ECONOMIC FACTORS AFFECTING TECHNICAL

EFFICIENCY OF SMALL-HOLDER COFFEE FARMING IN THE KRONG ANA

WATERSHED, DAK LAK PROVINCE, VIETNAM

A THESIS SUBMITTED TO THE DEPARTMENT OF NATURAL RESOUCES

AND ENVIRONMENTAL MANAGEMENT OF THE UNIVERSITY OF HAWAI῾I AT IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR

ACKNOWLEDGEMENTS

Since I enrolled in my graduate program at the University of Hawaii at Manoa, I have received tremendous assistance and encouragement from individuals and institutions First, I would like to express sincere thanks to my academic advisor, Prof

Dr John F Yanagida, who have continuously encouraged and assisted by providing me invaluable comments and suggestions for my thesis and academic plan as well I am deeply indebted to him for all endless support and kindhearted assistance during my study period

I would like to express my endless gratitude to Dr Prabodh Illukpitiya for his helpful comments and encouragement He provided extremely gracious suggestions with productive discussion throughout the thesis research I am grateful to Dr Tung Bui for his invaluable advice on my proposal, keeping track of research implementation conducted in Vietnam I hold my committee members, Dr John F Yanagida, Dr Tung Bui and Dr Prabodh Illukpitiya, in deep reverence

I gratefully acknowledge the International Fellowship Program and the East West Center for providing financial support for my graduate studies

Much appreciation is extended to students from Tay Nguyen University, namely Tran Ngoc Nam, Ha Van Dung and Do Thanh Tuyen, who helped with data collection I also extend my special thanks to faculty members from Department of economics, Tay Nguyen University for their advice and encouragement during my graduate studies

I am grateful to my family and friends, especially my parents, my wife, Hoang Thi Thu and my son, Ho Hoang Quan, who have exhibited the supreme virtue of patience and understanding

I am indebted to more people than I can name Thank you all for being present in this unforgettable stage of my life

ABSTRACT

Title: Analysis of socio-economic factors affecting technical efficiency of

small-holder coffee farming in the Krong Ana Watershed, Dak Lak Province,

Vietnam

Coffee is a major crop in Vietnamese agriculture and plays an important role in the country‟s economy, especially in the Central Highlands of Vietnam Coffee production is a major source of income for farmers in the Dak Lak province Although Vietnam is well-known as one of the largest coffee producers in the world, there is minimal research and information identifying the technical efficiency and socio-economic factors contributing to production efficiency of coffee

The overall objective of this study was to estimate the technical efficiency of coffee production and evaluate factors affecting the level of technical inefficiency of small holder coffee farmers in the Krong Ana Watershed of Dak Lak province The specific objectives were to (i) identify the factors affecting coffee production, (ii) estimate the technical efficiency of coffee farming and (iii) identify factors contributing

to technical inefficiency by analyzing the relationship between estimated efficiency levels and farm specific socio-economic factors

The study was conducted in four districts of the Dak Lak province Since pooling data was not possible in all districts based on the results of the Chow Test, separate analyses were conducted for Cu Kuin and three combined districts (Krong Ana, Krong Bong and Lak) Maximum likelihood estimates for all the parameters of the stochastic frontier and inefficiency model were simultaneously generated The variance parameters were estimated in terms of parameterization By employing the stochastic frontier

approach, the results reveal that selected variables significantly affect coffee output (i.e., cost of organic fertilizers, pesticide expenditure, amount of irrigation water and coffee trees for the Cu Kuin district model; labor, inorganic and organic fertilizer expenditure and age of coffee trees for the three combined districts model) The estimated mean technical efficiency scores were 0.7466 and 0.6836 respectively for the Cu Kuin district and the three combined districts Formal education of the household head, amount of credit, ethnicity, and coffee farming experience were key factors which can reduce technical inefficiency of coffee production for the combined districts sub-region For the

Cu Kuin district, extension services can be used as a conduit to reduce technical

inefficiency of coffee production, while ethnicity has the opposite effect as compared to a priori expectations This latter result requires further research and analysis Improvement

of technical efficiency by 10% could generate a substantial amount of additional income for coffee farmers The overall findings suggest that water conservation practices, and the proper choice of fertilizers and pesticides could lead to improvements in coffee yields Expanding coverage of formal education and making credit more available can help farmers enhance technical efficiency of coffee production in the combined districts Improvements in both quantity and quality of extension services may increase technical efficiency of coffee production for farmers in the Cu Kuin district More in-depth investigation into population policies is necessary to identify the effects of family labor, number of children and family size on improving technical efficiency of coffee production for both sub-regions

