Does Credit Access Improve Firm Output? Evidence from a Field Experiment in Bangladesh

Does Credit Access Improve Firm Output? Evidence from a Field Experiment in Bangladesh Nusrat Abedin Jimia,b, Subal Kumbakara, Plamen Nikolova Mohammad Abdul Malekc a Department of Economics, State University of New York, Binghamton c Research and Evaluation Division, BRAC b Corresponding author (njimi1@binghamton.edu) Selected Paper prepared for presentation at the 2016 Agricultural & Applied Economics Association Annual Meeting, Boston, Massachusetts, Jul 31-Aug 2. Copyright 2016 by Nusrat Abedin Jimi and Plamen Nikolov. All rights reserved. Readers may make verbatim copies of this document for non-commercial purposes by any means, provided that this copyright notice appears on all such copies. Does Credit Access Improve Firm Output? Evidence from a Field Experiment in Bangladesh Abstract: Poor financial environment of rural developing economy leads to underinvestment and inefficiency of marginal and small-scale farm households. Development interventions for providing improved market access and credit at subsidized interest rate to small farm households are therefore considered as the preconditions in the transformation process of rural agrarian economy. Hence, the question of whether access to resources influences farm households’ production decisions, performance and efficiency is very important. In this paper, we attempt to estimate the impact of a subsidized credit on farm output and efficiency of small and marginal rice farmers of Bangladesh. Using survey data of a field experimental study, we show that relaxing the credit constraint has significant positive impact on farm output and efficiency. On an average, small-scale rice farms with access to subsidized credit are found to be 13% more efficient than farms with no credit access. The increase is 76% on average when we use the randomized access to credit as instrument for farm credit. We also examine the impact heterogeneity of access to credit by rice variety. We find that cultivation of modern Hybrid rice variety is significantly higher (on average 17%) for treatment farm households compare to the control group. However, we do not find much evidence of heterogeneous productivity impact of access to credit by HYV vs. Hybrid rice. Combining the results, we conclude that access to credit is effective in improving the overall output and efficiency of marginal and small-scale rice farm households. Thus, policies enhancing the credit access of marginal farmers are important for sustainable agricultural development of rural developing economy. Introduction Small and informal firms are central source of employment for half or more of the labor force in most developing countries (World Bank, 2004). However, marginal and small-scale farm households in rural developing economy face imperfect financial markets as well as limited resources. This poor financial environment shapes the investment pattern of small farm households and might lead to underinvestment and inefficiency (Townsend 1994, Morduch 1993, D Karlan et. all 2013). Recent causal evidence on microcredit impacts confirms the existence of severe credit constraints of marginal firms in developing economies. Marginal rate of return to capital of small marginal farms is very high compare to the market interest rate. Thus, relaxing the liquidity constraint induces these farms to finance their production and affects farm sales and profit. (McKenzie, Woodruff 2008; Banerjee, Duflo 2013). However, capital or credit constraints alone are not the problem especially in the context of rural agrarian economy. Risk matters and it influences the investment decision and farm income even after relaxing the credit constraint (D Karlan et.all 2013). Therefore, development interventions for providing resource constrained farm households with credit at subsidized interest rate, inputs, improved financial instruments, market access etc. are considered as preconditions for the transformation process of rural developing economy. Given this context, the question of whether small-scale farms allocate limited resources to their best use, whether access to credit and resources influences small farm household’s performance and efficiency etc. are very important. This paper, using survey data of a field experimental study in Bangladesh, aims to estimate the impact of a subsidized credit on farm output and efficiency of small and marginal rice farmers of Bangladesh. Measuring the efficiency of marginal and small-scale farmers and attempts to identifying credit as one of the possible determinants of efficiency has long been the research interest among agricultural economists. Lots of empirical papers attempt to test the effectiveness of subsidized agricultural microcredit on income and efficiency of subsistence and marginal farmers. However, the