國立臺灣大學生物資源暨農學院農業經濟研究所 碩士論文
Department of Agricultural Economics College of Bio-resources and Agriculture
National Taiwan University Master Thesis
中美洲農業部門效率與生產力變動之研究
A Study of Efficiency and Productivity Changes for Agricultural Sector in Central America
勞亦理
Eliseo Arauz Quiroz
指導教授：黃芳玫 博士
Advisor:Fung-Mey Huang, Ph.D.
中華民國 106 年 7 月 July, 2017
i Abstract
The purpose of this study is to identify the main sources of agricultural growth for Central America (C.A.) between 1979 and 2014 using data for six countries of the region.
Two analytical methods are used. The first method is data envelope analysis (DEA), which is a non-parametric estimation. The second one applies a parametric function by using stochastic frontier analysis (SFA) which also includes “exogenous variables”. The results from DEA indicate that agricultural growth in countries of C.A. is driven by technical change (TC) rather than efficiency change (EC). The results from SFA suggest that just upper-middle income (UM) economies are driven by TC (or frontier shift) while lower-middle income (LM) economies in C.A. are mostly driven by EC (or catch-up). The empirical results from the exogenous variables also suggest that the more the use of irrigation, human capital, and specialization of crops, the more efficient a country’s agricultural sector becomes.
Keywords: Agriculture, Efficiency, Productivity, DEA, SFA, Central America
ii Dedication
I would like to dedicate this thesis to my family, my friends, and all those who have believed in me. Also to the memory of those who are no longer in this world, but in my memory and my heart forever.
iii
Acknowledgement
I would like to give thanks to my advisor: Professor Fung Mey-Huang (Ph.D.), who contributed and helped me during the process of writing this thesis. Thanks to all my professors and staff of the Department of Agricultural Economics at NTU for their guidance, knowledge, and hospitality.
Special thanks to my family and lovely mother: Linda Quiroz, for her continuous support.
Furthermore, to my friends, classmates, and colleges for their academic support during the course of living, studying, and writing this thesis.
iv
Table of Contents
Abstract ... i
Dedication ... ii
Acknowledgement ... iii
Chapter 1 Introduction ... 1
1.1 Introduction and motivation ... 1
1.2 Objectives of this study ... 4
Chapter 2. Overview of Central America Agriculture ... 5
2.1 Economies ... 5
Chapter 3. Literature Review ... 10
Chapter 4. Data and Methodology ... 12
4.1 Data Envelope Analysis (DEA) ... 13
4.1.1 Malmquist productivity equation ... 15
4.1.2 Decomposition of the Malmquist index ... 16
4.2 Stochastic Frontier Analysis (SFA) ... 18
4.2.1 Stochastic Frontier Model ... 19
4.2.2 Empirical Stochastic Frontier Model ... 23
4.2.3 Elasticities ... 24
4.3 Data Sources and Variables ... 24
4.3.1 Output ... 25
4.3.2 Inputs ... 26
4.3.3 Exogenous Variables ... 31
Chapter 5. Results and Discussion ... 36
5.1 Data Envelopment Analysis (DEA) ... 36
5.1.1 TFP and its components ... 36
5.1.2 Country Analysis of TFP and its components ... 42
5.2 Stochastic Frontier Analysis (SFA) ... 48
5.2.1 Parametric model ... 48
5.2.2 Elasticity Estimation ... 50
v
5.2.3 Hypothesis Test ... 52
5.2.4 TFP and its component ... 53
5.2.5 Country Analysis of TFP and its components ... 57
5.2.6 Efficiency Effect ... 59
Chapter 6. Conclusion ... 61
6.1 Study Approach ... 61
6.2 The Results ... 61
6.2.1 Exogenous variables ... 63
6.3 Policy implication ... 64
6.4 Further research ... 65
References... 66
Appendix ... 69
vi
List of Figures
Figure 2.1 Map of Central America... 5
Figure 2.2 Composition of population and poverty line in Central America ... 6
Figure 2.3 Employment share in Central America ... 7
Figure 2.4 Sector value added (% GDP) ... 8
Figure 2.5 Agricultural share as percentage of GDP ... 9
Figure 4.1 Technical Efficiency from an Output Orientation. ... 14
Figure 4.2 Malmquist Productivity Indices ... 17
Figure 4.3 Output and Input growth (1979-2014) ... 30
Figure 5.1 Cumulative TFP change and its components (DEA) ... 40
Figure 5.2 Guatemala cumulative TFP change and its components (DEA) ... 43
Figure 5.3 Honduras cumulative TFP change and its components (DEA) ... 44
Figure 5.4 El Salvador cumulative TFP change and its components (DEA) ... 45
Figure 5.5 Nicaragua cumulative TFP change and its components (DEA)... 45
Figure 5.6 Costa Rica cumulative TFP change and its components (DEA) ... 46
Figure 5.7 Panamá cumulative TFP change and its components (DEA) ... 47
Figure 5.8 Cumulative DEA & SFA TFP change and its components, 1979-2014 ... 56
Figure 5.9 Low-middle income economies cumulative TFP change and its components (SFA). ... 57
Figure 5.10 Upper-middle income economies cumulative TFP change and its components (SFA) ... 58
vii
List of Tables
Table 4.1 Primary Crops ... 26
Table 4.2 Land Use for Agriculture... 27
Table 4.3 Machinery use for Agriculture ... 28
Table 4.4 Labor in Agriculture ... 28
Table 4.5 Fertilizer in Agriculture ... 29
Table 4.6 CO2 in agriculture ... 32
Table 4.7 Human Capital ... 33
Table 4.8 Area equipped for irrigation ... 33
Table 4.9 Rural Population ... 34
Table 4.10 Life Expectancy ... 34
Table 4.11 Herfindahl-Hirschman index ... 35
Table 5.1 Annual mean of TFP change and its components, 1980-2014 (DEA). ... 36
Table 5.2 Summary description TFP change and its components, 1980-2014 (DEA) ... 37
Table 5.3 TFP change and its components by period (DEA) ... 38
Table 5.4 Summary description of TFP change and its components, 1990-2014 (DEA) ... 39
Table 5.5 Summary description of technical efficiency, 1979-2014 (DEA) ... 40
Table 5.6 Summary description of TFP change and its components by decade, 1980-2014 ... 41
Table 5.7 Maximum-likelihood estimates of the translog production frontier model... 49
Table 5.8a Summary elasticities of inputs by country, 1979-2014 ... 50
Table 5.8b Summary elasticities of inputs by country, 1990-2014 ... 51
Table 5.9 Summary elasticities of inputs by period ... 51
Table. 5.10 Test of Hypothesis for Parameters of the Stochastic Frontier and Inefficiency Model for the agricultural sector in Central America... 53
Table 5.11 Summary description of TFP change and its components, 1979-2014 (SFA) ... 54
Table 5.12 TFP change and its components by period (SFA) ... 55
Table 5.13 Summary description of technical efficiency, 1979-2014 (SFA) ... 55
1
Chapter 1 Introduction
1.1 Introduction and motivation
Technical change (TC) and efficiency change (EC) are two important sources of agricultural productivity growth. Because the rapid growth of world population leads to an increasing demand of food, many countries have set their goal to increase the agricultural sector’s productivity. To improve agricultural productivity, new varieties of seed, advanced equipment, and new methods of production have been implemented, which have improved the agricultural capability of some countries to face food insecurity and increase competitiveness among regions. However, the improvements of agricultural productivity deal with the effects of environmental and social conflicts which could hinder the best performance of the sector in Central America (C.A.).
