Association for Public Policy Analysis & Management

Many development interventions have encouraged decentralization to improve government services in developing countries: by bringing government services closer to the populations they serve, citizens should find it easier to access services and hold politicians accountable. However, decentralization may create political incentives for local governments, which may be more vulnerable to elite capture. In this paper, I examine how political actors (politicians and voters) react to an intervention which provides public services through local governments. I find political responses to the announcement of the decentralization (such as contesting local elections), particularly by those who may anticipate larger private benefits, although these effects seem not to persist after implementation.

I look at the political effects of a land administration decentralization in Burkina Faso, undertaken with support from the Millennium Challenge Corporation (MCC). In this reform, land offices, responsible for providing documentation of customary land claims, were created in local municipalities. Previously, documentation of land rights was rare. I argue that these land offices represent a potentially valuable resource for control by elected officials, as documenting informal rights creates opportunities for elite capture.

MCC conducted an impact evaluation from 2013-2015 during the pilot phase, using a paired-RCT design at the municipality level. I use official electoral returns data from three cycles of multi-seat municipal council elections to investigate differences in political behavior between 'treatment' municipalities and matched 'control' municipalities. I compare changes in outcomes such as political competition, voter turnout, and voter attitudes between 'treatment' and 'control' municipalities, from a pre-program baseline in 2006 to the announcement in 2012, and then after implementation in 2016. The timing of these elections allows me to distinguish anticipatory effects of the announcement of pilot municipalities from the political effects of implementation. To explore mechanisms, I supplement these with Afrobarometer survey data about perceptions of corruption, measures of ethnic fragmentation, and local government budgets.

I first consider anticipation effects, when land office locations have been announced but not yet created. I find an increase of 1.38 political parties contesting elections in municipalities which will receive pilot-phase land offices over a control-group mean of 4.25 parties [p-value = .01]. However, there are no significant differences in measures of electoral competitiveness. I also find a 4% [p-value = .02] decrease in voter turnout in these municipalities. Additionally, residents of treatment municipalities perceive government officials to be more corrupt during the 2012 election. These results are primarily driven by pastoralist areas, where customary land rights are less stable, making formal documentation more politically important. By the 2016 elections, however, municipalities that received local land offices again look like their control counterparts. I explore potential mechanisms for why political behavior responds more strongly to anticipation than to implementation of this decentralization.

Although this paper exploits a randomized piloting of a decentralization program for identification, many causal political questions do not occur in experimental contexts. Therefore, I compare these experimental results with quasi-experimental methods such as synthetic control, in order to assess their validity against the benchmark provided uniquely by my context.