Throwing the rascals out is tricky, but not impossible

It is a commonly understood feature of democratic political systems that under-performing governments tend to be ejected by voters in favour of the most palatable alternative, but is that really the case? Catherine E. de Vries, drawing on new research, argues that voter sophistication and issue salience each play a key role, concluding that incumbent governments can expect to earn credit for favourable, or blame for unsatisfactory, policy outcomes.


Voters replaced Gordon Brown with David Cameron (Credit: Phil CC BY NC ND 2.0)

Are governments punished for past performance at election time? The idea that voters use elections to hold governments to account for policy performance lies at the heart of representative democracy. For democratic government to work well, voters need to be able to mandate governments to implement policy and to hold them accountable for their actions after a term in office.

Voters casting a vote based on policy performance can act as the rational gods of vengeance or reward. For a long time, political scientists and economists praised the performance voting model (sometimes called the retrospective voting model) and showed that voting based on incumbent policies was a uncomplicated way for largely ill-informed and low interested voters to keep their decision making simple yet still making the right choice.

More recently however, experimental studies provide a more sombre assessment and demonstrate that performance voting leads to biased or even bad ballot choices as voters are myopic and easily swayed by rhetoric. These findings give rise to a puzzle: How can we resolve the seemingly inconsistent findings that a) voters do engage in rational punishment and reward at election time yet b) at the same time have too short time horizons to properly evaluate policy outcomes and pay more attention to rhetoric than to facts?

The Argument

In a study recently published in the European Journal of Political Research, Nathalie Giger and I suggest that heterogeneity among voters might go a long way in uniting these different views. Our argument is simple:

  • First, in order for voters to judge if an incumbent deserves to be re-elected based on past performance, they need to be aware of government actions and the outcomes of these activities. By consequence, performance voting should be accompanied by a sizeable sophistication gap. Only voters with high levels of political sophistication are able to relate policy outcomes to government activities and to the vote.
  • Second, increased salience of a policy area will lower the cognitive costs associated with performance voting and thus narrow the sophistication gap. Voters at lower levels of political sophistication might generally not be willing to pay the price of getting informed about policy, but when they care enough about a policy outcome, they will.

The Evidence

In order to test these intuitions we use data from elections between 2002 and 2006 in the 25 advanced democracies included in the Comparative Study of Electoral Systems (CSES). This dataset allows us to examine performance voting across a large area of policy areas, not just concerning the economy, but also the environment, social policy, or immigration policy. This is a crucial feature of the data as it will allow us to examine if the salience of the policy area narrows the sophistication gap.

This chart shows the coefficients for the direct effects of voters’ political sophistication (labelled Sophistication), voters’ evaluation of government performance across different policy areas (labelled Performance), voters’ evaluation of government performance in the policy area they think is most important to their country (Salient Performance) and the interaction effects between the performance evaluations and voters’ level of political sophistication (labelled as Performance*Sophistication and Salient Performance*Sophistication respectively).

Several findings stand out. First, while the direct effect of voters’ political sophistication (here measured through an additive scale of political knowledge) has no direct effect of voting for incumbent party, performance evaluations do have a clear positive effect.

This corroborates the performance voting model, that is to say the more positive voters are about policy outcomes the more likely they are to vote for a party in government, and vice versa. Second, we find evidence of a sophistication gap in performance voting based on policy evaluations generally. When we calculate predicted probabilities, we find that if political sophistication increases from one standard deviation below the mean to one standard deviation above the predicted probability of incumbent vote due to policy performance evaluations by roughly 3 percentage points.

Although at first this might seem like a small effect, in real-world elections 3 percentage points might lead to an electoral victory rather than a loss or lead to coalition inclusion or not. This moderating effect of political sophistication is robust even when we control for partisanship. Third, the sophistication gap disappears if we take into account if voters view the policy area as important. If the latter is the case, we no longer find differences between high and low sophisticates in terms of their performance vote (this can be inferred from the fact that the confidence interval associated with the Salient Performance*Sophistication coefficient includes the null).

One final point that is worth highlighting in this respect is that when we study performance voting, we should not only look at the economy as most work to date does. This chart  displays the importance voters from the different countries included in this study attach to the economy. It becomes clear that the economy is one of the three most important policy areas for voters in 21 out of the 25 countries shown here, only around a quarter of the population in these countries rates this issue as the most salient (24.9 per cent; see the reference line in the figure). This leaves about 75 per cent of the population deeming other policy areas more important. While the economy, with roughly 80 per cent, is by far the most important issue for German voters in the 2002 election, less than 10 per cent of British or Japanese voters classified the economy among the top three policy areas in the 2005 and 2004 election, respectively.

The Verdict

These results provide a mixed assessment of the performance voting model and the way in which elections safeguard government accountability. On the one hand, they lead to a cautious tale about the degree to which electoral accountability by itself is properly suited to incentivise public officials to craft policy solutions that benefit all voters, given that only a subset of these voters have the capacity to hold elites accountable. Low sophisticates often lack the ability (or will) to access the information needed to reward or punish governments. Indeed, political facts are not easy to come by given that both politicians, aiming to sway public sentiment in their favour to secure re-election, and the media, wanting to maintain consumer interest, often lack the incentives to provide them.

On the other hand, our findings suggest that as long as voters care enough about government activities in a particular policy area the sophistication gap in performance voting narrows substantially. This suggests that salience may partially negate the informational costs associated with performance voting. Incumbents can thus expect to earn credit for favourable, or blame for unsatisfactory, policy outcomes that voters care about, and this should provide some impetus for responsive policy making.

Note: this article is based on a piece in the European Journal of Political Research which can be found here. It represents the views of the author and not those of Democratic Audit or the LSE. Please read our comments policy before posting. The shortened URL for this post is:

OLYMPUS DIGITAL CAMERACatherine E. de Vries is Professor of European Politics at the University of Oxford. For more information on Catherine E. de Vries’ research see her website.

Similar Posts