findings are not conclusive. Also, most of the studies rely exclusively on observational studies, plagued by endogeneity concerns. Thus, there exists a lack of credible estimates of the impact of credit expansion on farm households’ efficiency. This paper attempts to address this gap. We add to the existing literature by using data from a natural field experiment to estimate the causal effect of credit expansion on output and efficiency of small-scale farm households. In particular, we attempt to address two questions: (1) Does credit expansion improve small firms’ output and technical efficiency of small and subsistence level agricultural farm households? (2) Whether there is any heterogeneity in the effect of credit on farm household efficiency (e.g. by adoption of modern rice variety, land size etc.)? To answer the questions, we begin by estimating a simple log linearized Cobb- Douglas production function of rice and get the predicted rice output. Next, we take the difference between actual and predicted output (i.e. the residuals). Then we regress the actual and predicted output difference on treatment variable to estimate the impact of access to credit on efficiency of rice farms. However, as the program take-up rate was around 20%, this estimate gives us the intent-to-treat effect of the credit expansion program- i.e. how access to credit, on average, affects outcome variable of interest. To estimate the impact of credit on efficiency of rice production, we follow the IV approach. We instrument credit by the treatment assignment and get the treatment on the treated effect. We also examine the impact of access to credit on the technology choice or adoption of modern rice variety by farm households and then test the impact heterogeneity of access to credit by rice variety. Specifically, we estimate whether impact of credit access on rice productivity is different for HYV (high yielding variety) vs. modern Hybrid rice. To answer this question, we create an interaction term of modern rice variety (Hybrid dummy) and treatment assignment and check whether the coefficient estimate of this interaction term is statistically significant or not. We control for baseline covariates for all the regressions and the standard errors are clustered at the branch level. Our findings are as follows. Small-scale rice farm households with access to subsidized credit are, on average, significantly more efficient (by 13%) than farms with no credit access. The increase is 76% on average when we use the randomized access to credit as instrument for farm credit. We also find that cultivation of modern Hybrid rice variety is significantly higher (on average 17%) for treatment farm households compare to the control group. However, we do not find evidence of heterogeneous productivity impact of access to credit by HYV vs. Hybrid rice. Combining the results, we conclude that access to credit is effective in improving the overall output and efficiency of marginal and small rice farm households. Credit and firm output Literature The rapid increase in development funding being channeled to microfinance organizations highlights the nexus between credit access and small farm efficiency. Channelizing credit in the resource utilization may not automatically ensure the full potential of output in agriculture and thus the notion of ‘technical efficiency in production’ arises (Taylor et al, 1986a, p.111). Schultz’s “poor but efficient” hypothesis argues that in context of rural developing economy, merely a provision of agricultural credit at subsidized rate may not be effective in achieving higher technical efficiency due to limited investment opportunities. Therefore, attempts to test the effectiveness of subsidized agricultural microcredit on income and efficiency of subsistence and marginal farmers is not new. Economic efficiency at the micro level focuses on the ability of farms to utilize the best available technology and to allocate resources productively. Measurement of technical efficiency of farmers and identifying credit as a possible determinant of efficiency has long been the research interest among agricultural economists. The question is also important from policy perspective as governments of developing countries, with their limited resources and urge to ensure the food security and well-being of poor people, aim to find out the effective ways to improve the productivity and efficiency of the small and subsistence farm households. Findings of empirical studies on the effectiveness of subsidized credit are mixed. Studies in Brazil and Latin American did not find subsidized credit programs as useful for improving traditional farm efficiency (Taylor et. al 1986; Adams 2001). On the other hand, studies in Philippine, West Bengal, Pakistan, Bangladesh, using stochastic frontier analysis, identified credit as an