TABLE OF CONTENTS

ACKNOWLEDGEMENTS i

ABSTRACT iii

TABLE OF CONTENTS v

LIST OF TABLES vii

LIST OF FIGURES: viii

Chapter 1 INTRODUCTION 1

1.1 Background information 1

1.2 Problem statement 3

1.3 Objectives of the study 5

1.4 Testable hypotheses 5

1.5 Outline of thesis 6

Chapter 2 LITERATURE REVIEW 7

2.1 Factors governing coffee production 7

2.2 Overview of technical efficiency concepts 8

2.3 Factors affecting technical inefficiency 10

2.4 Conceptual framework 13

2.5 Knowledge gaps 16

Chapter 3 RESEARCH METHODOLOGY 18

3.1 Research design outline 18

3.2 Theoretical Model 18

Chapter 4 DATA SPECIFICATIONS AND COLLECTION PROCEDURES 26

4.1 Approach 26

4.2 Data sources 27

4.3 Variable descriptions 27

4.4 Sampling Procedures 29

4.5 Descriptive Statistics 31

4.5 Common Tests for the robustness of the models 32

4.5.1 Chow test 32

4.5.2 Collinearity Testing 33

4.5.3 Testing for Heteroskedasticity 34

Chapter 5 EMPIRICAL MODELS AND ESTIMATION RESULTS 36

5.1 Empirical Models 36

5.2 Results and Discussion: 38

5.2.1 Factors contributing to coffee production 38

5.2.2 Maximum Likelihood Estimates (MLE) and Technical Efficiency 39

5.2.2.1 Analyses of Maximum Likelihood Estimates 41

5.2.2.2 Technical Efficiency 46

5.2.2.3 Analyses for technical inefficiency models 47

5.2.2.4 Efficiency improvements in coffee production 51

Chapter 6 SUMMARY, CONCLUSIONS AND RECOMMENDATIONS 54

6.1 Summary 54

6.2 Conclusions 55

6.3 Limitations of the study 57

6.4 Policy recommendations 58

Appendix A: Descriptive Statistics 71

Appendix B1: Correlation matrix of production factors for Cu Kuin district 72

Appendix B2: Correlation matrix of production factors for the combined districts 73

Appendix C: Pairwise t-test for the mean 74

LIST OF TABLES

Table 4.1: Description of production variables 28

Table 4.2: Description of variables of the efficiency model 28

Table 4.3: Descriptive statistics for sample size determination 29

Table 4.4 Sample distribution* 30

Table 4.5 Coffee output per hectare statistics of 4 districts 31

Table 4.6 Chow test for different impacts of independent 32

Table 4.7 Multicollinearity Testing 34

Table 5.1 OLS estimates fo coffee production function models 38

Table 5.2 MLE of stochastic production frontier and technical inefficiency models 40

Table 5.3 Frequency distribution of technical efficiency estimates 46

Table 5.4 Scenario for increasing the efficiency 52

Table 6.1 Change in revenue as the inputs increase by 1% 59

LIST OF FIGURES:

Figure 1.1: Krong Ana Watershed, Vietnam 2 Figure 2.1 Output- oriented technical efficiency 9 Figure 3.1: Stochastic frontier production function (Battese 1992) 24

Chapter 1 INTRODUCTION

1.1 Background information

Coffee farming was first introduced to Vietnam by the French in the 1850‟s; however, following that introduction, coffee production remained relatively low It was not until after the country‟s re-unification in 1975 and implementation of the national policy, involving setting “new economic zones” in the Central Highlands, those migrants from across the country settled in the Central Highlands to instigate coffee production in this area After this occurrence, Vietnamese coffee production increased significantly and Vietnam soon became the fourth largest coffee exporter in the world To be specific, in 1998, Vietnam accounted for approximately 6.5% of the world‟s coffee production This number continued to increase in subsequent years contributing to 13.2%, 14.6% and 12.4% of the world‟s coffee production in crop year‟s 2000/01, 2003/04, and 2008/09, respectively (de Fontaney and Leung, 2002; ICO, 2010) According to the International Coffee Organization (ICO), Vietnam had become the second largest coffee producer worldwide trailing only behind Brazil between the periods of 2000/01 to 2008/09

Figure 1.1: Krong Ana Watershed, Vietnam

Coffee is the primary export crop for Vietnamese agriculture and plays an important role in the country‟s economy This is especially true for the Central Highlands of Vietnam and its role in the world coffee market In these Central Highlands, farmers‟ incomes are totally dependent on coffee in the Dak Lak Plateau and the Krong Ana watershed in particular The Dak Lak province has been the largest coffee producer in terms of both coffee yield and land area in Vietnam Therefore, it is apparent that agricultural production in this area has been dominated

by coffee production (Business Portal)1 To illustrate, land area, as it relates to coffee farming in the Dak Lak province, represents approximately 236,200 hectares The remainder of the country‟s coffee-farming land area is approximately 270,000 hectares and represents more than ten provinces Next, the Krong Ana watershed, known for its large coffee plantations in the province, lies along the Krong Ana River and includes a part of Krong Bong, Lak, Cu Kuin, and Krong Ana districts within the Dak Lak province

These areas represent major contributors to the coffee industry not only in coffee production within Vietnam, but also in development, production and research

as it relates to the world coffee industry Thus, it can be proposed that changes in coffee production within the Dak Lak province, will not only affect farmers‟ livelihood and the provincial economy, but also have an influence in the global coffee market

through their exploitation Therefore, by improving technical efficiency this will result in not only efficient use of these natural resources, but also an increase in coffee production and subsequently an increase in profits for coffee farming Pursuit of technical efficiency also ensures that negative impacts on socio-economic factors are minimized and thereby, serves as an added incentive for sustainable development

