| This dissertation is an attempt to analyze the provision of public goods determined by public choice mechanisms. We first consider a pure public good with a service area covering two or more different jurisdictions. The provision can be determined by two alternative voting regimes: a centralized vote or a decentralized vote in each jurisdiction. Intuitively the provision through the decentralized regime should be lower, but I establish the opposite result over a subset of the parameter space.;Next, we analyze the decennial data of 48 states since 1960 to find the determinants of the provision of public primary and secondary education with a fixed effects panel data model. After correcting for serial correlation with GLS estimators, we find that the effects of income and percentage of households occupied by owner are significantly positive, and the percentage of 5 to 17 year-old (R517) and the percentage of 65 and over (R65) are significantly negative. Additionally, a dynamic model is estimated with instrumental variable, which shows that the one-period-lagged per-child spending is significantly positive but only R517 and R65 remain significant. We also evaluate the effect of the state supreme court decisions on school-finance litigations, which is insignificant.;Finally, we include state tax revenue in the analysis with annual data from 1970 to 2000. The volatility in expenditure on public education varies across states and is correlated with state tax revenue. In a regression where we include the volatility in tax revenue, geographical location, per-capita income, public school enrollment, R65, and state-aid formula, the first two variables are significant in explaining the volatility in expenditure. For each state, the best VAR model is identified in order to assess the time series properties. In 34 states the spending is significantly related to its first lag while in other states it is related to, in addition, its second or third lag. The relation seems to be one-way because in most models the past of the tax revenue affects the spending but not the other way around. |