In order to identify the main sources of agricultural growth in C.A. and understand the region’s agricultural evolutions, we analyze the total factor productivity and their two driving engines: efficiency change and technical change from 1979 to 2014. This analysis may grant an opportunity to introduce appropriate national agricultural policies according to the heterogeneity of the region. Two approaches are used: the data envelope analysis (DEA) and the stochastic frontier analysis (SFA).
This study contributes to the existing literature on different fronts. First, we focus on the analysis of agricultural productivity growth in C.A. Previous studies examined only selected countries in this region which limit the capacity for analysis and the comparability across countries. Second, this study extends the previous analysis by using data in more recent years. Much of the existing literature analyzed the agricultural productivity growth in
2
C.A with very short periods of time. To understand the development of productivity in current periods and to contrast this development of productivity with some recent economic and social phenomena is important. Third, we have classified the countries according to their income level and identified possible associations between the driven factors of productivity and the level of income of the countries. In studies of productivity in Latin America, there is no evidence that shows advantages of rich countries in the absorption of technology, or evidence suggesting that poor countries compensate for the lack of technology by implementing more efficient production mechanisms. Fourth, it incorporates the effect of exogenous variables in the analysis. The incorporation of exogenous variables in the model allows us to analyze the efficiency of productivity taking into account the environmental, social, economic or infrastructure differences among the countries of the region. Finally, two different approaches are used to analyze the growth pattern of the region. In most studies the researchers decide to use a single method of productivity analysis, especially to avoid confusion in the results, however, this study allows to evaluate the productivity in the region from 2 different approaches, taking into account the advantages of one study over another.
According to the World Bank (WB), there are 5% of the population in C.A. living with less than $1.90 per day, and 13% living with less than $3.10 per day in 2014. Because the agriculture sector represents an opportunity of subsistence for the rural area and poor population, improving agricultural productivity is a potential solution to increase their welfare. In 2014, the agricultural share of GDP in C.A was around 11%, but the proportion varied substantially between countries with low-middle income (LM) and countries with
3
upper-middle income (UM) countries.1 The agricultural share of GDP in the LM income countries is usually higher than the proportion of the UM income countries.
A closer diagnostic reveals that agricultural production for the region has grown considerably since 1979, without any substantial change concerning its agricultural structure, which is mostly concentrated in the production of a few crops, including: sugar cane, banana, maize, plantain, coffee, and cotton.
The results of this study suggest that countries in C.A. might still improve their efficiency. In addition, the region experimented a persistent growth in agricultural productivity, which is mainly driven in middle-upper incomes economies by TC.
Interestingly, the agricultural growth of countries with middle-lower income is driven by EC instead of TC.
The remainder of this thesis is organized as follows. In chapter 2, I will briefly introduce the essential aspects of the economy of C.A. in general focusing on agriculture of this region, and the theory used in the analysis of productivity. Following the theoretical chapters, the data and methodology will be described. The last two chapters contain the results and conclusion, followed by suggestions for further investigation.
1 According with the World Bank classification in 2014 the low-middle income countries are those with a Gross National Income (GNI) per capita between $1,046 and $4,125; the upper middle income are those with a GNI between $4,126 and $12,735.
4 1.2 Objectives of this study
The objectives of this study are to analyze the productivity growth of the agricultural sector, to identify the driving factors of agricultural productivity growth, and to examine the effect of socioeconomic variables on agricultural productivity growth.
5
Chapter 2. Overview of Central America Agriculture
2.1 Economies
Central America (C.A.) is a region in the middle of the American continent with a land area of approximately 507,966 square kilometers. The region consists of seven countries:
Guatemala, Honduras, El Salvador, Nicaragua, Costa Rica, Panamá and Belize. In this study Belize is excluded due to the unavailability of information. C.A. is considered an important platform for commerce; bordered by Mexico to the north, and Colombia to the southeast, and between the Caribbean Sea and the Pacific Ocean, this region have a highly coveted strategic (Figure 2.1).
Source: Geology.com
Figure 2.1 Map of Central America
6
The total population of the six countries analyzed in this study in 2014, according to the World Bank, was 44,723,934 persons. Among those, 44% are considered rural population and the remaining are urban population.
In terms of population growth, C.A experienced a nearly 2% growth in its population, increasing from 22.37 million people in 1979 to 44.72 million people in 2014. (See Figure 2.2).
Source: World Bank
Figure.2.2 Composition of population and poverty line in Central America.
In spite of the increase in population, the average percentage of employment in agriculture decreased (Figure 2.3). The share of the total employment in agriculture between 1979 and 2014 was around 30%. In 2014, the percent of the population engaged in agricultural activities was below 25%, with significant differences among countries. It is worthwhile mentioning that, Honduras had the highest participation rate of people working
0.4 0.9 1.4 1.9 2.4
0 10 20 30 40 50
1979-1989 1990-1999 2000-2009 2010-2014
Poverty Line Millions
Total Population Millions
Note: $ (Purchasing power parity of 2011)
Composition of Population in Central America
Rural Population Urban Population
Poverty headcount ratio at $1.90 a day Poverty headcount ratio at $3.10 a day
7
in agriculture (37%) while Costa Rica had the lowest participation rate (14%). According to estimates of the FAO (2014), almost half of the rural population of Latin America (L.A.) is poor and one-third live in extreme poverty.
Source: World Bank
Figure 2.3 Employment share in Central America.
Because farming is an important way of subsistence for the families living in rural areas, increasing productivity for agricultural sector means, at the micro scale, improving the life of families in poverty and, at the macro scale, increasing the comparative advantage and opportunities to compete with industrialized and advanced economies. Therefore, this study is important from a political perspective as the government desires to improve the welfare of farmers and to increase the competitiveness of the agricultural sector.
38.0 28.2 26.5 25.9
19.3
22.3 20.9 18.9
42.7 49.6 52.6 55.2
0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 90.0 100.0
1979-1989 1990-1999 2000-2009 2010-2014
Percentage
Year
Employment share in Central America
Employment in services (% of total employment) Employment in industry (% of total employment) Employment in agriculture (% of total employment)
8
Source: World Bank
Figure.2.4 Sector value added (%GDP)
Since 1980, the agricultural sector in C.A. has shown dramatic changes in terms of Gross Domestic Product (GDP)’s contribution. According to the data released by the World Bank (WB, 2014), the average share of agricultural value added on GDP in Central America decreased from 16% in 1980’s to 11% in 2014 (Figure 2.4).