The Central Highlands of Vietnam is well-known as an area rich in natural resources with favorable climatic conditions for agricultural development; however, human resource quality is still poor compared with other regions across the country A study by ICARD and Oxfam (2002) shows that ethnic minority farmers have been limited in the adoption of coffee producing technology and have experienced limited access to support for their coffee production This is due to the limited formal education and the lack of skills necessary to keep up with the continuously changing coffee industry; as such, these ethnic farmers in the Central Highlands of Vietnam are experiencing more difficulty in reaping benefits from this fast growing coffee industry (Marsh, 2007) It is this relationship among socio-economic factors such as education level which plays an important role in problem solving, e.g., the efficient use of natural resources and developing strategies to achieve sustainable development in the study area

Although Vietnam is known as one of the largest coffee producers in the world, there is minimal research and information regarding coffee production in this area, in particular, efficient coffee production and socio-economic factors influence coffee production efficiency This study is important because it investigates resource use in coffee production and subsequently efficient use of these resources Although this study examines the potential development in the Dak Lak province, these results can be applied to other areas having similar characteristics and problems The

findings of this research will be a useful source of information for coffee producers and policy makers seeking production strategies for coffee development

1.3 Objectives of the study

The overall objective of the study is to estimate the technical efficiency of coffee farming and to evaluate factors affecting the technical inefficiency in small holder coffee farms in the Krong Ana Watershed, Dak Lak province, Vietnam Results from this study could suggest policy measures to improve coffee farming in

this area The specific objectives are three-fold:

(1) Identify factors governing coffee production in the Krong Ana watershed, Dak Lak province, Vietnam;

(2) Estimate technical efficiency in coffee farming through the stochastic production frontier;

(3) Identify factors contributing to technical inefficiency by analyzing the relationship between estimated efficiency levels and farm specific socio-economic and geographic factors

1.4 Testable hypotheses

Whether the Ordinary Least Square (OLS) or Maximum Likelihood Estimate (MLE) is adequate to capture the data? The null hypothesis is that the stochastic production functions for coffee production in the research sites do not exist Rejecting the null hypothesis implies that the conventional estimation (OLS) is not adequate to represent the data and MLE is more appropriate

By estimating a Cobb-Douglas function, the role of each input factor in coffee production can be analyzed In order to do so, one should determine the appropriate

The testable hypotheses for factors affecting coffee output are whether the estimated regression coefficients associated with coffee input variables are statistically equal to zero Rejection of the null hypotheses is expected for estimated regression coefficients Therefore, how and by how much these factors affect coffee output will be determined and analyzed

The existence of technical inefficiency of coffee production will also be tested The null hypothesis for the existence of technical inefficiency is that there are no inefficiency effects in coffee production The rejection of this null hypothesis implies that technical inefficiency does exist in coffee production

As a corollary to the existence of technical inefficiency, the hypothesis involving the effects of socio-economic factors on technical inefficiency is tested The null hypotheses are that estimated coefficients of socio-economic variables are zero The rejection or acceptance of these null hypotheses identifies factors which significantly affect technical inefficiency of coffee production

Chapter 2 LITERATURE REVIEW

There have been a number of studies on the economics of resource issues and farm specific factors affecting production efficiency e.g., Huang and Liu (1994), Cardenas et al (2005), and Nchare (2007) Socio-economic factors such as labor, fertilizers and pesticides and resources such as irrigation water, and land play vital roles in modern-day economies and society as they relate to coffee production Each

of these contributors has particular characteristics that can be quantified and appropriately applied in multiple ways For example, Nachare (2007) found that land, labor, fertilizers, pesticides and capital positively affected Arabica coffee output

Approaching social effects and economic resources in terms of efficiency is especially necessary for every economy as it can help to see how technical efficiency affects resource use by comparing actual production levels with the maximum possible output levels If a producer is able to obtain a high technical efficiency score, which is considered as a point close to the production frontier, this infers that the use

of resources is efficient and that the socio-economic factors positively affect efficiency Moreover, the increase in production potential, that is achievable, may be determined

2.1 Factors governing coffee production

A number of studies have examined input factors affecting agricultural production in general and coffee production in particular For instance, Nchare (2007) considered labor in person-day, fertilizers in physical units, pesticide costs and age of coffee producers as input factors for coffee production function Similar application of factors for the production frontier function of small food crop

production in Cameroon was also found by Kolawole and Ojo (2007), implying the important roles of labor and fertilizers and other inputs as well In addition, Illukpitiya and Yanagida (2004) showed that labor, inorganic and organic fertilizers are primary input factors for paddy production in Sri Lanka These factors were also employed by Chen et al., (2009) for Chinese farms, (also see Alvarez and Arias (2004); Binam et

al (2004); and Illukpitiya and Yanagida (2010))

Irrigation is another key factor, influencing coffee production in Vietnam (Chi and D‟haeze 2005; D‟haeze 2004) However, previous studies suggest conflicting optimal strategies with varied water application amounts For example, Cheesman et

al (2007) found coffee smallholders in the Dak Lak province tend to over-irrigate by

a factor between 2.2 and 5 times the dosage required to achieve the most favorable water condition for a coffee plantation that can maximize a coffee tree‟s yield While D‟haeze et al (2003), in the Centre Highlands, Vietnam, recommended an optimal efficient use of water to be around 1,000 cubic meters per hectare during flower setting of coffee production Local authorities suggest that during the dry season, from November to April of the following year, irrigation application of 3,300 cubic meters per hectare is the most favorable condition for coffee plantations (D‟haeze, 2004 and Luu, 2002)

