Measuring Technical Efficiency in Agricultural Extension Services

Open Access
Author:
Yang, Qiong
Graduate Program:
Agricultural, Environmental and Regional Economics
Degree:
Doctor of Philosophy
Document Type:
Dissertation
Date of Defense:
September 24, 2013
Committee Members:
  • Spiro E Stefanou, Dissertation Advisor
  • Spiro E Stefanou, Committee Chair
  • Stephan J Goetz, Committee Member
  • Edward C Jaenicke, Committee Member
  • Debashis Ghosh, Committee Member
Keywords:
  • agricultural extension
  • stochastic production frontier
  • technical inefficiency
Abstract:
With rapid technological advances in the second half of last century, the increasing dependence on science-based agriculture has placed greater importance on the fast and efficient transfer of the advanced knowledge to farmers. For most of countries, without an efficient agricultural sector, a country is severely constrained in its ability to feed itself and in consuming other goods and its development. Agricultural extension is one of the most common forms of public support to enhance the agricultural productivity. With costs rising, limited resources available, decreasing proportion of agricultural sector in the economy and changes in the prevailing philosophy of the appropriate extent of government intervention, it becomes more imperative to improve the performance of agricultural extension. Our research focuses on quantifying these changes brought up by agricultural extension. Existing literature on this quantification work on agricultural extension is scarce and inconclusive. Measuring the technical efficiency in a variety of fields has achieved rapid progress via the newly adopted econometric toolboxes; however, the corresponding work in agricultural extension failed to keep in pace. One particular aspect that has been overlooked in agricultural extension literature is the possibility of sample selectivity. Given the public goods nature of agricultural extension, one should distinguish between the two production processes that are involved. In the first stage, an intermediate output is produced using only discretionary inputs (i.e.,variable and controllable by the decision making unit). In the second stage, final outcomes are determined by the level of the intermediate output and by environmental (i.e.,fixed) variables. In quantifying the technical efficiency levels present at the farm level, we extend the existing framework on measuring technical efficiency to account for sample selectivity. Stochastic production frontier approach with accommodations for sample selectivity is used to measure technical efficiency and technical efficiency scores of individual farm. In addition, we test for sample selectivity from the statistical causal effect perspective. An empirical study is conducted based on this extended framework and some hypothesis testing of our interest will be done as well. Our study confirms the existence of sample selectivity from both econometrical and statistical approaches. Efficiency scores are calculated based on our econometrical estimation procedure. Comparisons between farms receiving and those not receiving extension visits are made. We find that there is no significant difference in the level of technical efficiency between these two types of farms. However, farms receiving the extension services have higher productivity level than those not receiving extension services. This finding is consistent with the existence of sample selection. Farms' herd behavior, debt/asset ratio and education level are identified as possible sources of sample selectivity. We also reinforce the conclusion of the existence of sample selectivity following causal inference approach. Several matching algorithms are conducted. Similar conclusions are drawn as in the econometric approach. At the macro level, a static game theoretic model with incomplete information adapted from mechanism design is developed to characterize the interactions among all participants in the two production processes. Extension agents are assumed to have up to three different types characterized by productivity levels. The government's objective is to maximize social welfare by setting up a truth-telling mechanism subject to a set of constraints. Equilibrium results and policy implications are discussed. To our knowledge, this is the first attempt to analyze the interactions of the government, extension agents and farms using game theoretic model. A simulation study is conducted to learn the comparative statics, which is based on the two nonlinear equation systems provides us insight into the impact of productivity difference among agents on economic variables of our interest. Based on our study, policy implications can be multi-fold. At the micro level, policy makers should alleviate the possible influence of sample selection in farm's choices of extension services. Several instruments identified in our analysis including educational level, debt status can serve for such a purpose. At the macro level, we establish a solid foundation that government should be less involved in specific production process of extension service. Our findings are consistent with the decreasing resource allocation to agricultural sector and the increasing financial, human resource input by nongovernment organization and many other private sectors. In addition, the different reaction patterns to the changes in relative productivity may guide policy makers to design mechanisms suitable to different types of extension agencies. In particular, the different price reaction curves to the change in relative productivity may be studied so that extension agencies charger different prices to different farm types. In this case, the true demand rather than the demand distorted by sample selectivity can be more accurately identified. Following future research is of our particular interest. Our estimation of stochastic production frontier function is limited due to data availability in several aspects. In particular, farm types are classified as receiving and not receiving extension services. The exploration may go deeper if we can further distinguish between private and public extension service provision. Increasing the number of individual farms in our data set would make it possible to compare the performance of private extension services to that of public extension services.