It is notable that the agricultural sector’s share of GDP varies between countries considerably. In Panamá, for instance, the value was below 4% while in Nicaragua the value was around 19% (Figure 2.5).
16.3 15.6 11.8 11.0
26.1 26.7 26.7 26.2
57.8 57.7 61.5 62.8
0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 90.0 100.0
1979-1989 1990-1999 2000-2009 2010-2014
Percentage value added
Year
Contribution to GDP in Central America
Services (% of GDP)
Industry (% of GDP)
Agriculture (% of GDP)
9
Source: Food Agricultural Organization Figure 2.5 Agricultural share as percentage of GDP
Panamá’s economy is mainly based in the service sector. According to the International Labor Organization (ILO, 2014), the percentage of labor working in service represents around 65% of the employment; while the employment in agriculture represents around 16.7% in 2014. Nicaragua has the highest percentage of the population engaged in agriculture. According to the ILO, the labor in agriculture represented around 32.2% of total employment in the region.
The Economic Commission for Latin America and the Caribbean (ECLAC) appointed significant advances in the labor sector in C.A. these countries however, also show that structural weaknesses persist, especially due to the large informality of the labor market.
0 5 10 15 20
Guatemala Honduras El Salvador Nicaragua Costa Rica Panama
%
Country
Agricultural share as percentage of GDP in Central America, 2014
Agricultur e, value added (%
of GDP)
10
Chapter 3. Literature Review
Improving productivity of agriculture can increase agricultural production and minimize the cost incurred by farms. Bravo-Ortega, C. (2004) determined several elements that may improve efficiency and productivity in agriculture. In addition to this study, Adelman et al. (1988) identified different patterns in agriculture among regions. This result suggests that some elements mentioned by Bravo-Ortega may have different effects across countries and regions. In 1970 Hayami. et al. (1970) studied the difference in productivity among countries; however, this study was not focused on Latin America and did not take account of the singularities among regions.
To investigate in detail the sources of growth in L.A., Dias, A. et al. (2010) identified some specific characteristics of the agricultural sector. Ludena, C. et al. (2010) extended this study and applied a different measure as suggested by Fulginiti, L. (1998) using two important components including efficiency change and technical change for productivity growth.
Based on the finding of previous researchers, improving the productivity of agriculture may have several positive effects on economies. Although the results extend previous information about productivity analysis, little of them account for the possible implication of inclusion of environmental variables for productivity change in C.A. This knowledge void warrants further investigation and exploration.
There are few studies considering the countries in C.A. for agricultural productivity.
The results from Coelli (2005) on agricultural productivity for 93 countries (including five countries from C.A), shows that the main source of growth was TC and there is no evidence
11
that suggests technological regression on agriculture. These findings can also be applied to C.A with an exception for Nicaragua, where the growth was driven mainly by EC. Ebata (2014) conducted an analysis of 14 countries of C.A. and the Caribbean and found that TC was the most important source of growth in agriculture. For those countries in C.A., the growth in agriculture was driven by TC as well with exceptions in some periods.
A recent study by Ludena (2010) about agricultural productivity growth in Latin America and the Caribbean (LAC) from 1967 to 2010, shows that the highest growth within the region occurred during the 1990’s and 2000’s, as result of increases in efficiency and the adoption of new technologies. This result is supported by Diaz, A. (2010) who found that the performance of total factor productivity (TFP) for Latin America was better in the last decade of his study (1990-2000).
To the best of our knowledge, the previous studies found that Costa Rica has the highest TFP growth of the region, and Panamá has the lowest one. Overall, these studies highlight patterns of growth for the agricultural sector and provide important insights concerning the main sources of growth.
12
Chapter 4. Data and Methodology
Normally productivity in economic is a ratio of outputs to inputs. Sometimes this ratio involves a basic quotient of a single dependent and an independent variables, however, in most cases, constructing a productivity measure requires the aggregation of multiple inputs.
In such cases, the aggregation of all the factor of production may provide a better diagnosis of the reality than when we just partially analyze the productivity.
In this analysis, we measure the TFP in C.A. by using two alternative methods to estimate frontiers: data envelope analysis (DEA) and stochastic frontier analysis (SFA).
DEA is considered a non-parametric approach that among its advantages has the fact of not requiring information about prices. In this approach the TE is not measured according to the average performance of the companies or countries but in relation to the maximum performance done. Because we also have a particular interest to analyze the efficiency based on the average performance we use SFA which is a parametric method that in this study also includes exogenous variables that could affect the efficiency of a country.
For missing values, we regress the variables on time by using different functional forms (i.e., linear, quadratic and cubic) and select the model with the best R-square to obtain the predicted values. We replace the missing values with its predicted one.2
2 Percentage of predicted values :
Employment: 44%, Machinery: 28 %, Fertilizer: 0.01%, Land: 0%, and Primary Crops: 0%.
13 4.1 Data Envelope Analysis (DEA)
First, we analyze the efficiency by using the non-parametric frontier Malmquist index which is an alternative to the SFA approach. This method uses a linear programming to construct an envelope frontier which considers all the countries in the sample and considers that any possible deviation from frontier is as a result of inefficiency.
DEA is based on the Farrel (1957) theory of efficiency measurement method and was introduced by Charnes et al. (1978). This method can be either input or output oriented. In the case of input-oriented, the frontier is defined by reducing inputs while holding output quantity constant. The output-oriented method, conversely, seeks to increase output bundle with inputs level held constant.
In this analysis, we use a variable return to scale (VRS) output-oriented DEA involving 6 countries and 216 observations over a 36-years period with one dependent variable and four independent variables.
In our case, the output orientation is considered more in agriculture farms and countries that do not have particular orders to fill output quantities, and also because farmers have more control over inputs than over output.
We can illustrate the output-oriented measures by using a case where one input (𝑥) and two output (𝑞1and 𝑞2 ) are used (Figure 4.1). The measure of technical efficiency (TE) is represented in by OA/OB, which is the difference between the observed point of production and the point on the production possibility frontier (PPF) which is defined by GG’. Normally researchers consider more appropriate to use the assumption of constant return to scale (CRS)
14
especially when financial constraint, imperfect competition, etc., may not affect the countries.
However, as Coelli et al. (2005) mention, this seems less realistic. Therefore, we use VRS instead of CRS. Note that A is operating in the area bounded by the PPF, at this area any country operation is consider inefficient.
Source: Coelli et al. (2005)
Figure 4.1 Technical Efficiency from an Output Orientation.
By using linear programming for each country, we can calculate the output-oriented DEA.
𝑚𝑎𝑥
∅𝜆∅
st
−∅q𝑖 + Qλ ≥ 0 −∅q𝑖 + Qλ ≥ 0 x𝑖 + Xλ ≥ 0 I1′ λ = 1λ ≥ 0 (1) A
B
O G’
G
𝑞1 𝑥1
⁄ 𝑞2
𝑥1
⁄
15
where 1 ≤ ∅ < ∞, and ∅ − 1 is the proportional increase in output that an i-th country, can achieve by held input constant. ∅ is a scalar, and λ is a I × 1 vector of constants. The resulted TE, which is defined by 1⁄∅, is the score reported by DEAP Version 2.1.3
The main purpose of this stage is to measure the changes of TFP and decompose it into TC and EC. This study use Färe et al. (1994) output-based Malmquist productivity change index).
