UC Riverside

Decentralization has dramatically altered governance in developing countries. However, the empirical evidence regarding its effects on the provision of public goods has been limited and ambiguous. In the first chapter I argue that this ambiguity stems from insufficiently disentangling partial from full decentralization. I differentiate between these two types by comparing administrative decentralization, political decentralization, and their complementarities in Indonesia. The paper employs a unique Indonesian panel of village level outcomes and a difference-in-differences estimation strategy with village level fixed effects. I show that use of a naïve specification that only considers political or administrative decentralization as separate treatments while neglecting their complementarities leads to an omitted variable bias problem. Results from a more complete specification suggest that districts that were treated with both types of decentralization, i.e. full decentralization, display significantly greater provision of public goods compared to those that experienced partial decentralization in the form of only political or administrative decentralization.

The second chapter of this thesis contributes to a related literature. It focuses on the causal effect of occupational transitions on consumption changes and poverty. Recent research has pointed out that sectoral transitions from the agricultural to the non-agricultural sector could be a successful pathway out of poverty due to higher productivity in the non-agricultural sector. But these studies face several limitations, such as the use of cross sectional or short panel data. We address some of these gaps and introduce two novel ways of defining sectoral transitions. Each of these definitions is used to exploit a fixed effects and an instrumental variable strategy with long run panel data on Indonesian households. Under both strategies we find that consumption growth is conditional on initial economic status and the nature of the transition---the growth was relatively higher only for those households who were either poor and agricultural in the baseline, or non-poor and non-agricultural. In terms of poverty, we find longer non-agricultural employment resulted in a positive probability of exiting poverty and a negative probability of becoming poor. Based on these findings we propose that pro-poor policies must be tailored to the agricultural or non-agricultural status of a household.

The third chapter is a natural extension of the first one. In this chapter I test whether individual outcomes associated with publicly provided goods, such as schools and health-centers, depend on decentralization complementarities. Based on three different datasets I conclude that individual welfare outcomes were significantly better for those belonging to fully decentralized districts compared to those in partially decentralized districts.