important determinant of farm efficiency and suggest that access to credit has the potential to increase the production and cost efficiency (Edward Martey et. al 2015; Laha Arindam 2013; Zahidul 2011). Some recent studies in Ghana and Philippine have found land ownership and tenure system to be more important for farm efficiency and argue that only credit is not effective in improving farm efficiency (Donkor, 2014; Koirala et. al 2015). Also, most of these studies rely exclusively on observational studies and thus plagued by endogeneity concerns. Although recent causal evidence on microcredit impacts informs economic theory and policy debates about its effectiveness as a development tool, none of the recent evidence addresses credit impacts on firm efficiency (Banerjee, Karlan, Zinman, 2015). Therefore, no credible estimates exist of the impact of credit expansion on firm efficiency and there is gap in the existing literature and we attempt to address this gap in this paper. Rest of the paper proceeds as follows. In Section 1, we describe the credit program and the experimental design. Section 2 presents our estimation approach. Sections 3 discuss the results and Section 4 concludes. 1. Project Background 1.1. BCUP Credit Program Ineffectiveness of conventional microcredit and formal banking systems for the farming community especially the tenant farmers in Bangladesh is prevalent. In this context, BRAC 1 introduced the ‘Tenant Farmers Development Project’ (Borga Chashi Unnayan Prakalpa (BCUP)), with a low-interest revolving fund from the Bangladesh Bank (Central Bank of Bangladesh) under its financial inclusion strategy. The project was initiated in December 2009 (Boro season of 2010) with Tk. 5,000 million (USD 70 million) as revolving loan fund with an interest rate of 5 per cent per month (the rate at which commercial banks can borrow fund from the Central Bank). The fund was given initially for three years with a target of reaching 300,000 farmers with credit within this period. In 2012, Bangladesh Bank approved extension of the project for another three years. 1 BRAC is the largest NGO in Bangladesh and the world. The BCUP program provides a customized credit services to farmers who cultivate land owned by themselves (owner farms) and by others either fully or partially (pure tenant, tenant-owner etc.). Loans are provided at subsided rate of interest which is a flat rate of 10 per cent per year. The effective rate of interest comes to about 15 to 20 per cent on declining balance depending on the mode of repayment of the principal and interest due. The credit limit is $62 to $375 (taka 5,00030,000); duration is 6-10 months; grace period is 1 month and the Instalment is monthly. An important feature of BCUP program is the formation of Village Organization (VO) as the platform for the delivery of services. In the VO, members are grouped in teams of five members, and four to eight teams consisting of 20 to 40 members forming the village level informal tenant farmer association. The VO meets once every month on a fixed day and time which is attended by the BCUP Program Organizer and an agricultural technician. Apart from the discussion of loan proposal and collection of repayment of the instalment dues and the deposit of savings, farmers can get agricultural information, advice from the Agriculture Development Officer (ADO). Households are selected for loan disbursement by few stages of verification. After screening and verification of information, members are formed into small groups. A Village organization (VO) is formed by combining at least three small groups. BCUP is targeted to reach all 484 Upazilas (sub-districts) of Bangladesh in successive phases. By September 2012, it reached to 212 Upazilas. The main objective of the program was to reduce the dependence of tenant farmers on high-cost informal markets for financing their working capital needs, since they do not have access to formal financial institutions for various reasons. By reducing the credit constraints, BCUP program aimed at improving farm productivity and livelihoods of rural small-scale farm households of Bangladesh significantly. 1.2. Experimental Design For the purpose of the evaluation of the BCUP program, a clustered randomized control trial (RCT) design was adopted. Initially, the program identified the potential 40 branch offices to scale up the program in the year 2012. The research team then randomly drew twenty (20) treatment branches for intervention and the rest twenty (20) branches were designated as control branch offices. Six villages were selected randomly from each treatment and control branch among all the villages within the 8 kilometer radius of the branch catchment area. Thus, 240 villages were selected for this study. Then, a census was conducted in 240 villages covering 61,322 households for detecting eligible households. Using this household census information and based on the program eligibility criteria- willing to take loan, maximum land holding limits 33 decimal- 200 decimal etc. 7,563 households were identified as eligible to participate in the BCUP program- 4,228 and 3,335 in the treatment and control areas, respectively. Finally, among these eligible households, a total of 4,331 households were selected randomly for the quantitative baseline survey -2,155 in treatment areas and 2,146 in control areas. The baseline survey was conducted on 2012 and end line survey was conducted in 2014. The program take-up rate was around 20%. 