Democracies might be in jeopardy if voters are not able to properly assess policymakers due to the presence of incomplete information. A few existing studies cast some doubts on such situations by providing evidence of voters punishing/rewarding incumbent politicians in the presence of events that are orthogonal to government policies, such as shark attacks, the performance of local sport teams, or lottery winnings (Achen and Bartels 2004, Bagues and Esteve-Volart 2016, Healy et al. 2010). Yet, to assess properly whether incumbent politicians are held accountable for policy choices, it is important to understand whether voters are sufficiently sophisticated to go beyond the intrinsic noise present in the realised effects of public policies.
Since politicians endogenously choose their policies to enhance their re-election probability, it is typically difficult to have an adequate counterfactual to judge what would have been the voters’ response if different policies were implemented. In order to identify whether and how voters respond to government policies, the ideal experiment would require the government to randomly manipulate the content of a policy and then this random manipulation would map onto different outcomes. For example, to analyse voters’ response to a tax increase or cut by the central government, it would be necessary to observe a random variation of its effects across lower levels of government, for instance municipalities. Or, given a tougher crime control policy, it would be necessary to observe locally random variation in crime rates. Indeed, conditional on the ideological preferences of voters on these types of policies, the variation in local response would identify the causal effect of the policy outcome on voters’ electoral behaviour, if any. That is, it is necessary to observe variations in the effects of the policy that are independent both from the voters’ and the government's characteristics.
Our case study: Criminal justice in Italy
In a recent paper (Drago et al. forthcoming), we address this issue by focusing on criminal justice. We exploit a natural experiment that allows us to have a proper counterfactual to evaluate the voters’ response to the consequences of the policy, keeping their ideology and the impact of the policy on ideological stands constant.
In July 2006, the Italian government implemented an (unanticipated) collective pardon due to dramatic overcrowding in prisons at that time. As a result, a subset of the prisoners with less than 36 months of residual time to serve were released on 1 August 2006 (Figure 1).
Figure 1 Incarceration rates
Note: The figure illustrates the variation in the incarceration rate (i.e., per 100,000 people) in Italy before and after the collective pardon bill.
The design of the policy was such that released prisoners who recidivate within a five-year period would be charged an additional sentence equal to their residual sentence at the time of their release under the pardon. This created an incentive to refrain from reoffending for pardoned individuals that increases in the length of the residual sentence. Such an incentive, as shown in previous research (Drago et al. 2009), turns out to be exogenously distributed across released prisoners. Two identical individuals that entered prison the first time with a sentence of 50 months, at the time of pardon in August 2006 may have two different residual sentences, and thus a different incentive to recidivate, because they entered prison in different periods. Since the date of entry into prison is plausibly exogenous to future criminal behaviour, the Italian collective pardon provides a unique opportunity to evaluate voters’ response to the realised effects (recidivism rates) of the pardon. Indeed, the heterogeneity in the residual sentence remains even when aggregating the individual heterogeneity at the municipality where pardoned inmates lived (Figure 2).
Figure 2 Geographical distribution of the incentive to recidivate
Hence, by using the variation in the incentive to recidivate across municipalities, we can assess the extent to which voters respond to the effects of the crime control policy by holding all the rest equal. In our study we first show that, as expected, municipalities where pardoned individuals had a higher incentive to recommit criminal behaviour experienced higher recidivism. Then, we document that individuals do take into account the observed effects of the policy in their voting decisions. In municipalities with a higher incentive to recidivate, voters ‘punished’ the political coalition that put forward such a pardon (centre-left) in the first post-pardon parliamentary elections. The effect is quantitatively relevant. A one standard deviation increase in the incentive to recidivate (corresponding to an increase in recidivism of 15.9%) led to a 3.06% increase in the margin of victory of the centre-right coalition in the post-pardon national elections (2008) relative to the last election before the pardon took place in 2006.
This shows that worse observable effects of the policy at the local level imply worse electoral outcomes for the politicians responsible for such policy. What are the mechanisms that drive this result? We show that where the incentive to recidivate is higher, newspapers report more crime news on pardoned individuals recidivating. Moreover, voters update their beliefs about the competence of the incumbent coalition to deal with crime. Importantly, a higher incentive to recidivate was not associated with individuals being more likely to perceive crime as the most important issue in Italy or in their city. This suggests that votes correctly associated the pardon with the recidivism of pardoned inmates and not with crime in general.
General implications for policy
Our case study shows that voters respond to the observed effects of a public policy (both in terms of beliefs and behaviour) in a way that is consistent with retrospective voting models of electoral accountability (Persson and Tabellini 2002, Besley 2006, Ashworth 2012). Or, at least, they seem to be able to do so as long as they observe the effects of a policy, either through direct observation and word of mouth or, as we document, via news media.
Authors’ note: These findings are relevant for the political debate in Europe and abroad. Voters may hold politicians accountable for the realised effects of public policies as long as it is possible to identify who is responsible for such policies.
Achen, C H, and L M Bartels (2004), “Blind retrospection: Electoral responses to drought, flu, and shark attacks”, Working Paper.
Bagues, M, and B Esteve-Volart (2016). “Politicians’ luck of the draw: Evidence from the Spanish Christmas lottery”, Journal of Political Economy, 124 (5), 1269-1294.
Healy, A J, N Malhotra, and C Hyunjung Mo (2010), “Irrelevant events affect voters' evaluations of government performance”, Proceedings of the National Academy of Sciences 107(2): 12804-12809.
Ashworth, S (2012), “Electoral Accountability: Recent Theoretical and Empirical Work”, Annual Review of Political Science 15: 183–201.
Besley, T (2006), Principled Agents? The Political Economy of Good Government, Oxford University Press on Demand.
Drago, F, R Galbiati, and F Sobbrio (forthcoming), “The Political Cost of Being Soft on Crime: Evidence from a Natural Experiment”, Journal of the European Economic Association.
Persson, T, and G Tabellini (2002), Political Economics: Explaining Economic Policy, The MIT Press.