Three Essays on Evaluation and Measurement in Developing Countries

The overall theme of this dissertation is on evaluation and measurement in developing countries. All three essays make a contribution to the literature--either with the use of a new empirical method for evaluation or a unique or new data set, or the evaluation of a tool or program not yet evaluated, or by evaluating an aspect within the evaluation literature usually neglected. The results of the three essays have policy relevance and can be used to design or modify anti-poverty programs geared towards the social and productive sectors so that they can have a greater impact on poverty. Chapter 2 uses data from small scale potato farmers in Ecuador to examine the impact of the program Plataformas de Concertación on productivity growth. Using propensity score matching combined with a Stochastic Production Frontier model recently introduced by Greene (2010) that corrects for sample selection bias, we disaggregate the yield growth attributable to the program into technological change (TC) and technical efficiency (TE). While the results do not exhibit a clear indication of selection bias, the analysis does show that on average beneficiaries exhibit higher yields than control farmers given the same input levels, but lower TE with respect to their own frontiers. These results suggest that while the program raised the technology gap in favor of beneficiaries, it had a negative effect on TE on the short run. The latter finding is consistent with the notion that beneficiaries enjoyed a significant change in production techniques, but it is very likely that they were still in the "learning by doing" stages at the time the data was collected. In fact, the results suggest a fast recovery on TE levels on the part of beneficiaries as time with project increased. Chapter 3 uses data from Peru to evaluate the Progress out of Poverty Index (PPI), which is a low-cost and easy to use poverty scorecard developed by Mark Schreiner of Mircrofinance Risk Management, L.L.C for the Grameen Foundation. The PPI estimates the likelihood that a household has expenditure below a given poverty line. The scorecard is a practical way for anti-poverty programs or pro-poor NGOs to monitor poverty rates among their clients, track changes in poverty rates over time, and target services. The results of Chapter 3 show that the PPI is very accurate (in terms of bias and targeting accuracy) when it is applied at the national level and in urban areas. However, when it is applied to rural areas, the tool is not able to discriminate well between the poor and non-poor. After making several modifications to the PPI the results show that these modifications do not translate into significant improvements in accuracy (neither for bias nor for targeting accuracy). We conclude that the PPI might benefit from the creation of a separate scorecard for rural areas that is based on a set of indicators that might be more relevant for rural areas. Chapter 4 uses data from an unconditional cash transfer program, the Child Grant Program (CGP), implemented as a Randomized Control Trial in Zambia, to assess the impact of the program on agricultural production, productive investments, and technical efficiency. We find that the program did not have an impact on the value of harvest or gross margins. However, the program had a positive impact on expenditures in seeds, the share of households that spend on crop production, including miscellaneous expenditures. Moreover, the increase in crop expenditures seems to be reflected in the amount of harvest sold for maize and rice, in the total value of sales, primarily driven by maize sales, and the share of households that sell part of their harvest, primarily driven by sales of maize and rice. We find that the CGP had a positive and significant impact on total operated land (an increase of 0.23 ha.). The impact of the program on farm tools was very modest, although the impact on farm animals was greater: the CGP only increased the share of households that own a hammer (4%); while the program had a positive and significant impact on the number of cattle owned (increase of 0.44), goats (0.13), chickens (1.11), and ducks (0.2). To provide a more nuanced picture of the impact of program, we ran the same set of analyses across three household types based on differences in size of household labor, defined as household members 17 years old or older and younger than 65: (1) single-headed households; (2) nuclear households; (3) and surplus-labor households. The program impacted each of these types of households differently: single-headed households who are labor-constrained seem to have retrenched from agricultural activities, while nuclear and surplus-labor households have intensified and expanded their agricultural endeavors. Importantly, while nuclear households benefitted the most from the program in terms of the number of indicators that were positively impacted, surplus-labor households benefited the most in terms of the magnitudes of the benefits. The results show that the program had no impact on technical efficiency. The results of the ancillary equation for the inefficiency part of the error structure shows that education and access to credit can significantly improve TE levels.