2.2 Overview of technical efficiency concepts

The term technical efficiency of production can be described as the producer‟s ability to develop the greatest amount of output possible from a fixed amount of inputs In other words, an efficient producer is one that given a state of technical know-hows can produce a given quantity of goods by using the least quantity of possible inputs (Nchare, 2007) Another description of technical efficiency includes

the ability to minimize input use in the production of a given output vector, or the ability to obtain maximum output from a given input vector (Kumbhakar and Lowell, 2000)

According to Nchare (2007), the first analyses of efficiency measures were initiated by Farrell (1957) Drawing from Debreu (1951) and Koopmans (1951),

Farrell proposed a division of efficiency into two components: technical efficiency,

which represents the producer‟s ability to maximize output from a given level of

inputs, and allocative efficiency, which is the ability of a producer to use inputs in

optimal proportions, given their respective prices and available technology Subsequently, a combination of the two measures yields a measure of economic efficiency

Figure 2.1 Output- oriented technical efficiency

Figure 2.1 shows the output-oriented measures of technical efficiency; for example, an observed production plan (XB, YB) is illustrated by point B, in Figure 2.1 The vertical length that connects points A and B (AB) is the amount of output that could increase while input N remained fixed on the x-axis; thus, according to

Kumbhakar and Tsionas (2006), the ratio of AB/AN represents output-oriented technical inefficiency

2.3 Factors affecting technical inefficiency

Since the study completed by Farrell (1957), a number of additional empirical studies have evaluated production efficiency Many studies such as Aigner et al., (1977), Battese (1992), Bravo-Ureta and Pinherio (1993), Battese and Coelli (1995), Binam et al (2004), Illukpitiya and Yanagida (2004) and Rios and Shively (2007) have highlighted effects of common socio-economic variables such as age, education, experience, gender etc on technical efficiency Among the literature on technical efficiency, several selected studies are relevant to the topic of agricultural production and specifically to coffee production

A study on the paddy farming sector in Sri Lanka (Illukpitiya and Yanagida, 2004) showed that educational level, experience of the head of household, age of the head of household and extension contacts, had positive and significant impacts on technical efficiency This study found that farmers with more experience were superior technically in the efficient use of resources Technical efficiency in this study

is represented by an estimated mean value of 74% and ranging from 34% to 95% The finding of this study also showed that there is potential for increasing the efficient use

of resources of paddy farming and thereby improving farm revenues

In their study on technical efficiency among smallholder farmers in Cameroon, Binam et al (2004) selected several interesting variables affecting technical efficiency in the agricultural sector For example, variables such as the distance of the farm plot from the main access road, the soil fertility index and credit access, all had a significant relationship with the efficiency of farmers among farming

systems in the slash and burn agriculture zone On the other hand, educational levels had a significant impact on the technical efficiency of maize mono-cropping systems and an insignificant influence on the technical efficiency of other crops The study indicates that the mean technical efficiency levels for groundnut, maize-groundnut and maize are 71%, 73% and 75% respectively

Binam et al (2004), in Cameroon, also reports that the variable, club membership of a household member, plays a role of social capital in providing incentives for efficient production This variable has a negative and statistically significant effect on technical inefficiency Also, this study finds that access to cash credit is likely to increase technical efficiency of farmers in the slash and burn agriculture zone of Cameroon, by providing funds to reduce capital constraints hindering agricultural households

In addition, an analysis of production efficiency involving Mexican producing districts by Cardenas, et al (2005), showed that from 1997 to 2002, coffee producers experienced a high level of technical efficiency on average, which increased by 1.04% (0.9586 to 0.9720) during this period The estimation of technical efficiency ranged from 0.89 (least efficient) to 0.996 (most efficient) This increase in technical efficiency relates to infrastructure use, therefore, having a positive influence

coffee-on the efficiency of coffee producticoffee-on, an unexpected observaticoffee-on More specifically, this study hypothesized that the availability of roads is a factor impacting coffee productivity Results revealed that road availability was negatively related to agricultural production, an unexpected outcome On the other hand, factors such as population density, altitude, and the proportion of cash crops to coffee based on monetary value positively contributes to the level of technical efficiency in coffee production

In a different study, in the Cote d‟Ivoire, Binam et al (2003) report that technical efficiency results are much lower than previous studies of technical efficiency in the agricultural sectors This study shows that technical efficiency scores vary widely, ranging from 0.02 to 1, with a mean value of 0.36 when using the Charnes et al model (1978), and from 0.05 to 1, with a mean value of 0.47 when using the Banker et al model (1984) This could imply that improvements in coffee yield are significant because there were a number of households with actual output far below the production frontier