4.1.1 Malmquist productivity equation
The Malmquist productivity equation is the following
𝑚
𝑜(𝑞
𝑠, 𝑞
𝑡, 𝑥
𝑠, 𝑥
𝑡) = [𝑚
𝑜𝑠(𝑞
𝑠, 𝑞
𝑡, 𝑥
𝑠, 𝑥
𝑡) ∗ 𝑚
𝑜𝑡(𝑞
𝑠, 𝑞
𝑡, 𝑥
𝑠, 𝑥
𝑡)]
12= [(
𝑑𝑜𝑠(𝑥𝑡,𝑞𝑡)𝑑𝑜𝑠(𝑥𝑠,𝑞𝑠)
) ∗ (
𝑑𝑜𝑡(𝑥𝑡,𝑞𝑡)𝑑𝑜𝑡(𝑥𝑠,𝑞𝑠)
)]
1 2
(2)
This equation compares two points at time 𝑠 and 𝑡 and can be decomposed as EC and TC. The first distance function with technology in period 𝑡, 𝑑𝑜𝑠(𝑥𝑡, 𝑞𝑡), is defined as the references; it measures the maximal proportional change in output required to make (𝑥𝑡, 𝑞𝑡) achievable in relation with technology in period 𝑡. The second distance function in the denominator 𝑑𝑜𝑠(𝑥𝑠, 𝑞𝑠) measures the reciprocal proportional expansion of output vector 𝑞𝑠 given the input vector 𝑥𝑠.
3 Data Envelopment Analysis (Computer) Program.
16
Similar analysis for the second term in the bracket can be applied when we use technology in period 𝑠 as the reference
4.1.2 Decomposition of the Malmquist index
Färe, et al. (1992a, 1992b) rearranged the Malmquist productivity index as follows:
𝑇𝑜𝑡𝑎𝑙 𝐹𝑎𝑐𝑡𝑜𝑟 𝑃𝑟𝑜𝑑𝑢𝑐𝑡𝑖𝑣𝑖𝑡𝑦 𝐶ℎ𝑎𝑛𝑔𝑒
=
𝑑𝑜𝑡(𝑥𝑡,𝑞𝑡)𝑑𝑜𝑠(𝑥𝑠,𝑞𝑠)
[(
𝑑𝑜𝑠(𝑥𝑡,𝑞𝑡)𝑑𝑜𝑡(𝑥𝑡,𝑞𝑡)
) ∗ (
𝑑𝑜𝑠(𝑥𝑠,𝑞𝑠)𝑑𝑜𝑡(𝑥𝑠,𝑞𝑠)
)]
1
2
(3)
Where,
The first expression illustrates the efficiency corresponding to the countries with best practices in the sample. Therefore, the performance captured by this component can be interpreted as the catching up effect.
𝐸𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑐𝑦 𝐶ℎ𝑎𝑛𝑔𝑒
= 𝑑
𝑜𝑡(𝑥
𝑡, 𝑞
𝑡)
𝑑
𝑜𝑠(𝑥
𝑠, 𝑞
𝑠) = 𝑞
𝑡/𝑞
𝑐𝑞
𝑠/𝑞
𝑎The second expression inside the brackets can be interpreted as TC, and this measures the shift of the frontier over the period of study. Fare et al. (1994), considers any improvement of this component as innovation by the country.
𝑇𝑒𝑐ℎ𝑛𝑖𝑐𝑎𝑙 𝐶ℎ𝑎𝑛𝑔𝑒
= [
𝑑𝑜𝑠(𝑥𝑡,𝑞𝑡)𝑑𝑜𝑡(𝑥𝑡,𝑞𝑡)
∗
𝑑𝑜𝑠(𝑥𝑠,𝑞𝑠)𝑑𝑜𝑡(𝑥𝑠,𝑞𝑠)
]
1
2
= [
𝑞𝑡/𝑞𝑐𝑞𝑠/𝑞𝑎
×
𝑞𝑡/𝑞𝑐𝑞𝑠/𝑞𝑎
]
1
2
(5)
(4)
17
To confirm a positive or negative productivity change will depend on the value of the TFP for any country. A TFP greater than one implies that a country has a positive change in productivity, while a TFP value lower than 1 implies a negative change. However, if the country exhibits the same productivity in respect the previous period, then the TFP is equal to 1, implying that the country TFP has not changed.
The previous decomposition is illustrated in Figure 4.2 by considering a case of constant return to scale, where one input and one output are used.
Source: Coelli et al. (2005)
Figure 4.2 Malmquist Productivity Indices.
In Figure 4.2 a country operates at two different points (D and E) in time (𝑠 and 𝑡). In both cases, the production is consider inefficient because the country is not operating in the frontier but below the technology for the corresponding period.
𝑞
E
D
Frontier in period 𝑠
0 𝑥
𝑞𝑐
Frontier in period 𝑡
𝑞𝑏 𝑞𝑡
𝑞𝑎 𝑞𝑠
𝑥𝑠 𝑥𝑡
18
In addition to this method, we use the SFA approach including exogenous variables to verify if the results are similar or differ significantly with the inclusion of exogenous variables and the application of a different approach. It’s important to mention that we do not pretend to compare the results of both method; if that were the case, we would have to regress the efficiency score of DEA upon the exogenous variables that we want to analyze, which is also known as second stage analysis of DEA or exclude from the SFA model the exogenous variables, so the results can be as comparable as possible.
4.2 Stochastic Frontier Analysis (SFA)
The stochastic frontier model (SFM) was proposed by Aigener et al. (1977), and by Meeusen and Vanden Broeck (1977) simultaneously. This method is regularly used for the purpose of efficiency analysis, especially because it takes into account possible noise in the information and measurement errors ( 𝑣𝑖𝑡− 𝑢𝑖𝑡).
In general, this method uses two approaches: Ordinary Least Square (OLS) and Maximum Likelihood method (ML). The first approach assumes that all the countries are efficient because the error term(𝜀𝑖𝑡) doesn’t consider technical inefficiency (TI), which is given by a non-negative random term denoted as 𝑢𝑖𝑡.
The second approach, unlike the first, consider technical inefficiency because the error term has two separate components (𝜀𝑖𝑡 = 𝑣𝑖𝑡− 𝑢𝑖𝑡).
The component 𝑢𝑖𝑡 represents the unobserved variables such as climate, quality of land, or other exogenous elements not defined in the production function (these variables are
19
associated with technical inefficiency). This component is non-negative and assumed normally distributed with distribution pattern truncated normal, half-normal, or exponential (half normal for this study). The component 𝑣𝑖𝑡 represents the random shock variable which have mean value (𝜇𝑖) equal to 0, variance constant or N (0, 𝜎𝑣2), and normally distributed, where 𝑢𝑖𝑡 and 𝑣𝑖𝑡 are distributed independent of each other.