2. Empirical Strategy 2.1 Data This study uses data from a randomized control evaluation design adopted to evaluate the BCUP (Barga Chashi Unnayan Prakalpa) program in Bangladesh. The level of analysis is farm household. Baseline and end line information on economic, demographic variables of farm households; input, output, price and other variables of rice farm land etc. are used. 2.2 Empirical Strategy Our aim is to estimate the impact of credit on technical efficiency2 of rice production. We begin by estimating a simple log linearized Cobb- Douglas production function of rice. 𝑦𝑖 = 𝛽0 + 𝛽1 𝑁𝑖 + 𝛽2 𝐿𝑖 + 𝛽3 𝑀𝑖 + 𝛽4 𝐾𝑖 + 𝜖𝑖 (1) 𝑦𝑖 = Value of rice output of household 𝑖 𝑁𝑖 = Labor cost of household 𝑖 𝐿𝑖 = Land cost of household 𝑖 𝑀𝑖 = Seed, Fertilizer and Pesticide cost of household 𝑖 𝐾𝑖 = Irrigation, Land preparation and other cost of household 𝑖 𝜖𝑖 = Error term i = 1 𝑡𝑜 𝑛 All the variables are from the end line survey data of 2014 and measured in log form. Estimation of the rice production function (equation 1) gives us the 𝑦̂𝑖 - predicted value of rice output. We take the difference between actual and predicted output i.e. the residuals, 2 Technical efficiency is the effectiveness with which a given set of inputs is used to produce an output. A firm is said to be technically efficient if a firm is producing the maximum output from the minimum quantity of inputs, such as labor, capital, land etc. 𝜖̂𝑖 = 𝑦𝑖 − 𝑦̂. ̂𝑖 on 𝑖 Then we run the regression of the actual and predicted output difference 𝜖 treatment variable 𝑍𝑖 𝑚 𝜖̂𝑖 = 𝛾0 + 𝛾1 𝑍𝑖 + ∑ 𝛾𝑗𝑖 𝑋𝑗𝑖 + 𝜗𝑖 𝑗=2 𝑖 = 1 𝑡𝑜 𝑛 (2) 𝑍𝑖 = 1 if treatment household (Access to credit) and 0 otherwise. 𝑋𝑗𝑖 ′𝑠 are baseline covariates j of household i. Our parameter of interest is 𝛾̂1 which shows the impact of access to credit (treatment assignment) on efficiency of rice farms. However, as the program take-up rate was around 20%, this estimate gives us the intent-to-treat effect of the credit expansion program- i.e. how access to credit, on average, affects outcome variable of interest. To estimate the impact of credit on efficiency of rice production, we follow the IV approach and run the following version of regression (2) 𝑚 𝜖̂𝑖 = 𝛾0′ + 𝛾1′ 𝐷𝑖 + ∑ 𝛾𝑗𝑖′ 𝑋𝑗𝑖 + 𝜗𝑖′ 𝑗=2 𝑖 = 1 𝑡𝑜 𝑛 (2′) 𝐷𝑖 = 1 if the household took credit from BCUP and 0 otherwise. We instrument 𝐷𝑖 by 𝑍𝑖 . Thus, 𝛾̂1′ gives the impact of credit on efficiency- treatment on the treated effect. Standard errors are clustered at the branch level. There are two major rice producing seasons in Bangladesh – Boro and Aman3. We refer to both Boro and Aman production by All Season. 3 Boro and Aman are major rice growing seasons in Bangladesh. December-February is the plantation time of Boro and April–May is the harvesting time. Boro production is heavily dependent on irrigation as it covers the dry seasons. Boro production, including the highyield varieties, expanded rapidly until the mid-1980s with the green revolution and rapid increase in use of irrigation. Aman season has We estimate equation (2) and (2′) for All season as well as for each of the seasons separately. 2.3 Heterogeneous Treatment Effects Relaxing credit constraint induces small-scale farms to finance more in production. (McKenzie, Woodruff 2008; Bnaerjee, Duflo 2013). However, capital or credit constraints alone are not the problem especially in the context of uncertainty of rural agrarian economy. Risk matters and it influences the investment decision of farm and farm income even after relaxing the credit constraint (D Karlan et.all 2013). We expect that BCUP credit enables the resource constrained marginal farms to buy fertilizer, seeds and do irrigation at proper time and hence increase their output and productivity. Credit can also induce the small farms to adopt/ produce more Hybrid rice variety which is comparatively more costly (requires more complementary inputs) but also has higher yield on average than HYV rice. Thus credit might have differential impact on the technical efficiency of different rice varieties. To test these hypotheses, at first, we examine the impact of access to credit on adoption of modern rice variety by the following equation plantation time of April- May and harvesting time of November-December. Some rice for this season harvest is sown in the spring through the broadcast method, matures during the summer rains, and is harvested in the fall. 