Binam et al (2003) also suggest in their study that membership to a farmer‟s club or association, family size, and residence status of the farmer, have significant influences on technical efficiency of coffee production in the Co te d‟Ivoire area To

be specific, the membership to a farmer‟s club or association negatively impacts coffee yield while the residence status of the farmer significantly influences technical efficiency This study also reveals statistically significant results that younger heads

of households are more efficient than older heads of households In contrast, accessibility to credit, farming system, land tenure, farm management contract and distance from the house to the farm for this study are not statistically significant A notable finding in the Cote d‟Ivoire area, compared to the study by Illukpitiya and Yanagida (2004) in Sri Lanka The latter result is that the level of formal education, commonly measured by the number of years of schooling, does not clearly influence technical efficiency in coffee farming However, Nchare (2007) found that there were positive relationships between technical efficiency scores and educational level of coffee producers in Cameroon and access to credit also significantly and positively affected technical efficiency of coffee production Additionally, Illukpitiya and

Yanagida (2010), and Kolawole and Ojo (2007) also indicated a similar results of the effects of education and credit on technical inefficiency of agricultural production

Muhammad-Lawal et al., (2009) revealed that young people may not be able

to contribute to agricultural production activities because young people are likely to

be in school Thus, the ratio of number of children and family size can be explained as

a child dependency index that may affect the technical inefficiency of agricultural production

A recent study examining technical efficiency of organic and conventional coffee farming in Kona, Hawaii (Masuda, et al 2010), shows that reducing chemical inputs in conventional Kona coffee farming can improve coffee yields The estimated efficiency levels were 0.5775 and 0.4696 respectively for organic and conventional coffee farming

From this overall analysis, the educational level of the household head is positively related to the technical efficiency (see for example, Illukpitiya and Yanagida, 2004; Binam et al 2004; Kehinde, et al 2010; Nchare, 2007), whereas, Rios and Shively (2007) report that increased educational attainment by household heads of small farms is a factor reducing efficiency This is perhaps because more education increases opportunities for off-farm work and thereby decreases on-farm management intensity Access to credit has a negative influence on technical inefficiency which is statistically significant (Nchare, 2007) while other studies did not incorporate this factor

2.4 Conceptual framework

Masuda, et al (2010) and Binam et al (2003) indicate that technical efficiency scores in coffee production are lower than other agricultural production; however,

Cardenas, et al (2005) found that the level of technical efficiency for coffee production in Mexico, specifically, to be very high Moreover, since Illukpitiya and Yanagida (2004) show potential efficiency improvement for paddy farming in their study, then coffee, which commands a much higher market price, could in turn have greater potential than rice production if improved with the same efficiency For example, if inefficiency is evidenced in coffee farming systems, there is a potential to improve income of farmers by reducing technical inefficiency Therefore, this potential increase in technical efficiency could lead to higher income from coffee Thus, a study examining the technical efficiency of coffee production for the Dak Lak province, known as the largest coffee plantation in Vietnam, has significant importance and value to coffee producers and the economy of Vietnam

The OLS method is frequently employed to estimate the production function This procedure estimates the production function with the assumption that there is no efficiency term However, to investigate technical efficiency and socio-economic factors affecting technical efficiency, the MLE is a commonly used method for analyzing inefficiency effects (see Kumbhakar et al (1991) and Reifschneider and Stevenson (1991)) This approach is known as the one-step procedure to simultaneously estimate parameters for both the stochastic production function and inefficiency model Huang and Liu (1994) proposed that the technical inefficiency effects are a function of a number of the firm‟s specific factors and of interactions among these factors

In this study, the dependent variable is coffee yield for a particular crop year The explanatory (independent) variables that will be analyzed include labor, farm‟s expenditures on inorganic fertilizers, organic fertilizers and pesticides, amount of irrigation water applied, and age of coffee trees Previous studies also utilized similar

variables, but not in the same manner and combination used in this proposed model For instance, Masuda et al (2010) employed fertilizers, labor, and other inputs in coffee production Binam et al (2004) simply specified land, labor and capital as other production factors in their model In addition, Cardenas et al (2005) used area planted in coffee, fertilizer use and machinery use as their selected input variables Nchare (2007) employed total area planted in coffee, amount of labor including family and hired labor, total quantity of chemical fertilizers, cost of pesticides, age of coffee trees, and capital as explanatory variables in the coffee production function for Cameroon

None of the previously mentioned studies analyzed irrigation water as an important input in the coffee production function However, the studies by D‟haeze et

al (2003) and Cheesman et al (2007) mentioned that coffee plantations should apply ample amounts of water for high coffee yields Thus, irrigation is widely viewed as an essential input in coffee production and should be included in the proposed production model

To model technical inefficiency, key socio-economic variables affecting technical efficiency in coffee production are examined These factors affecting technical efficiency in the proposed model are age of the household head, formal educational level of the household head, ethnicity, access to extension services, amount of credit loaned, experience in coffee farming, and child dependency index Most of these variables were analyzed in previous studies such as Iinuma, et al (1999), Binam et al (2003), Binam et al (2004), Cardenas et al (2005), and Rios and Shively (2005) However, the estimated effects from these explanatory variables on efficiency (or inefficiency) were found to be different among previous studies This is

due to regional differences that are likely to be different in farm‟s specific characteristics

