Because there is a controversy about the relationship between democracy and development, my objective is to clarify this confused intellectual landscape by using Korean regional data covering the years 1971–2001. This is a period that spans Korea’s transition from dictatorship to democracy. In this dissertation, I use civil society organizations (CSOs) as a yardstick for quantifying the level of regional democracy. And, I test for the impact of CSOs on economic growth and key public goods – such as education and health services, and social overhead capital (SOC) – under the dictatorship and under a democracy. The democratization of Korea in 1987 forms the boundary between the autocratic and democratic regimes. And, it also provides a good opportunity to identify and compare the efficacy of fiscal decentralization in the different political systems. Accordingly, estimating the impact of fiscal decentralization on socioeconomic development before and after 1987, I examine whether the level of fiscal decentralization stimulated by democratic system leads to better public goods. First, the least-squares analysis with fixed effects shows that political civil society has beneficial effects on education and health services but unfavorable effects on SOC and regional domestic product, experiencing a democracy. Moreover, nonpolitical CSOs have also had an ambiguous influence on economic outcomes under a democracy. This is consistent with existing literature arguing that the CSOs can be either rent-seekers and/or proponents for social welfare. A surprising finding, however, is that in general political and nonpolitical CSOs are strongly associated with good economic outcomes under the autocracy. I develop a conceptual framework for interpreting these findings. My analysis also indicates that fiscal decentralization is not always more efficient under a democracy. In fact, with the exception of health services and SOC, fiscal decentralization has a negative association with many economic outcomes under a democracy. All results are robust to the specification checks and the panel corrected standard errors (PCSEs) method.