Although it is true that this method has more advantages compared to DEA, it requires specifying a functional form that involves the use of econometric techniques, especially with the most functional forms.4
For this analysis, we use the translog functional form, a widely used functional form in the study of agricultural productivity which is more flexible than the linear and Cobb- Douglas functions.
4.2.1 Stochastic Frontier Model
The functional form of a translog production model when the ML method is utilized in SFA can be expressed as:
ln(𝑌𝑖𝑡) = 𝛽0+ ∑ 𝛽1𝑙𝑛𝑋𝑖𝑡+1
2∑𝑛𝑖=1𝛽𝑖𝑖𝑙𝑛𝑋𝑖𝑡2 + ∑𝑛𝑖=1∑𝑛𝑗≠𝑖=1𝛽𝑖𝑗𝑙𝑛𝑋𝑖𝑙𝑛𝑋𝑗 + 𝑣𝑖𝑡− 𝑢𝑖𝑡
𝑛𝑖=1
(6)
where 𝑌𝑖𝑡 denotes output 𝑋𝑖𝑡 = inputs variables
4 Second-order flexible forms.
20
𝛽0, 𝛽𝑖, 𝛽𝑖𝑖= the unknown parameters to be estimated. 𝑣𝑖𝑡= ramdom variables associated with disturbance in production; 𝑢𝑖𝑡=country specific/social economic characteristics. Where the technical inefficiency effects can be specified and defined as:
𝑢𝑖𝑡 = 𝑑0+ 𝑑1𝑧𝑖𝑡+ 𝑑2𝑧𝑖𝑡+ ⋯ (7) where the z’s are social-economic variables which explain inefficiency and 𝑖𝑡= i-th country in the t-th period.
The ML method is similar to the OLS method, with the difference that OLS assumes that all the countries are efficient, and then the random error of the OLS (𝜀𝑖) is equal to the random variables of the ML, that is 𝜀𝑖𝑡 = 𝑣𝑖𝑡.
In the ML method, the error term has two components. According to Coelli et al.
(1998), the first component 𝑣𝑖𝑡 is assumed to be iid N(0,𝜎𝑣2) random error and independent of the second term (𝑢𝑖𝑡). The second term is considered as a non-negative random variable and assumed to be distributed iid N(𝜂𝑖𝑡,𝜎𝑢2) and truncated at zero. Kumbhakar and Lovell (2003) also incorporate a composed error structure with a two-sided symmetric term and a one-sided component. When the assumptions concerning to the distribution of error term are valid, the ML irrespective to the type of model estimated has many desired large sample (i.e asymptotic) properties, which is also the reason why ML is more popular in the analysis of productivity.
21
Because technological advances often affect the economic relationship among variables, especially with production functions, this model includes a time variable or time trend (𝑡𝑖) that accounts for non-neutral technical change, and a time square term to allow for a non-monotonic technical change; this is
ln(𝑌𝑖𝑡) = 𝛽0+ 𝛽𝑡𝑡 +1
2𝛽𝑡𝑡𝑡2 + ∑𝑛𝑖=1𝛽𝑡𝑖𝑡𝑙𝑛𝑋𝑖𝑡+ ∑𝑛𝑖=1𝛽𝑖𝑙𝑛𝑋𝑖𝑡+1
2∑𝑛𝑖=1𝛽𝑖𝑖𝑙𝑛𝑋𝑖𝑡2 +
∑𝑛𝑖=1∑𝑛𝑗=1𝛽𝑖𝑗𝑙𝑛𝑋𝑖𝑙𝑛𝑋𝑗+ 𝑣𝑖𝑡 − 𝑢𝑖𝑡 (8)
where 𝑡 is a time trend representing technical change;
𝛽: the unknown parameters to be estimated;
𝑣𝑖𝑡 : the random error, assumed to be i.i.d and have N(0, 𝜎𝑣2) distribution, independent of the 𝑢𝑖𝑡𝑠.
And 𝑢𝑖𝑡 the technical inefficiency effect.
The EC and TC of each country can be predicted using previous approaches. In this parametric case, the EC in equation (4) of DEA analysis can be compared directly with equation (9).
𝑇𝐸𝑖𝑡 = 𝐸(exp(−𝑢𝑖𝑡)/𝑒𝑖𝑡) (9)
Where 𝑒𝑖𝑡=𝑣𝑖𝑡− 𝑢𝑖𝑡 can be used to calculate the component of efficiency change. By observing that 𝑑𝑜𝑡(𝑥𝑖𝑡, 𝑞𝑖𝑡)=𝑇𝐸𝑖𝑡 and 𝑑𝑜𝑠(𝑥𝑖𝑠, 𝑞𝑖𝑠)=𝑇𝐸𝑖𝑠 we calculate the efficiency change by dividing the TE in period t by the TE in period s.
We can also define the TE as the ratio resulting between observed production and the production output from the frontier production function.
22 𝑇𝐸𝑖𝑡 = 𝑌𝑖𝑡
exp (𝑋′𝑖𝑡𝛽+𝑣𝑖𝑡) =exp (𝑋′𝑖𝑡𝛽+𝑣𝑖𝑡+𝑢𝑖𝑡)
exp (𝑋′𝑖𝑡𝛽+𝑣𝑖𝑡) = exp(−𝑢𝑖𝑡) (9a)
In the same way, we can estimate TC using the parameter of our ML model. The TC index between periods for each of the countries can be calculated using the following equation:
𝑇𝑒𝑐ℎ𝑛𝑖𝑐𝑎𝑙 𝐶ℎ𝑎𝑛𝑔𝑒 = exp (1
2[𝜕𝑙𝑛𝑞𝑖𝑠
𝜕𝑠 +𝜕𝑙𝑛𝑞𝑖𝑡
𝜕𝑡 ]) (10)
This is, TC is equal to the exponential of the arithmetic mean of the log derivatives of the production function with respect to time using the data for each country in period s and t.
