Date of Award

Spring 1-1-2013

Document Type


Degree Name

Doctor of Philosophy (PhD)



First Advisor

Charles de Bartolome

Second Advisor

Terra McKinnish

Third Advisor

Robert McNown

Fourth Advisor

Francisca Antman

Fifth Advisor

Amy Liu


The analyses in my dissertation are related to how government expenditures are spent and how they should be spent as a whole. In the first chapter, the evidence of manipulation of expenditure is found and the explanation with rational voter is provided. The basic idea is that two types of incumbents will spend more money in their less-preferred sector to attract median voter when they face a small chance of winning in the following election. This expectation is confirmed using a unique Korean local government data set. The fact that there have been two dominant types of parties in Korea, and that expenditures are classified by the social development and the economic development enable this analysis. This analysis is unique in that this is the first empirical analysis trying to see how the opportunistic behavior of incumbent is done according to the probability of winning under heterogeneous preference about composition.

The second chapter examines the reason why governments transfer in-kind goods instead of cash. While there have been several analyses of possible explanations for in-kind transfers, the analyses based on a political view are scarce. This analysis provides the empirical evidence supporting the in-kind transfers as a political equilibrium. Two different hypotheses inducing different expected empirical results based on a political view are provided and are tested using cross-county data. The findings show that the in-kind transfers are increasing as the country becomes more democratic and the median voter is more away from median income.

The last chapter investigates the relationship between child labor and the participation in CCTs (conditional cash transfers) in Mexico. Since the CCTs are introduced to eliminate child labor and incentivize them to attend school, it has been thought that households with child labor are less likely to participate in the program. This chapter doubts this common belief. Because the conditionality in CCTs is normal goods and child labor could be the result of household characteristics making conditionality more valuable. The estimated probability model shows that households having child labor are actually more likely to apply to the program.