𝑚 𝑌𝑖 = 𝛽0 + 𝛽1 𝑍𝑖 + ∑ 𝛾𝑗𝑖 𝑋𝑗𝑖 + 𝜖𝑖 𝑗=2 𝑖 = 1 𝑡𝑜 𝑛 (3) 𝑌𝑖 = 1 if the farm i produces Hybrid rice, 0 otherwise. 𝑍𝑖 if is the treatment dummy (Access to credit). 𝑋𝑗𝑖 𝑠 are baseline covariates j of household i. Our parameter of interest is 𝛽̂1 . We also estimate impact of credit on Hybrid rice adoption by instrumenting credit 𝐷𝑖 by treatment assignment𝑍𝑖 in equation (3). We then test the impact heterogeneity of access to credit by rice variety by creating an interaction term of modern rice variety (Hybrid dummy) and treatment assignment and run the following version of equation 2 𝜖̂𝑖 = 𝛾0∗ + 𝛾1∗ 𝑍𝑖 + 𝛿1∗ 𝐻𝑦𝑏𝑟𝑖𝑑𝑖 + 𝛿2∗ 𝐻𝑦𝑏𝑟𝑖𝑑𝑖 𝑚 ∗ 𝑍𝑖 + ∑ 𝛾𝑗𝑖′ 𝑋𝑗𝑖 + 𝜗𝑖∗ 𝑗=2 (2′′) ̂2∗ is our parameter of interest. Again we estimate equation (3) and (2′′) for All season as 𝛿 well as for each of the Boro and Aman season separately. We control for baseline covariates and standard errors are clustered at the branch level. 3. Results and Discussion We use data from RCT study for evaluating the BCUP (Barga Chashi Unnayan rakalpa) program in Bangladesh. Sample size of the study is 4301 households. For this paper, we only consider the rice producing farm households. Our sample size is 3752 households. Table 1 shows the baseline characteristics and balancing between treatment and control households. Treatment and control households had significant difference in terms of credit access to other NGO and informal sources. We also did the join significance test and we could not reject the Null hypotheses of no joint significance. For all our regression analysis, we control for Baseline covariates. At first, we estimate the impact of credit access on farms input allocation and adoption of modern Rice varieties. Table 2 summarizes the result. We find that Hybird rice production was on average 17% higher for the treatment rice farms compare to the control farm households. Treatment farm households significantly reduced their input investment for HYV rice and increased input allocation for modern Hybrid rice. The finding was consistent for each of the rice producing seasons. Then, we estimate the impact of access to credit on efficiency of rice farms. Table 3 presents both the ITT and IV estimates. We find that small-scale rice farms with access to subsidized credit are on average 13% more efficient than farms with no credit access. The increase is 76% on average when we use the randomized access to credit as instrument for farm credit. Finally, we examine the impact heterogeneity of access to credit by rice variety (Table 4). However, we only find heterogeneous productivity impact of access to credit for Aman season. Treatment farm households who produced Hybrid rice are found to be 37% more efficient than HYV rice producers. Combining all the results, we conclude that access to credit is effective in improving the overall output and efficiency of marginal and smallscale rice farm households. 4. Conclusion Using survey data of a BCUP program evaluation study of Bangladesh, we estimate the impact of a subsidized credit on farm output and efficiency of small and marginal rice farmers. We show that relaxing the credit constraint has significant positive impact on farm output and efficiency. On an average, small-scale rice farms with access to subsidized credit are more productive than farms with no credit access. We also find that removing credit constraints helps small and marginal farms to reallocate their production inputs and invest more in modern rice variety. Cultivation of modern Hybrid rice variety is significantly higher for treatment farm households compare to the control group. However, there is not much evidence of heterogeneous impact of access to credit by rice variety except the Aman season produciton. Combining the results, we conclude that access to credit is effective in improving the overall output and efficiency of marginal and smallscale rice farm households. Thus appropriate policies targeting at enhancing the credit access of small farms are important for the sustainable agricultural growth and improved efficiency of rural agrarian economy. References: References: Arindam Laha ; 2013. Technical efficiency in agricultural production and access to credit in west Bengal, india: A stochastic frontier approach Ashraf, Karlan, Yin (2009), “Female Empowerment: Impact of a Commitment Savings Product in the Philippines” World Development Vol. 38, No. 3, pp. 333–344, 2010 Banerjee, A.V, Gertler, P.J. & Ghatak, M. (2002). Empowerment and Efficiency: Tenancy Reform in West Bengal. The Journal of Political Economy, 110, 239-280. Banerjee, Duflo (2014), “Do firms want to borrow more? Testing Credit constraints using a directed credit program” Review of Economic Studies (2014) 81 (2): 