This paper shows reliable evidence that the timing of 2009 swine flu in Mexico affected voting behavior in the next election. However, there may be different interpretations regarding the influence of other causal factors.
RR:C19 Evidence Scale rating by reviewer:
Reliable. The main study claims are generally justified by its methods and data. The results and conclusions are likely to be similar to the hypothetical ideal study. There are some minor caveats or limitations, but they would/do not change the major claims of the study. The study provides sufficient strength of evidence on its own that its main claims should be considered actionable, with some room for future revision.
The paper studies the electoral consequences of the 2009 swine flu outbreak in Mexico. The authors find that the magnitude of the outbreak negatively affected the election results of the incumbent party, the PAN: the negative variation in PAN vote shares in precincts more exposed to the epidemic was larger than in precincts less exposed. Besides, conditional on the magnitude of the peak, this effect seems to be independent of the timing of the peak and how the curve evolved during the outbreak.
The paper undoubtedly addresses a very relevant question with obvious implications to the current COVID-19 context. In general, I find the main claims of the paper reliable and trustworthy, for they are based on a competently executed empirical exercise. I have, though, a couple of remarks that I think should be considered when interpreting the results.
First, the obvious main threat for the identification of the causal effect is the possible existence of correlates of exposure to the pandemic that affect over-time variation in election results at the precinct level. The main model does not include precinct-level regressors other than the epidemic exposure itself (nor their interaction with the year dummies). Indeed, in the absence of such information (and also to account for unobservable differences between precincts) the fact that the 2003 coefficient is close to zero is reassuring, but it would be nice if more pre-treatment time periods are added to confirm this. In any case, for the sake of transparency, it would be nice correlates of exposure across areas are shown descriptively.
Second, I am not fully convinced of one of the lessons the authors extract from the evidence. They claim that one of the implications is that policies aimed at flattening the epidemic curve seem to be politically valuable because the 1) the electoral sanction is independent of the timing of the peak (early or late during the outbreak) + 2) the effect of variation in exposure is weaker in low peak precincts. However, I believe this evidence shows, at best, that the political response was irrelevant for the incumbent electoral results. As long as low peak areas are probably also less exposed, it is difficult to disentangle whether the heterogeneous effect is driven by the peak or the magnitude of the outbreak itself, especially if one considers the possibility that the effect of the number of cases might be non-linear (i.e. stronger once a certain threshold of cases is reached). Besides, Figure SI9 in the supplemental information casts doubt on the robustness of the heterogeneity of the effect in which the claim is based.
Finally, a more minor point is why would the authors not consider a simpler measure of exposure that is relative to the precinct population, apart from the z-score standardization they propose. Simply showing the number of cases per N inhabitants would make the results more transparent.
In any case, again, I think the paper makes a rigorous, important, and timely contribution to the discussion on the political consequences of epidemic outbreaks.