Three essays in agricultural development in Central America: Semiparametric analyses using panel data

Date of Completion

January 2010


Economics, Agricultural




The main goal of this dissertation is to investigate the complex rural sector for two of the poorest countries in Central America, Honduras and Nicaragua. The first essay analyzes the main determinants of farm labor productivity and off-farm labor supply by heads and spouses in rural Nicaragua. A semiparametric approach is used, which controls for biases not only from individual and farm time-invariant characteristics but also from sample selection. A three year balanced panel dataset for 1998, 2001 and 2005 period was used. The main results indicated that when marginal productivity of farm labor goes up, heads and spouses reduce their hours allocated to off-farm activities. Moreover, education, age, remittances, household size, and whether sons and daughters work are found to be related to off-farm labor supply with significant differences between heads and spouses.^ Still focusing on Nicaragua and using the same dataset, the second essay investigates how land endowments and other farm output determinants have contributed to agricultural income. Some key non-observable components (e.g., farmer ability and soil productivity) are controlled for by using parametric and semiparametric methods for panel data. Overall, the differences between the results of the parametric and the semiparametric specifications were small. The value of the partial elasticity of production for land, the key variable of interest, ranged from 0.17 to 0.37, depending on the estimation method used. Moreover, more robust results were obtained under the semiparametric specification and the results from the latter revealed that land tenure, education (especially for women), agricultural training, and community associations play significant roles in increasing farm income.^ The third essay evaluates the impact of the MARENA program using different matching techniques on the Total Value of Agricultural Output (TVAO) in Honduras. The empirical approach is based on the average treatment effect on the treated (ATET) framework and propensity score matching (PSM) methods. Propensity scores are estimated parametrically and semiparametrically. The results suggest that MARENA has had a positive and significant effect on TVAO for all matched samples constructed. In addition, there was no major difference between the average impact calculated from propensity scores derived from parametric and semiparametric models. However, more robust results (i.e., lower standard errors) were found under the semiparametric model compared to the parametric estimation. ^