572-607. Banerjee, Karlan, Zinman, (2015), “Six Randomized Evaluation of Micro Credit: Introduction and Further steps”American Economic Journal: Applied Economics 2015, 7(1): 1–21 http://dx.doi.org/10.1257/app.20140287 Battese, G.E., and Coelli, T.J. 1995. A Model for Technical Inefficiency Effects in a Stochastic Frontier Production Function for Panel Data. Empirical Economics 20. Bayes A and Hossain M (2011). Change in tenancy and labour markets and impact on livelihoods in rural Bangladesh: findings from a longitudinal survey 1988-2007, Paper presented in the conference of the Asian Society of Agricultural Economists (ASAE), Held in Hanoi: Vietnam. BB (2010). Speech of the Governor of BB (Bangladesh Bank) in the reception for the tenant famers on achieving food security and other contributions. Dhaka: Army stadium. Chang, H. Boisvert, R., N., and Hung, L. 2010. Land subsidence, production efficiency, and the decision of aquacultural firms in Taiwan to discontinue production. Ecological Economics 69. D Karlan et. all (2014) “Agricultural decisions after relaxing the Credit and Risk constraints” The Quarterly Journal of Economics 129 (2): 597-652 Duflo, Glennerster, Kremer (December 2006), “Using Randomization in Development Research: A toolkit”, NBER Technical Working paper Series Edward Martey et. all (2015) “Impact of Farmer Mentorship Project on Farm Efficiency and Income in Rural Ghana”Journal of Agricultural Science; Vol. 7, No. 10; Emmanuel Donkor, Victor Owusu, 2015; effects of land tenure systems on resource-use productivity and efficiency in Ghana’s rice industry. African Journal of Agricultural and Resource Economics Volume 9 Number 4 pages 286-299 Friebel G.; Ivaldi, M. and C. Vibes, 2003, Railway (De)Regulation: a European Efficiency Comparison, IDEI report, no. 3 on passenger rail transport, University of Toulouse. Greene, W. H. 1993. "Frontier Production Functions", EC-93-20. Stern School of Business, New York University. Heshmati, S. and Kumbhakar, S.C. 1997. Estimation of technical efficiency in Swedish crop farms: A pseudo panel data approach. Journal of Agricultural Economics 48: 22-37. Hossain M and Bayes A (2009). Rural economy and livelihoods insights from Bangladesh. Dhaka: AH Development Publishing House. Hossain M (1977). Farm size, tenancy and land productivity: An analysis of farm level data. Bangladesh Development Studies 5(3): 285-348. Jonathan Morduch (1995), “Income smoothing and consumption smoothing” Journal of economic perspective, vol 9, No 3 Krishna H. Koirala et all (2015), Farm Productivity and Technical Efficiency of Rural Malawian Households: Does Gender Make a Difference? McKenzie, Woodruff (2008), “Returns to capital in Microenterprises: Evidence from a field Experiment” The Quarterly Journal of Economics 123 (4): 1329-1372 Paul Thangata 2015; The Impact of Agricultural Extension Services on Farm Household Efficiency in Ethiopia Robert M. Townsend (May, 1994),, “Risk and Insurance in Village India” Econometrica Vol. 62, No 3. pp. 539-591 Seyoum, E.T., Battese, G.E., and Fleming, E.M. 1998. Technical efficiency and productivity of maize producers in eastern Ethiopia: a study of farmers within and outside the Sasakawa-Global 2000 project. Agricultural Economics 19. Sharma K.R., Leung. P., Zaleski H.M. 1999. Technical, allocative and economic effiencies in swine production in Hawaii: a comparison of parametric and nonparametric approaches. Agricultural economics 20. Taylor et al, (1986) “Inter-breed relationship of maintenance efficiency to milk yield in cattle.” Anim. Prod. 43:37–61. Uaiene, R.N., and Arndt, C. 2009. Farm household efficiency in Mozambique. Paper presented at the International Association of Agricultural Economist Conference, Beijing, China. Wouterse, F. 2011. Social Services, Human Capital, and Technical Efficiency of Smallholders in Burkina Faso. IFPRI Discussion Paper 01068. Zahidul et.all (2011) “Technical, Economic and Allocative Efficiency of Microfinance Borrowers and NonBorrowers: Evidence from Peasant Farming in Bangladesh” European Journal of Social Sciences – Volume 18, Number 3 Table 1: Baseline Characteristics and Balancing Variables Household Composition Sex of household head (male=1) Age of household head Household size Number of adult members (>16) Number of child (<16) Household head no education Amount of Land Own cultivated land (decimal†) Rented in land (decimal) Rented out land (decimal) Total cultivated land (decimal) Access to credit Loan from other NGO (dummy) Any formal and informal loan (dummy) Output and Input for Rice Production (yearly) All Season production (kilogram) HYV rice (kilogram) Hybrid rice (kilogram) Land (decimal†) Labor days Seed, Fertilizer and Pesticide (taka††) Irrigation, Land preparation and other cost (taka††) Price of Inputs Price of rice§ (kilogram) Price of land (decimal†) Labor Wage§§ (taka††) (1) Control (2) Treatment (2)-(1) t value 0.95 44.45 4.75 3.1 1.65 0.45 0.91 45.07 4.93 3.1 1.83 0.42 -1.68 0.911 1.044 0.0324 1.544 -0.565 38.83 51.31 7.78 90.14 37.38 51.12 8.39 88.5 -0.442 -0.0252 0.517 -0.177 0.07 0.17 0.05 0.11 -1.91 -3.059 1988.56 1550.05 97.39 115.44 51.96 5654.67 6819.5 1999.32 1602.95 79.5 117.98 54.87 5134.75 6311.16 0.0531 0.19 -0.353 0.189 0.627 -0.842 -0.539 15.34 9250.09 16.11 11731.28 1.282 1.428 229.9 276.42 2.387 Joint significance test F statistic = 45.10 § §§ Notes to table 1: †247 decimal=1 hectare. ††$1=80 taka. village level rice price. Village level wage per day. Unit of observation: household. Standard errors (in parentheses) are clustered at Branch level. Sample includes all households surveyed at baseline (2012); for input output section, only rice producing farm households are considered. Informal lenders includes moneylenders, loans from friends/family, and buying goods/services on credit from sellers. Table 2: Effect on Inputs of Rice and Adoption of Modern Rice varieties All Season# (1) (2) Boro season Rice## (3) (4) Aman Season Rice### (5) (6) Adoption of Rice varieties HYV rice (dummy) -0.10 Hybrid rice (dummy) Total Rice Production Inputs (yearly) Land (decimal) Labor days Seed, Fertilizer and Pesticide (taka††) Irrigation, Land preparation and other cost (taka††) HYV Rice Production Inputs (yearly) Land (decimal†) Labor days Seed, Fertilizer and Pesticide (taka†) Irrigation, Land preparation and other cost (taka††) Hybrid Rice Production Inputs (yearly) Land (decimal†) Labor days Seed, Fertilizer and Pesticide (taka††) [0.07] 0.17*** [0.03] -0.09 [0.06] 1.18*** [0.18] -0.047* [0.02] 0.16*** [0.04] -0.276** [0.11] 0.95*** [0.20] 0.01 [0.12] 0.09*** [0.028] 0.039 [0.66] 0.51*** [0.172] -5.40 [10.47] -8.17 [7.20] -62.47 [65.69] -27.32 [47.72] -12.32** [5.64] -8.69 [5.40] -72.22** [33.54] -50.87* [28.74] -7.09 [10.45] -4.64 [4.28] -39.12 [57.02] -25.61 [23.69] -1,466** [613.6] -5,973* [3,531] -1,050*** [317.8] -6,156*** [2,06] -932.2** [366.4] -5,146* [2,878] 542.60 [492.1] -247.10 [660.9] -2829.00 [1,945] -16581.00 [11,50] -540.40 [703.7] -2,983 [3,780] -23.97* [12.73] -14.25 [8.49] -93.67 [66.21] -47.81 [47.05] -19.85** [5.72] -12.41** [5.39] -116.4** [31.78] -72.71*** [26.99] -13.37 [11.18] -5.92 [5.769] -73.79 [71.25] -32.67 [34.25] -1,891** [699.1] -8,553** [3,435] -1,480** [331.9] -8,675** [1,922] -848.00 [532.2] -4,681 [3,626] -4,473** [1,892] -25,167** [11,075] -3,768** [1,840] -22,086** [10,826] -1,164** [447.3] -6,423** [3,268] 8.46*** [2.23] 3.77*** [1.08] 53.70*** [9.20] 25.82*** [5.59] 7.78*** [2.22] 3.78*** [1.14] 45.61*** [9.64] 22.14*** [5.76] 4.05** [1.94] 1.65** [0.63] 22.38* [12.15] 9.10** [4.033] 371.8*** [130.8] 2,820*** [682.0] 434.3*** [155.6] 2,546*** [668.6] 113.9** [43.84] 629.0*** [226.0] Irrigation, Land preparation, other cost (taka††) Observations 791.4*** [225.6] 3,062 5,799*** [1,347] 2498.00 943.0*** [291.5] 2439.00 5,527*** [1,366] 2439.00 167.0* [91.29] 1671.00 922.1 [574.1] 1671.00 Notes to table 2: †247 decimal=1 hectare. ††$1=80 taka. *** p<0.01, ** p<0.05, * p<0.1. Sample includes only rice producing farm households. Column (1), (3) and (5) shows the intent to treat (ITT) effect of the treatment on outcome of interest. Column (2), (4) and (6) shows the impact of credit (Treatment on the Treated) on outcome of interest where credit is instrumented by the treatment assignment. Standard errors (in parentheses) are clustered at Branch level. For all the regression we control for baseline covariates. # We refer to both Boro and Aman production by All Season. These are two major rice seasons and covers most of the rice cultivation in the country. ## Boro is a rice growing season in Bangladesh. December-February is the plantation time of Boro; April–May is the harvesting time. Boro production is heavily dependent on irrigation as it covers the dry seasons. Boro production, including the high-yield varieties, expanded rapidly until the mid-1980s with the green revolution and rapid increase in use of irrigation. ### Aman is a rice growing season in Bangladesh with plantation time April- May and harvesting time November to December. Some rice for this season harvest is sown in the spring through the broadcast method, matures during the summer rains, and is harvested in the fall. Table 3: Effect of Credit on Efficiency of Rice Production All Season Total Rice production (1) (2) Observation 0.13*** [0.05] 0.14*** [0.05] 0.11 [0.07] -0.08 [0.083] 0.76*** [0.27] 0.77*** [0.28] 0.37* [0.23] -0.55 [0.66] 2,474 0.089** [0.04] 0.08* [0.041] 0.03 [0.08] 0.57** [0.24] 0.52** [0.25] 0.116 [0.33] 2,197 0.22* [0.11] 0.26** [0.12] 1.18* [0.61] 1.21* [0.64] 1,635 # Total HYV Total Hybrid Total Traditional Variety Boro Season## Total Rice production Total HYV Total Hybrid Aman Season### Total Rice production Total HYV 2,289 365 441 2,141 278 1,176 