Is Smaller Better? An Assessment of the Causes and Consequences of Subnational Fragmentation in Brazil
In an effort to spur economic development and entrench democratization in the late-twentieth century, many developing countries decentralized various central government functions. By reducing the geographic distance between governments and their constituents, many posited that decentralization would make politicians more responsive and accountable to the unique needs of their communities and, thereby, improve the provision of public goods and services and enhance political participation. Whether decentralization actually generated these anticipated outcomes remains up for debate.
As countries have devolved power to smaller political units, some have experienced subnational fragmentation - a process in which some local communities split into two (or more) new communities. Increasingly common throughout the developing world, subnational fragmentation is a phenomenon worthy of study on its own, and the within-country comparison units it generates provide a means of introducing rigorous new evidence into unresolved debates on the effects of decentralization. In my dissertation, I leverage the within- country comparison units afforded by subnational fragmentation to assess its causes and consequences.
I study subnational fragmentation in Brazil. Brazil is one of the most decentralized countries in the world, where the most disaggregated political unit (the municipality) has increased in number by more than 25% since the country’s recommitment to democracy in 1985. To assess the causes and consequences of subnational fragmentation in the country, I rely primarily on an original dataset of municipality creation tracing the administrative, political, fiscal, and demographic histories of all 5,570 Brazilian municipalities over the 1988- 2015 period. I supplement this municipality creation dataset with data compiled from archival research in the Legislative Assembly of the Brazilian state of Minas Gerais. With these data sources, I use conventional regression analyses to uncover the causes of fragmentation and implement matching and difference-in-difference analyses to estimate its consequences.
In the first empirical chapter of my dissertation, I study the causes of subnational fragmentation. Specifically, I address the following question: What are the causes of and roadblocks to subnational fragmentation in Brazil? Contrary to conventional wisdom, I find that the structures of intergovernmental transfer formulas are not necessarily the primary motivations underlying subnational fragmentation in Brazil. Regression anal- yses implemented with my data on municipality creation suggest that grievances stemming from geographic expansiveness, political heterogeneity, and municipality age also spark political action and the initiation of fragmentation proceedings. However, through archival research in the Legislative Assembly of the Brazilian state of Minas Gerais and assessing plebiscite results obtained through the state’s Regional Electoral Court, I discover that not all municipalities that pursue fragmentation actually realize their end goal of dividing into new polities.
After local leaders initiate fragmentation proceedings by submitting fragmentation proposals to their re- spective state legislatures, the state legislature and local voters (through plebiscites) have veto power over fragmentation. In the second part of this chapter, I study the conditions that predict proposal “success" in the state legislature and with voters. In doing so, I provide more comprehensive insight into fragmentation in Brazil than others have done to date and contribute to broader literatures on legislative behavior and participation in direct democracy.
Regression analyses suggest that political alignments factor into state governments’ evaluations of fragmentation proposals. I find that state governments are more likely to approve of fragmentation proposals submitted by local leaders with whom they are politically aligned and that state governments are more likely to reject fragmentation proposals submitted from municipalities governed by politically unaligned mayors (due to the presumption that fragmentation could harm the political prospects of politically aligned mayors). My results suggest that plebiscites are more likely to pass when voters have sufficient information to make inferences about their future under new political arrangements. I suggest that this information exists as a function of the number of districts involved in proposal initiation.
In the second and third empirical chapters of my dissertation, I assess the consequences of fragmentation for the provision of public goods and services supplied at the community level and for political participation in local elections. In these chapters, I assess the extent to which decentralization lives up to its promises by asking: Does subnational fragmentation improve the quality of public primary education? Does subnational fragmen- tation increase turnout and reduce the proportion of null and blank ballots cast in local elections? In order to address these questions, I pre-process the data that I have collected on municipal fragmentation via matching to account for pre-fragmentation characteristics that differentiate fragmenting and non-fragmenting municipal- ities. The results from my first empirical chapter on the causes of fragmentation inform the characteristics on which fragmenting and non-fragmenting municipalities are matched. Then, I implement difference-in-difference analyses with this data.
With respect to the effects of fragmentation on the provision of public services, I find that fragmenting municipalities initially underperform in education infrastructure quality in comparison to their non-fragmenting counterparts but that the gap in education infrastructure quality between fragmenting and non-fragmenting municipalities diminishes over time. I attribute this result to the increased administrative attention and concentration of resources in new municipalities in the aftermath of fragmentation. With respect to the effects of fragmentation on political participation, I find that fragmentation initially enhanced turnout, increased the valid vote share, and decreased both the blank and null vote shares in local Brazilian elections but that these positive effects did not persist beyond the first local elections in the post-fragmentation period. Overall, my results suggest that, in the long run, fragmentation achieves some anticipated improvements in service delivery but does not sustain enhanced political participation. More broadly, the results of my dissertation suggest that reducing the distance between governments and constituents can yield some, but not all, of its promises.
Chapters of my dissertation are available upon request.