The findings by Binam et al (2004) are useful for this proposed study Given the agro-climatic environment of the Dak Lak province, i.e., the climatic conditions and soil quality, this area is the most appropriate for coffee farming and coffee production in Vietnam Thus, factors such as the soil fertility index and climatic conditions may not vary amongst farms in the Krong Ana watershed

2.5 Knowledge gaps

Given this information, the intended research will focus on specific factors identified through literature surveys of model specifications in measuring productivity

of coffee farming and technical efficiency; however, knowledge gaps still persist

To begin with, a number of studies mentioned above did not examine the water scarcity variable or the irrigation variable in depth Thus, the effect of irrigation application on crop productivity has not been extensively analyzed in previous studies As mentioned earlier, irrigation water plays an important role in coffee production Although climatic conditions in the Central Highlands of Vietnam are quite stable, the quantity of irrigation water in some crop years has seriously affected coffee yield in the Dak Lak province For this study, the amount of water application

is reviewed as an important factor in the stochastic production function

Nchare (2007) also found no significant relationship between experience of producers and coffee productivity in Cameroon, West Africa According to the literature survey, there are other different findings showing opposite impacts from education on agricultural production in general and coffee production in particular

It is apparent from the survey literature that there were only a few studies analyzing other factors such as ethnicity Dak Lak province in Vietnam has a very diverse ethnic composition due to a variety of different cultures It is these variations

in traditions and values which affect work ethic and agricultural production Although indigenous Vietnamese farmers account for approximately 70% of coffee farmers in the proposed site, the two largest minority group, Ede and M‟nong, comprise approximately 20% of the provincial population (Dak Lak people committee website)2 This suggests that further examination of ethnicity of the household head warrants scrutiny as a variable in the efficiency model

2 Dak Lak people committee website Basic Socio-economic information of Dak Lak province, available at http://cema.gov.vn/modules.php?name=Content&op=details&mid=7786 (accessed on

Chapter 3 RESEARCH METHODOLOGY

3.1 Research design outline

This study focuses on coffee production in the Krong Ana Watershed, Dak Lak province, Vietnam Data collection of both primary and secondary data involving socio-economic characteristics of coffee production and the farm household will involve use of a questionnaire

From the data collected, a Cobb-Douglas production function and a stochastic production frontier will be estimated Data collected on a per farm basis will include coffee production, farm size, labor allocation, fertilizer & pesticide application, application of irrigation water, age of coffee trees, and total area planted with coffee The theory employed to estimate the production function and stochastic production frontier include elements of production economics, household production, and econometrics

Other important components of this study include estimating the technical inefficiency function and quantifying technical efficiency levels of all farm observations The Frontier 4.1 software will be used for the econometric estimation of the Cobb-Douglas and stochastic frontier function

3.2 Theoretical Model

Since the seminal article on efficiency measurement by Farrell (1957), the basic stochastic frontier model was independently proposed by Aigner et al., (1977) and Mueeusen and Van den Broeck (1977) Various other models have been suggested and have applied in the analysis of cross-sectional and panel data The efficient frontier can be considered as either the maximum level of output for a given set of

inputs (an output orientation), or the minimum set of inputs required to produce a given set of output (an input orientation) (Tingley et al 2005)

There are two principal methods for measuring efficiency, which are the stochastic frontier production function with the associated technical inefficiency model and data envelopment analysis (DEA) proposed by Charnes et al., (1978) DEA is based on linear programming, consisting of estimating a production frontier through a convex envelope curve formed by line segments joining observed efficient production units This approach is called a non-parametric method with no assumption about the error term The stochastic frontier is known as a parametric method and considered to

be more fitting than DEA in agricultural applications, especially in developing countries, where errors in collecting and measuring data may occur along with effects from climate, weather conditions, diseases, etc., which are uncontrollable (Coelli et al 1998) Compared to the stochastic production frontier approach, DEA is unable to test important hypotheses such as whether technical inefficiency exists Also, DEA does not capture measurement errors and random effects Moreover, DEA is very sensitive

to extreme values and outliers, since this method is based on comparison of individual farms and their peers (Nchare, 2007)

According to Kumbhakar and Lovell (2000), there are two main approaches used to analyze the determinants of technical efficiency with the stochastic production function framework Early studies adopted a two-step approach, in which efficiencies are estimated in the first stage, and in the second stage, estimated efficiencies are regressed against a vector of explanatory variables This two-step approach has been used by authors such as Pitt and Lee (1981) and Ben-Belhassen (2000) in their studies However, this approach reveals inconsistencies In the first step, inefficiency effects are assumed to be independently and identically distributed Nevertheless, in

the second step, the technical efficiency predictors obtained are assumed to depend on

a certain number of farm‟s specific characteristics as explanatory variables This suggests that the inefficiency effects are not identically distributed unless all the estimated coefficients of the factors considered happen to be simultaneously significant (Nchare 2007)

To deal with the drawback of the two-step approach, Kumbhakar et al (1991), Reifschneider and Stevenson (1991), Huang and Liu (1994) and Battese and Coelli (1995), have adopted a single-stage approach in which explanatory variables are incorporated directly into the inefficiency error component In this method, the variance of the inefficiency error component is hypothesized to be a function of firm‟s specific factors