Considering equation (7), the derivative of output respect time can be rewritten as:
𝑇𝑒𝑐ℎ𝑛𝑖𝑐𝑎𝑙 𝐶ℎ𝑎𝑛𝑔𝑒 = 𝑒𝑥𝑝 {1
2[(∑𝑛𝑖=1𝛽𝑡𝑙𝑛𝑋𝑖𝑡+ 𝛽𝑡𝑡∗ 𝑡 + 𝛽𝑡)𝑝𝑒𝑟𝑖𝑜𝑑 𝑠 + (∑𝑛𝑖=1𝛽𝑡𝑙𝑛𝑋𝑖𝑡+ 𝛽𝑡𝑡∗ 𝑡 + 𝛽𝑡)𝑝𝑒𝑟𝑖𝑜𝑑 𝑡]} (10a)
in the i-th country( 𝑖 = 1,2,3, . . ,6) in the t-th period (𝑡 = 1,2,3, . . ,36)
23 4.2.2 Empirical Stochastic Frontier Model
The functional form of translog production model in SFA used in this study is defined as follows:
ln(𝑌𝑖𝑡) = 𝛽0+ 𝛽1ln (𝐿𝑎𝑏𝑜𝑟𝑖𝑡) + 𝛽2ln (𝑀𝑎𝑐ℎ𝑖𝑛𝑎𝑟𝑦𝑖𝑡) + 𝛽3ln (𝐴𝑟𝑒𝑎𝑖𝑡) + 𝛽4ln (𝐹𝑒𝑟𝑡𝑖𝑙𝑖𝑧𝑒𝑟𝑖𝑡) + 𝛽5 (𝑇𝑖𝑚𝑒𝑖𝑡) +1
2[𝛽11ln (𝐿𝑎𝑏𝑜𝑟𝑖𝑡2) + 𝛽22ln (𝑀𝑎𝑐ℎ𝑖𝑛𝑎𝑟𝑦𝑖𝑡2) + 𝛽33ln (𝐴𝑟𝑒𝑎𝑖𝑡2) + 𝛽44ln (𝐹𝑒𝑟𝑡𝑖𝑙𝑖𝑧𝑒𝑟𝑖𝑡2) + 𝛽55 (𝑇𝑖𝑚𝑒𝑖𝑡2)] + 𝛽12ln (𝐿𝑎𝑏𝑜𝑟𝑖𝑡)*
ln(𝑀𝑎𝑐ℎ𝑖𝑛𝑎𝑟𝑦𝑖𝑡) + 𝛽13ln (𝐿𝑎𝑏𝑜𝑟𝑖𝑡)* ln(𝐴𝑟𝑒𝑎𝑖𝑡) + 𝛽14ln (𝐿𝑎𝑏𝑜𝑟𝑖𝑡)*
ln(𝐹𝑒𝑟𝑡𝑖𝑙𝑖𝑧𝑒𝑟𝑖𝑡) + 𝛽15ln (𝐿𝑎𝑏𝑜𝑟𝑖𝑡)* 𝑇𝑖𝑚𝑒𝑖𝑡 + 𝛽23ln (𝑀𝑎𝑐ℎ𝑖𝑛𝑎𝑟𝑦𝑖𝑡)* ln(𝐴𝑟𝑒𝑎𝑖𝑡) + 𝛽24ln (𝑀𝑎𝑐ℎ𝑖𝑛𝑎𝑟𝑦𝑖𝑡)* ln(𝐹𝑒𝑟𝑡𝑖𝑙𝑖𝑧𝑒𝑟) + 𝛽25ln (𝑀𝑎𝑐ℎ𝑖𝑛𝑎𝑟𝑦𝑖𝑡)* 𝑇𝑖𝑚𝑒 + 𝛽34ln (𝐴𝑟𝑒𝑎𝑖𝑡)*
ln(𝐹𝑒𝑟𝑡𝑖𝑙𝑖𝑧𝑒𝑟𝑖𝑡) + 𝛽35ln (𝐴𝑟𝑒𝑎𝑖𝑡)* 𝑇𝑖𝑚𝑒𝑖𝑡 + 𝛽45ln (𝐹𝑒𝑟𝑡𝑖𝑙𝑖𝑧𝑒𝑟𝑖𝑡)* 𝑇𝑖𝑚𝑒𝑖𝑡 + 𝑣𝑖𝑡−𝑢𝑖𝑡 , 𝑖 = 1,2,3, . . ,6 , 𝑡 = 1,2,3, . . ,36 (11)
where ln(𝑌𝑖𝑡) denotes the log of primary crops (1000 mt) 5 in country 𝑖 in year 𝑡.
𝛽0, 𝛽𝑖, 𝛽𝑖𝑖: the unknown parameter to be estimated.
The inputs include in the model, namely,
ln(𝐿𝑎𝑏𝑜𝑢𝑟) : log of total employment in agriculture (1000 persons).
ln(𝑀𝑎𝑐ℎ𝑖𝑛𝑎𝑟𝑦) : log of total machinery used in agriculture (1000 units).
ln(𝐿𝑎𝑛𝑑) : log of total land for agriculture (1000 ha) 6 .
ln(𝐹𝑒𝑟𝑡𝑖𝑙𝑖𝑧𝑒𝑟) : log of total fertilizer used in agriculture (1000 mt) .
𝑣𝑖𝑡= random variables associated with disturbance in production, and 𝑢𝑖𝑡=country specific/social economic characteristics.
𝑢𝑖𝑡 can be specified and defined as:
𝑢𝑖𝑡 = 𝛿0+ 𝛿1𝑐𝑜2𝑖𝑡 + 𝛿2ℎ𝑐𝑖𝑡+ 𝛿3𝑖𝑟𝑟𝑖𝑡+ 𝛿4𝑟𝑝𝑖𝑡+ 𝛿𝑙𝑒𝑖𝑡+ 𝛿6ℎℎ𝑖𝑖𝑡 (12)
5 mt stands for metric tons.
6 ha stands for hectares.
24 where
𝑐𝑜2𝑖𝑡= represent the emission of Co2 in agriculture in country 𝑖 in year 𝑡.
ℎ𝑐𝑖𝑡= human capital index in country 𝑖 in year 𝑡.
𝑖𝑟𝑟𝑖𝑡= irrigation area in country 𝑖 in year 𝑡.
𝑟𝑝𝑖𝑡= rural population in country 𝑖 in year 𝑡.
𝑙𝑒𝑖𝑡= life expectancy in country 𝑖 in year 𝑡.
ℎℎ𝑖𝑖𝑡= Herfindahl-Hirschman index (HHI) in country 𝑖 in year 𝑡.
4.2.3 Elasticities
Based on the coefficient estimates of MLE, the estimated frontier elasticity of each input can be calculated by using equation (13).
The elasticity is expressed as:
𝜀 =
𝜕ln𝑓(x,t)𝜕ln𝑥𝑚
= 𝛽̂
𝑚+ ∑
𝑛≠𝑚𝛽̂
𝑚𝑛ln𝑥
𝑛+ 𝛽̂
𝑚𝑛ln𝑥
𝑚+ 𝛽̂
𝑡𝑚𝑡
(13)4.3 Data Sources and Variables
To investigate the total factor productivity and its two driving engines: efficiency change and technical change from 1979 to 2014 in Central America, this study uses data from a variety of sources: The Food and Agricultural Organization (FAO), International Labor Organization (ILO), International Fertilizer Association (IFA), the United Nations Economic Commission for Latin America and the Caribbean (ECLAC), World Bank (WB), Penn World Table (PWT), country reports, and author estimations.
25 Country Coverage
This study intends to focus on a homogeneous set of countries located in C.A. These countries are somewhat similar in geographic, climatic, and political characteristics. The countries included in this study are: Guatemala, Honduras, El Salvador, Nicaragua, Costa Rica, and Panamá. Belize was omitted from this study due to the limitation of information.
Time Period
This study exam the TFP for the period 1979-2014. Since information about agriculture prior to 1979 is insufficient in C.A.
