Discretionary transfers: does alignment really matter?

Hand of man in suit offering money

Politicians often target discretionary funds based on political considerations that have little to do with welfare. Particularly, the literature suggests that the upper level of government in federal countries disproportionally allocates funds to co-partisan local governments, intending to favour the re-election of aligned officers.

In BSE Working Paper 1337, “Drought-reliefs and Partisanship,” Federico Boffa, Francisco Cavalcanti, Christian Fons-Rosen and Amedeo Piolatto analyse, theoretically and empirically, the extent of political favouritism in the allocation of discretionary central funds.

Alignment matters only before some local elections

In their theoretical and empirical analysis of the role of alignment in the assignment of discretionary transfers in a federal country, the authors recognise that the timing of the transfer matters in shaping both the decision makers’ incentives and the resulting (biased) allocation. 

In particular, in a setting in which central and local elections alternate, even a purely office-motivated central politician with perfectly informed voters displays partisan bias in the allocation only before local elections, while the allocation is based on welfare considerations before presidential elections. 

To fully understand why this occurs, notice that transfers are used to swing votes and they are more effective when assigned to voters that need the transfer the most. This is why welfare considerations enter into the decision mechanism of the office-motivated policymaker. 

Furthermore, federal elections are single-district, hence, every vote counts the same. Instead, what matters the most at local elections is, for each district, whether the transfer makes the electoral-outcome swing. 

When central politicians assign transfers before local elections, they take into consideration: i) how many votes swing in a municipality after the transfer (hence, welfare matters) ii) whether the shift is pivotal for the mayor’s re-appointment, iii) how much gratitude the reappointed mayor will show, measured as the mayor’s support/endorsement of the central politician at the subsequent presidential elections. Such gratitude depends on alignment and explains the allocative partisan bias.

The pivotality of transfers before local elections is crucial to understanding an important contribution of the work. The partisanship bias is only observed in a subset of municipalities where the need for the transfer is intermediate.

Drought aid reliefs in Brazil

In the context of drought aid reliefs in Brazil, the theoretical model shows that, under fairly standard concavity conditions, municipalities affected by the most severe droughts benefit the most from transfers. Hence, the model predicts that, before presidential elections, relief is allocated to districts based on aridity. However, before local elections, the allocation strikes the optimal balance between building a long-term relationship with mayors (in which case alignment matters) and the impact of transfers on voters (in which case the level of aridity matters).

The empirical analysis uses Brazilian data in the 2000-2016 time window. Brazil represents an ideal setting to test the effects of the election-timing on the transfer-allocation criteria since presidential and municipal elections alternate every two years. Data on precipitation and evapotranspiration allow the authors to compute the SPEI aridity index, which can be used to construct a benchmark optimal allocation. The evidence confirms the predictions of the model.

Overall, aridity matters for the allocation of transfers. Before mayoral elections, the probability of receiving aid relief increases respectively by 6.7 p.p. and 1.6 p.p. when transitioning from low to moderate aridity or from moderate to severe. A qualitatively identical pattern is observed before presidential elections, when the average probability increase is respectively by 13.4 p.p. (from low to moderate) and by 6.8 p.p. (from moderate to severe).

The figure below descriptively shows the relative probability of receiving drought-aid-relief, conditional on a given level of aridity. Municipalities with similar SPEI are grouped together. Positive values on the vertical axis represent a greater probability of drought-aid-relief in favour of aligned municipalities. It is only in the intermediate range of SPEI levels before mayoral elections that aligned municipalities are systematically more likely to receive those funds compared to non-aligned municipalities.

Graph of share of municipalities obtaining aid relief
Figure 1. Share of municipalities that obtained aid relief: difference between aligned and non-aligned municipalities.

Notes: The vertical axis represents the difference between the ‘share of aligned municipalities that received aid’ and the same share for the non-aligned municipalities: positive values correspond to when the share of aligned municipalities that received aid is larger than the one for non-aligned municipalities. Each dot corresponds to a different degree of aridity, measured by the Standardised Precipitation Evapotranspiration Index (SPEI ). The two dashed vertical lines delimit the area defined as moderate aridity.

The Regression Discontinuity Design analysis

The authors adopt a Regression Discontinuity Design (RDD), based on close municipal elections. This guarantees a causal identification of the effects, under the assumption that municipalities in which the election was decided by a small margin of victory are statistically indistinguishable.

Before local elections, aligned districts enjoy a 6.3 p.p. higher probability of receiving a transfer. While such bias is in line with the literature, the authors show the novel finding that the effect is fully driven by municipalities that suffered a moderate drought. In this case, aligned municipalities have 18.1 p.p. higher chances of receiving aid reliefs. Instead, there is no alignment bias when aridity is either low or severe. Furthermore, allocations are not distorted vis-à-vis the benchmark allocation before presidential elections.

The figure below shows the differences in predicted values around the RDD threshold between aligned and non-aligned municipalities. The authors split the sample of municipalities based on aridity (low, moderate, severe) and implement the regressions separately. The top three graphs correspond to the two years before mayoral elections and the bottom ones to the two years before presidential elections. Consistent with the model prediction, the only statistically significant discontinuity occurs in municipalities with moderate aridity preceding municipal elections.

Graph of share of municipalities that obtained aid relief
Figure 2: Regression Discontinuity Design.

Notes: Graphs represent predicted values of RDD. The dependent variable is aid relief, the forcing variable is the margin of victory of the candidate from the party of the incumbent president at the previous mayoral election. The top three graphs show the predicted values separately for the municipalities at each aridity level in the years preceding a municipal election. The bottom three graphs represent predicted values for the years leading up to a presidential election. Circles represent the local mean and dashed lines represent 95% confidence intervals.

Policy recommendations

The authors’ analysis contributes to a better understanding of the mechanism through which distributive politics operates. It is natural to assume that the socially optimal assignment of resources should follow a logic of needs and efficiency, rather than political opportunism. When it is not feasible to delegate the allocation to an independent decision-maker, the analysis suggests avoiding the allocation of transfers in the times just before local elections, at least in countries where central and local elections alternate. An alternative route to mitigate the distributional distortion could consist in designing geographically different districts for voting and transfer purposes, making it hard for the central government to target any specific voting district.