Elections and externalities of health expenditures : spatial patterns and opportunism in the local budget allocation

In this paper, we tested the hypothesis that policy makers' choices concerning health spent at the local level are spatially correlated and electorally oriented. We check the influence of a set of demographic, electoral and economic determinants of public health activity. We performed a spatial panel data analysis encompassing 399 municipalities in the period from 2005 to 2012, estimating six models, namely: all years, all years controlling by election type, non-electoral years, electoral years, central-elections years and local-elections years. Our contribution to the literature lies in three findings. Firstly, we show that health spent is driven for (global and local) positive spatial autocorrelation, and it is persistent, which means that independent of electoral calendar, there exists spatial effects in the allocation of spent. The parametric estimation that best fitted with data was Spatial Error Model Estimation (SEM), and the lambda value for all models were the same (0.099), indicating that the spatial correlation affects both electoral and non-electoral year's expenditures, in central and local elections. In other words, we found a positive spillover effect in the public health spent at the local level. Exogenous effects (normative power of the law when it comes to minimum values of health spent in the municipalities) and correlated effects (the role of local health infrastructure, particularly in local-election years) help to understand the channels of this spatial dependence. A second issue is that elections are strong enough to change the allocation pattern in local governments, probably as a political strategy to seduce voters. Health spent seems to have a politically motivated component, stronger to local elections (considering incumbent's reelection and/or his goal of electing the supported successor), but still significant in central elections. Considering that less than 5% of Brazilian tax burden is in the local level, the relevance of central elections is probably tied to grants access. The last point is the difference between central and local-elections effects. The main channels to explain the per capita health expenditure level in municipalities, according to all Models, are demographic issues related to aged people (positive effect) and young people (negative effect). Moreover, in election-years the positive aged people effect increases and the negative young people effect smooths. Another important issue is that population density variable changes from a negative effect (in all years and non-electoral years' models) to a positive effect in election-years models. This suggests that the more a municipality is dense populated, the more efficient will be the campaign spent to seduce voters, and one effective instrument to do that is by increasing public health expenditure in elections years.