4.3.1 Output
Primary Crops
Primary crops are according to (FAO, 2014):
“The crops that come from the land without having any indirect processing, cleaning or quality change. These variables are divided into temporary crop which are both sown and harvested during the same agricultural year (sometimes more than once), and permanent crops which are sown or planted once and not replanted after each annual harvest”. (See list of primary crops in table B in appendix)
26
Table 4.1 Summary of Primary Crops, 1979-2014.
Country Tons of Primary
Crops
*Weight Mean Std. Dev. Min Max
Guatemala 20003690.6 40% 20003.4 8975.2 7658.4 39039.4 Honduras 7209948.9 15% 7542.5 2395.0 4985.1 12229.1 El Salvador 5962254 12% 5962.3 1374.9 3690.3 8920.8
Nicaragua 5107003.6 10% 5108.7 1723.5 3145.6 9571.7 Costa Rica 8161927.3 16% 8161.9 2648.0 4574.5 12424.8
Panamá 3366367.1 7% 3366.4 332.8 2805.1 4290.9 Source: Food and Agriculture Organization
*Note: weight= production of primary crops in each country
/
total primary crops.4.3.2 Inputs
This analysis included four inputs: labor, machinery, land, and fertilizer. Details information of these inputs is presented in the following.
Land
The variable land is measured as agricultural area. This variable account for arable land, permanents crops, and pastures. However, in this study, the land under permanent meadow and pasture which is land used permanently for a period of five years or more, is not included.
According to the Food and Agriculture Organization Statistic (FAOSTAT, 2014) the definition of arable land and permanent crops are as the following:
“Arable land is defined as the land under temporary crops (multiple-cropped areas are counted only once) which is all land used for crops, not abandoned land resulting from shifting cultivation, and with a less than one-year growing cycle and which must be newly sown or planted for further production after the harvest.
27
Arable land consists of land temporarily fallow (less than five years), temporary meadows for mowing or pasture, and land under market and kitchen gardens”.7
“Permanent crops is defined as the land cultivated with long-term crops which do not have to be replanted for several years (such as coffee), land under trees and shrubs producing flowers (such as roses), and nurseries (except those for forest trees, which should be classified under “forest”).”
Table 4.2 Summary of Land Use for Agriculture, 1979-2014.
Country *Weight Mean Std. Dev. Min Max
Guatemala 27% 1975.4 224.2 1726 2564
Honduras 23% 1674.9 201.7 1427 2015
El Salvador 11% 851.4 71.0 727 974
Nicaragua 24% 1755.3 338.8 1240 2320
Costa Rica 7% 512.0 17.7 490 547
Panamá 9% 665.4 60.5 552 757.4
Source: Food and Agriculture Organization
*Note: weight=land used in agriculture in each country
/
total land used agricultural.Machinery
According to (FAO, 2014) variable Machinery refers to:
“Number of agricultural tractors in use, which commonly refers to wheel and crawler or track-laying type tractors (excluding garden tractors) employed in agriculture.”
7 Data for “arable land” are not meant to indicate the amount of land that is potentially cultivable.
28
Table 4.3 Summary of Machinery used for Agriculture, 1979-2014.
Country *Weight Mean Sts. Dev. Min Max
Guatemala 15% 4.26 0.145794 3.95 4.54
Honduras 16% 4.77 0.724735 3.2 5.46
El Salvador 12% 3.61 0.4214016 3.25 5
Nicaragua 10% 2.99 0.6571603 2.1 4.29
Costa Rica 23% 6.77 0.4555585 5.9 7.22
Panamá 24% 6.97 1.695099 5.05 9.89
Source: Food and Agriculture Organization
*Note: weight=machinery used in agriculture in each country
/
total machinery used agriculturalLabor
According to International Labor Organization (ILO, 2014) this variable considers the number of labor specifically working in agriculture. The indicator provides information on persons in working age (15-64) who, during a specified short period, were classified in the following categories: a) paid employment, and b) self-employment.
The economic activity is classified according to the main activity of the establishment in which a person worked during a specific period. It does not depend on the specific duties or functions implied by the job but rather depends on the characteristics of the economic unit in which this person works.
Table 4.4 Summary of Labor in Agriculture, 1979-2014.
Country *Weight Mean Sts. Dev. Min Max
Guatemala 37% 1241.9 457.4 84.2 2113.6
Honduras 20% 689.9 290.8 2.8 1043.8
El Salvador 13% 431.8 166.5 4.7 638.3
Nicaragua 17% 577.1 205.7 334.8 1058
Costa Rica 7% 247.4 23.0 172.8 285.1
Panamá 6% 191.5 33.3 154.9 270.2
Source: International Labor Organization
*Note: weight=agricultural labor in each country
/
total agricultural labor29 Fertilizer
According to International Fertilizer Organization (IFO), this variable considers the amount of fertilizer used in agriculture by considering the three most important plants nutrients.
These nutrients are nitrogen (N), phosphate (P205), and potash (K20). In C.A. the use of fertilizer is increasing, especially in the last decade. The use of fertilizer is not limited to these three components alone, but these are the most commonly used.
Table 4.5 Summary of Fertilizer in Agriculture, 1979-2014.
Country *Weight Mean Sts. Dev. Min Max
Guatemala 31% 169.2 62.3 68.1 319.5
Honduras 18% 97.9 77.4 15 291.8
El Salvador 13% 72.0 12.1 53.1 97
Nicaragua 8% 43.1 15.8 17.7 76.6
Costa Rica 23% 126.4 36.2 69.9 199
Panamá 6% 33.2 8.0 23.1 60.1
Source: International Fertilizer Organization
*Note: weight=fertilizer used in agriculture in each country
/
total fertilizer used in agricultural.In Figure 4.3 the output and input growth rate are presented. In general, the decade of 1980 was characterized by slow growth and a lot of ups and downs. In addition, we have included in appendix a graph for each input and output growth over the whole period.
30
Source: prepared by author
Figure 4.3 Output and Input growth (1979-2014)
In the case of the production of primary products, it shows a steady growth over the years. However, we can identify the slow growth of this variable during the first years of the study and between 1998 and 2002. This stagnation is explained by the sudden arrival of the hurricane Mitch in 1998 which was one of the worst storms in the last decades and severely destroyed much of the productive infrastructure of the region.
In terms of employment, abrupt changes were experienced in the results of the war in C.A. However, a steady employment growth can be seen after 1990, although some countries such as Panama experienced a slowdown in agricultural employment.
In the case of machinery and fertilizer, both variables presented steady growth, although the fertilizer presented two important breaks in 2002 and 2009. The sudden growth during these two years is attributed to efforts of some governments for compensating the
-60.00 -40.00 -20.00 0.00 20.00 40.00 60.00 80.00 100.00
1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014
Percentage
Output and Inputs Growth (1979-2014)
Employment Machinary Land Fertilizer Primary Crops
31
effects of climate change that were reflected in the greater prolongation and impact of el niño and la niña phenomena in the region.