Total Hybrid Total Traditional Variety 0.49** [0.18] -0.0004 [0.08] 1.15 [0.71] -0.00312 [0.64] 110 636 Notes to table 3: *** p<0.01, ** p<0.05, * p<0.1. Sample includes only rice producing farm households. Column (1) shows the intent to treat (ITT) effect of the treatment on outcome of interest. Column (2) shows the impact of credit (Treatment on the Treated) on outcome of interest where credit is instrumented by the treatment assignment. Standard errors (in parentheses) are clustered at Branch level. For all the regression we control for baseline covariates. # We refer to both Boro and Aman production by All Season. These are two major rice seasons and covers most of the rice cultivation in the country. ## Boro is a rice growing season in Bangladesh. December-February is the plantation time of Boro; April–May is the harvesting time. Boro production is heavily dependent on irrigation as it covers the dry seasons. Boro production, including the high-yield varieties, expanded rapidly until the mid-1980s with the green revolution and rapid increase in use of irrigation. ### Aman is a rice growing season in Bangladesh with plantation time April- May and harvesting time November to December. Some rice for this season harvest is sown in the spring through the broadcast method, matures during the summer rains, and is harvested in the fall. Table 4: Heterogeneous Effect of Credit on Efficiency Rice Production (HYV vs. Hybrid rice) Variables Treatment (access to credit) Hybrid rice production (dummy) Access to credit x Hybrid Dummy Control for Baseline covariates Observations All season# (1) (2) 0.134*** 0.160*** [0.047] [0.049] 0.141** [0.065] -0.08 [0.068] Yes Yes 2,474 2,038 Boro Season## (3) (4) 0.0897** 0.077** [0.0405] [0.038] 0.19** [0.074] -0.075 [0.079] Yes Yes 2,196 2,182 Aman Season### (5) (6) 0.221* 0.286** [0.109] [0.128] -0.125 [0.128] 0.369** [0.140] Yes Yes 1,635 999 Notes to table 4: *** p<0.01, ** p<0.05, * p<0.1. Sample includes only rice producing farm households. Results show the intent to treat (ITT) effect of the treatment on outcome of interest. Standard errors (in parentheses) are clustered at Branch level. For all the regression we control for baseline covariates. # We refer to both Boro and Aman production by All Season. These are two major rice seasons and covers most of the rice cultivation in the country. ## Boro is a rice growing season in Bangladesh. December-February is the plantation time of Boro; April–May is the harvesting time. Boro production is heavily dependent on irrigation as it covers the dry seasons. Boro production, including the high-yield varieties, expanded rapidly until the mid-1980s with the green revolution and rapid increase in use of irrigation. ### Aman is a rice growing season in Bangladesh with plantation time April- May and harvesting time November to December. Some rice for this season harvest is sown in the spring through the broadcast method, matures during the summer rains, and is harvested in the fall. Appendix Table A.1: Descriptive Statistics and Baseline Characteristics Variables Household Composition Sex of household head (male=1) Age of household head Household size Number of adult members (>16) Number of child (<16) Household head no education Amount of Land Own cultivated land (decimal†) Rented in land (decimal) Rented out land (decimal) Total cultivated land (decimal) Access to credit Loan from other NGO (dummy) Any formal and informal loan (dummy) Output and Input for Rice Production (yearly) All Season production (kilogram) HYV rice (kilogram) Hybrid rice (kilogram) Land (decimal†) Labor days Seed, Fertilizer, Pesticide Cost (taka ††) Irrigation, Land preparation, other cost (taka††) Price of Inputs Price of rice§ (kilogram) Price of land (decimal†) Labor Wage§§ (taka††) § N Mean Std. Dev. Min Max 4301 4301 4,301 4301 4,301 4301 0.93 44.76 4.84 3.10 1.74 0.44 0.25 11.64 1.76 1.34 1.24 0.50 0.00 7.00 1.00 1.00 0.00 0.00 1.00 99.00 20.00 12.00 11.00 1.00 4301 4301 4301 4,301 38.10 51.20 8.09 89.30 51.39 89.30 28.42 97.96 0.00 0.00 0.00 0.00 1003.00 2610.00 426.00 2610.00 4301 4,301 0.06 0.14 0.23 0.35 0.00 0.00 1.00 1.00 3752 3752 3752 3752 3,752 3752 3,752 1994.00 1576.76 88.36 116.73 53.43 5392.21 6562.89 1694.41 1531.38 499.82 102.12 47.71 4972.32 6329.18 30.00 0.00 0.00 4.00 0.00 0.00 0.00 23325.00 23325.00 9126.00 1320.00 575.86 53153.00 67201.00 3752 3752 3752 15.73 2.53 9.38 30.81 10502.59 18081.30 1.00 500000.00 253.38 66.29 100.00 450.00 §§ Notes to table A1: †247 decimal=1 hectare. ††$1=80 taka. village level rice price. Village level wage per day. Unit of observation: household. Sample includes all households surveyed at baseline (2012); for input output section, only rice producing farm households are considered. Informal lenders includes moneylenders, loans from friends/family, and buying goods/services on credit from sellers.