There are a number of previous studies which have successfully used the stochastic frontier There are two main types of production frontier models namely cross-sectional and panel data production frontier models According to Kumbhakar and Lowell (2000), both cases consist of estimates of parameters describing the structure of the production frontier, and estimates of the output-oriented technical efficiency for each producer

Reviews of some of these models and their applications are given by Battese (1992), Bravo-Ureta and Pinherio (1993), and Coelli (1995) Some models have been proposed in which the technical inefficiency effects in the stochastic frontier model are also modeled in terms of other observable explanatory variables (see Kumbhakar et al., 1991; Huang and Liu, 1994; Battese and Coelli 1995)

However, the estimation techniques applied depend on the richness of the available quantitative data In this study, the cross-sectional production frontier model was formulated to reflect that farmers in the research site do not normally keep

records on past activities hence data collection is dependent on recall method Farmers have fairly good knowledge on their input application and coffee production

on their plantations for the current crop year

The basic stochastic frontier model was simultaneously introduced by Aigner, Lowell, and Schmidt (1977) and Meeusen and van de Broek (1977) In this study, a stochastic production model proposed by Battese and Coelli (1995) has been chosen

(3.1)

Where yi is the production of the i-th firm, i = 1,…n; xi is a vector of m inputs

used by the i-th firm; β j is a vector of parameters to be estimated; the random error, V i

where i = 1,…n, captures the effects of statistical noise, which are assumed to be

independently and identically distributed as N(0, ); U i where i = 1,…n are

non-negative random variables, associated with technical inefficiency in production, which are assumed to be independently and identically distributed exponential or half-

normal variable [U i ~ (|N(0, )|)] The deterministic production function is written

as: f (x i ; β), while [f (x ij ; β j ) exp {v i}] is the stochastic production frontier Both exponential and half-normal distributions have been criticized for arbitrarily restricting the mean of technical inefficiency effects to zero and related consequences for estimated technical efficiency levels A few authors have proposed alternative specifications for the technical inefficiency effects For instance, Stevenson (1980)

proposed use of the truncated-normal distribution [U i ~ iid (|N(μ, )|)], Greene (1990) proposed the two-parameter gamma distribution, and the model for technical inefficiency effects proposed by Battese and Coelli (1995) Although there is generally no priori justification for the choice of any particular distributional forms for the technical inefficiency effects, the generalized truncated-normal distribution has been most frequently applied in empirical applications due to its computational

simplicity However, in recent years, the model for the technical inefficiency effects proposed by Battese and Coelli (1995) has become quite popular because of its computational simplicity as well as its ability to examine the effects of various firm-specific variables on technical efficiency in an econometrically consistent manner, as opposed to a traditional two-step approach (Sharma and Leung, 2001)

Technical efficiency of the ith producer can be described as:

This equation defines technical efficiency as the ratio of observed output to the

maximum feasible output in an environment characterized by exp {V i} It implies that

y i can obtain its maximum feasible value of [f (x ij ; β j ) exp {v i }] if and only if TE i = 1

Otherwise TE i < 1 provides a measure of the shortfall of observed output from

maximum feasible output in an environment characterized by exp {v i}, which is allowed to vary across producers

Equation (3.2) can be rewritten as:

(3.3)

Where TE i = This form is chosen because of the simplification

when taking natural logarithms The value of TE i falls between zero and 1, so that u i

must be greater or equal to zero

By taking natural logarithms on both sides of equation (3.3), it becomes:

Where, “ln” represents the natural logarithmic transformation, the subscript j

= 1 … m, is number of observed input factors V i is random error having zero mean, which is associated with random factors not under the control of the producer In the first phase of the analysis, technical efficiency effects for a cross section of coffee farmers will be modeled in terms of input variables in the production process Various

statistical hypotheses will be tested for the coffee farmers and their technical efficiencies of production will be estimated

A Cobb-Douglas production function and the maximum likelihood estimates for all the parameters of the stochastic frontier defined by equation (3.3) and the inefficiency model defined by following equation (10), will be simultaneously obtained by running program, FRONTIER Version 4.1 (Coelli, 1996) Also estimated are the variance parameters in terms of:

with

where, the parameter has a value between zero and one

Equation (3.3) constitutes the technical inefficiency effects model in the stochastic frontier of equation (3.2) Considering the stochastic frontier production

function defined by equation (3.2), the technical efficiency of farm i, written as TEi, is defined according to Battese et al (1998) as:

where u i are the non-negative random variables, called technical inefficiency

effects These u i are assumed to be independently distributed and defined by the truncated normal distribution, with mean, i, and variance u i is defined by:

where, W i for i = 1,…n are random errors, defined by the truncation of the

normal distribution with mean zero and variance, The point of truncation is - i.e., Wi ≥ - The Zis are the firm-specific variables which may also include input variables in the stochastic production frontier, provided that the technical inefficiency effects are stochastic For the case where the Z-variables include interactions between

firm-specific variables and input variables, the Huang and Liu (1994) non-neutral stochastic frontier is obtained

Figure 3.1: Stochastic frontier production function (Battese 1992)