Finally in the case of land for agriculture. This variable maintained a steady tendency to growth, however after 2003 this tendency start to decrease. The decreasing use of land for agriculture might be attributed to the expansion of the urban area but also to the abandonment of agricultural activities.
To extend the analysis, we have included in appendix Table C. containing the growth rates of productivity ratios for each of the factors of production used in this study. Taking into account the identity about partial productivities of Hayami and Ruttan (1985)8, and productivity ratios of machinery and fertilizer.
4.3.3 Exogenous Variables
In this study, we have included 6 exogenous variables that reflect differences in the quality of the inputs used in the production model but also variables outside the production that could explain qualitative differences among the countries of the region. The exogenous variables used in this study reflect variances in the quality of inputs used, infrastructure, environmental quality and social development of countries.
8 𝑃𝑟𝑖𝑚𝑎𝑟𝑦𝐶𝑟𝑜𝑝𝑠/𝐿𝑎𝑏𝑜𝑢𝑟 = (𝐴𝑟𝑒𝑎/𝐿𝑎𝑏𝑜𝑢𝑟)(𝑃𝑟𝑖𝑚𝑎𝑟𝑦𝐶𝑟𝑜𝑝𝑠/𝐿𝑎𝑛𝑑)
32 Carbon Dioxide (CO2)
According to FAO (2014), Carbon Dioxide (CO2) is defined as follows:
“Total emissions produced in the different agricultural emissions sub-domains, such as manure management, manure applied to soils or left on pastures, enteric fermentation, rice cultivation, synthetic fertilizers, cultivation of organic soils, crop residues, burning of crop residues, burning of savanna, and energy use.
Carbon Dioxide provides a picture of agricultural contribution to the total amount of greenhouse gas (GHG) emissions. GHG emissions from agriculture consist of two non-CO2 gases, namely methane (CH4) and nitrous oxide (N2O), produced by crop and livestock production and by management activities”. The unit of measure of this variable is gigagrams.
Table 4.6 Summary of Co2 in agriculture, 1979-2014.
Country *Weight Mean Std. Dev. Min Max Guatemala 22% 5877.8 1358.1 3711.98 8457.79
Honduras 19% 4876.4 597.7 3814.34 5954.95 El Salvador 10% 2724.3 258.2 2166.48 3294.49 Nicaragua 23% 5974.4 1186.4 3967.14 7759.27 Costa Rica 14% 3789.4 634.5 2765.79 4750.31 Panamá 12% 3029.0 249.6 2685.42 3585.04
Source: Food and Agriculture Organization
*Note: weight= emission of Co2 in agriculture in each country
/
total Co2 emissionsHuman Capital (HC)
We use the human capital index of Penn World Tables (PWT), which follow a standard approach in the literature, based on the construction of average years of schooling
33
from Barro and Lee (2013), and an assumed rate of return to education based on the estimations of Mincer equation around the world (Psacharopoulos, 1994).
Table 4.7 Summary of Human Capital, 1979-2014.
Country Rank Mean Std. Dev. Min Max
Guatemala 6 1.58 0.1 1.34 1.85
Honduras 3 1.86 0.2 1.60 2.22
El Salvador 5 1.77 0.2 1.41 2.14
Nicaragua 4 1.84 0.2 1.52 2.18
Costa Rica 2 2.33 0.2 1.92 2.60
Panamá 1 2.49 0.2 2.08 2.81
Source: Penn World Table Irrigation (IRR)
This variable is defined as the area equipped with irrigation infrastructure to provide water to the crops. This variable includes areas equipped for partial and full spate irrigation areas,control irrigation, and equipped wetland. The unit of measure is 1000 ha (FAO, 2014).
Table 4.8 Summary of Area equipped for irrigation, 1979-2014.
Country Rank Mean Std. Dev. Min Max Guatemala 1 192.8 100.7 84.0 338.0
Honduras 4 76.4 9.0 66.0 90.0
El Salvador 5 42.3 3.5 36.0 45.2 Nicaragua 2 100.7 53.9 60.0 199.0 Costa Rica 3 89.1 14.8 56.0 103.0
Panamá 6 32.0 2.3 28.0 35.0
Source: Food and Agriculture Organization Rural Population (RP)
According to the World Bank (WB), the rural population refers to people living in rural areas. It is calculated as the difference between total population and urban population.
34
Table 4.9 Summary of Rural Population, 1979-2014.
Country Mean Std.
Dev.
Min Max
Rural Population
*Rural Population
Share Guatemala 56.1 5.8 48.20 65.69 6081414.73 37%
Honduras 53.4 5.4 46.10 62.27 3150716.34 19%
El Salvador 50.0 5.8 41.02 59.05 2439878.17 15%
Nicaragua 46.1 6.8 37.21 56.09 2130258.64 13%
Costa Rica 42.6 7.4 30.61 53.66 1462073.62 9%
Panamá 39.3 8.3 24.09 51.12 1010115.06 7%
Source: World Bank
*Note: rural population share=rural population in each country
/
total rural populationLife Expectancy (LE)
Life expectancy at birth according to the World Bank (2014) refers to:
“Number of years that a newborn infant would live considering that the patterns of mortality throughout its lifetime were to remain the same”.
Table 4.10 Summary of Life expectancy, 1979-2014.
Country Rank Mean Std. Dev. Min Max
Guatemala 6 65.2 4.9 56.8 71.7
Honduras 3 68.1 4.2 58.8 73.1
El Salvador 5 66.0 5.3 56.4 72.8
Nicaragua 4 67.2 5.4 58.1 74.8
Costa Rica 1 76.5 2.1 71.5 79.4
Panamá 2 74.2 2.2 69.9 77.6
Source: World Bank
35 Herfindahl-Hirschman Index (HHI)
The index is named after economists Orris C. Herfindahl and Albert O. Hirschman and is usually used to measure the market concentration. In this study, we use this index to measure the concentration of primary crops in CA.
𝐻𝐻𝐼 = ∑ 𝑠𝑖2
𝑁
𝑖=1
Where 𝑠𝑖 refers to the share of primary crop of country 𝑖 and N is the total number of countries.
An HHI below 100 indicates a very high diversification in production of primary crops.
An HHI below 1,500 indicates a balanced diversification in production of primary crops.
An HHI between 1,500 to 2,500 indicates a moderate concentration in production of primary crops.
An HHI above 2,500 indicates a high concentration in production of primary crops.
We assumed with this approximation that countries with a higher concentration of products tend to be more specialized in the production of this one, thus increasing the efficiency of production.
Table 4.11 Summary of Herfindahl-Hirschman index, 1979-2014.
Country Rank Mean Std. Dev. Min Max Guatemala 2 5239.2 547.4 4207.9 6435.5
Honduras 5 3339.3 337.9 2800.2 3831.3 El Salvador 1 5368.3 706.7 3968.4 6549.2 Nicaragua 3 4892.6 366.0 4200.0 5551.7 Costa Rica 6 2757.0 537.3 2067.0 3799.8 Panamá 4 3637.6 482.7 2886.4 4536.8
Source: calculated by the author.
