This dissertation explores how decentralization and democratization shape the local politics of Latin American cities. It consists of three papers that investigate political barriers to democratic representation and equitable governance for urban residents segregated by socioeconomic class. The first paper introduces the Fragmented Latin American Urban Democracy (FLAUD) dataset. It estimates how divided over 1,000 cities in 17 countries are by politico-administrative boundaries across multiple levels of subnational territorial organization. By integrating geospatial data on metropolitan footprints, district boundaries, and population density, I generate new estimates of administrative fragmentation sensitive to the underlying distribution of residents across urban districts. I propose two derivative forms of fragmentation—electoral and sectoral—that characterize whether districts contain predominantly urban or rural constituents and the effect of fragmentation on governance across various policy areas. I demonstrate two potential applications of FLAUD by estimating the effect of urban fragmentation on the spatial distribution of amenities within cities and on reported satisfaction with local governance and service provision by social class. The second paper examines the effect of higher-level partisan voting on local representation by socioeconomic status. In cities segregated by social class and fragmented into numerous electoral districts, wealthier and poorer residents often form separate constituencies for local elections. When voters form class-based attachments to parties in higher-level elections and use party labels in local candidate selection, class-concentrated districts are more likely to elect representatives in uncompetitive races, possibly weakening democratic accountability. I create a panel dataset of census demographics and election returns for districts in Buenos Aires, Santiago, Mexico City, and Lima to estimate the degree to which voters from different classes have been exposed to competitive elections over the past two decades. I compare realized election results against simulated elections to benchmark the independent effect of segregation and determine whether and how higher-level political preferences linked to social class influence local representation in each city. The final paper investigates how the electoral geography of class-based voting can impede the election of popular mayors unaffiliated with dominant parties. Voters often rely on party labels from higher-level elections as heuristics when voting for local representatives. However, doing so may not reflect sincere preferences for those parties locally but rather a lack of reliable information on alternative candidates. I adopt as a case study Lima, Peru, where parties are weak and personalistic, yet party performance in provincial (city-wide) and district (sub-urban) elections are strongly correlated. Using two decades of municipal election results, I show that provincial voting patterns, which are candidate-centric and class-based, strongly predict district-level party vote share. Voters deviate from this pattern only when they have independent information about district candidates’ qualifications. Yet locally popular mayoral candidates are less likely to get elected when competitors affiliated with especially dominant provincial parties enter local races with substantial, often unjustifiable, electoral advantages.