Figure 3.1 indicates the basic structure of the stochastic production frontier

model (3) The figure describes the production activities of two firms, represented by i and j Firm j uses inputs with values given by x j (the vector x j) and obtains the actual

output, Y j , but the stochastic frontier output, Y j *, exceeds the value on the deterministic

production frontier, f(x j ;β), because its production activity is associated with

„favorable‟ conditions for which the random error, V i, is positive On the other hand,

firm i uses input with values given by x i (the vector x i ) and obtains the output, Y i,

which has corresponding frontier output, Y i *, which lies below the value on the

deterministic frontier function, f(x j ;β), because its production activity is associated with „unfavorable‟ conditions for which the random error, V i, is negative In both cases, the observed outputs are less than the corresponding frontier values, but the stochastic frontier production values lie around the deterministic production function

Deterministic production function

Observed output Y

i

associated with the producers involved It is also possible that a stochastic frontier

value lies on the deterministic frontier, if the random error, V, equals to zero This

case may happen if the observed output, stochastic production frontier value and

deterministic production frontier are all equal and the random error, V, and the technical inefficiency effects, U, both equal to zero

Chapter 4 DATA SPECIFICATIONS AND COLLECTION

PROCEDURES

4.1 Approach

The study was conducted in the research site consisting of 4 districts, namely,

Cu Kui, Krong Ana, Krong Bong, and Lak, in Dak Lak province, Vietnam, which lies

in the Krong Ana watershed Coffee is the major agricultural crop and plays a crucial role in the well-being of people in the province In 2009, total coffee planted area in the province was 181,960 hectares accounting for about 40% of the total agricultural area in the province while the research site has an area of 20,522 hectares for coffee farming (Dak Lak Statistics Office 2009)

The socio-economic situation of households in the Krong Ana watershed and their possible effects on the efficiency of coffee production was of particular interest

in choosing this study site For instance, socio-economic distribution is very different amongst the region This perhaps can be explained by soil quality and other geographic conditions in the Cu Kuin district being more favorable for agriculture and coffee farming

Population density in the Lak, Krong Bong and Krong Ana districts is also lower than in the Cu Kuin district Higher income in the Cu Kuin district is an explanation for the earlier migration To deal with these problems, a number of people

with various ethnic groups immigrated from the north (Cu Kuin and Krong Ana) into the southern Krong Ana watershed (Krong Bong and Lak) from 1994 - 1999 as a

result of the Government‟s program redistributing population to the economic zones (MRC 2011)3 Therefore, this study is interested in examining how socio-economic

3

Mekong River Commission (MRC) (2011) Watershed Baseline Survey – An Example from the

factors affect the efficiency of coffee production for different districts in the research site

4.2 Data sources

Primary data:

A questionnaire is developed to gather both farm level coffee production data

as well as household demographic information The method of data collection adopted

is face- to- face interviews The questionnaire consists of demographic information about household characteristics, input and output data for coffee production, and socio-economic and geographical information related to coffee production

The primary data collected was for the 2008/9 crop year The household survey was conducted during the period June – July, 2010 Stratified random sampling was the sampling technique used

Secondary data:

Materials and official documents related to coffee production in the research site were collected A major source of the secondary data collected was the Statistical Yearbook 2009 of the Dak Lak Province

Table 4.1: Description of production variables

ValueInorF (X2) Cost of inorganic fertilizer (million VNDs per hectare) ValueOrF (X3) Cost of organic fertilizer (million VNDs per hectare) ValuePes (X4) Cost of pesticide (million VNDs per hectare) Irrigamount (X5) Amount of water applied (thousand cubic meters s per hectare)

Theoretically, inputs are measured in physical units for production functions However, because a variety of fertilizers and pesticides were used in coffee production, these inputs are measured in monetary units Note that this was determined during the data collection phase that coffee farmers in the research sites normally used many different types of inorganic fertilizers, organic fertilizers, and pesticides Simply adding these different types of inputs together was not an appropriate procedure Rather, using costs or expenditures may be a “better” representation of these variables However, since monetary values are used, the estimated parameters for the Translog Cobb-Douglas production function are not output elasticities

Table 4.2: Description of variables of the efficiency model

Edu (Z 2 ) Number of years of formal education for the household head Eth (Z 3 ) Dummy variable for ethnicity (if Vietnamese Kinh = 1 or otherwise 0) Ext (Z 4 ) Extension services used (yes = 1 or otherwise 0) Cre (Z 5 ) Amount of credit loaned from banks and credit organizations (in million

VND) Exp (Z 6 ) Number of years (experience) in coffee farming by the household head Cdpindex(Z ) Number of children divide by family size for each individual household

4.4 Sampling Procedures

According to Johnson and Bhattacharyya (2006), the appropriate sample size depends on the researcher‟s desired level of confidence for estimated results, the population standard deviation and the desired error margin, which can be illustrated as the following formula:

(4.1)

Where n denotes the sample size,

represents the upper point of the standard normal distribution, is the population standard deviation and d is the desired error margin Population standard deviation is based on small sample data (i.e., pretesting) with the assumption that the population standard deviation is approximated by sample standard deviation

In this research, a 95% confidence level is used, thus statistically we can calculate the value of

= 1.96

Next, parameters and d are calculated from the pretesting data These parameters are shown in Table 4.3

Table 4.3: Descriptive statistics for sample size determination