As 2016 presidential hopefuls begin announcing their candidacies, Americans are readying themselves for more than a year and a half of political campaigning.  That’s a long time. Long enough to perhaps cost $5 billion.

However, there are benefits to a long campaign season. As Calvin Coolidge said, “The purpose of a campaign is to send an intelligent and informed voter to the ballot box.” Campaigns may help inform voters and enable them to develop more accurate assessments of the candidates. Long campaigns have the potential to do this even more effectively.

But there is a downside. In a new article (ungated here) Raphael Boleslavsky and I show that informative campaigns can also decrease the incentives for candidates to moderate their views. In other words, more informative campaigns encourage polarization between politicians, which tends to make voters worse off.

In our research, we draw on a game theoretic model.  But the intuition behind and implications of the model are easy to understand.

Imagine that Rick and Hillary are candidates. They are partisans with their own policy views, which are not necessarily aligned with those of the average voter. When announcing their candidacies, Rick and Hillary must each decide whether to run on a platform aligned with their own views, or one more closely aligned with the views of a majority of voters.

The key finding is that which platform they choose may depend on the length of the campaign.

Suppose first that Rick and Hillary live in a world in which there are no debates, media interviews, or public appearances. The candidates announce their policy positions, and then voters cast their ballots.

In this case, voters choose whom to vote for based entirely on how the candidates’ policy positions compare with their own, and any preconceived notions about how likely the candidates are to be competent, trustworthy leaders. In this case, both candidates are likely to choose the more popular policy, recognizing that in order to win election, they must announce a platform that the majority of voters prefer relative to their opponent’s.

This is the median voter theorem at work. The famous theorem says that competition between candidates should lead them to run on platforms aligned with the likely swing voter in the election.

But the median voter theorem does not hold when political campaigns allow voters to form opinions about candidate quality. Suppose now that Rick and Hillary live in a more realistic world in which campaigns last more than a year and a half, and involve numerous debates, interviews and other public appearances.

Over the course of a long campaign, voters will form opinions about how well the candidates think on their feet and react under pressure, how well they deliver a speech, and how well they manage their campaigns. Longer campaigns give voters more opportunity to assess the character, competence, and trustworthiness of the candidates, and form beliefs about which candidate will be most effective in office.

Perceptions about Rick and Hillary will evolve over the course of the campaign. Even Rick and Hillary themselves are uncertain about their performance on the campaign trail, and how the media and voters will respond to their personalities and personal histories. They don’t know whether Rick will come across as uninformed and ill prepared during a debate or when making a speech. They also don’t know whether Hillary will come across as unlikable, or whether a scandal will stick concerning her time as a cabinet secretary.

The longer the campaign, the more likely one of the candidates experiences scandal or a campaign misstep. The longer the campaign, the more time voters have to form opinions about leadership qualities, and the more likely one candidate stands apart on this dimension. This is consistent with recent evidence that voter preferences are more responsive to information on candidate quality than policy, and that moderate voters care mostly about character and competence.

On Election Day, the candidate whom the majority of voters believe is more likable, competent or trustworthy is likely to win election, as long as their policy is not “too extreme” compared with the other candidate. Policy platform matters a lot less here than in the case where a short campaign allows little opportunity to form opinions about leadership ability. In a longer campaign, Rick and Hillary recognize that a more moderate policy is less likely to be the deciding factor in the election. As a result, they have less incentive to moderate policy.

In this way, more informative campaigns can lead candidates to adopt less moderate policies. This increase in polarization hurts the average voter. At the same time, however, a more informed electorate is better able to assess the leadership characteristics of the candidates, which can lead to better decisions and benefit voters.

Whether or not longer, more informative campaigns make voters as a group worse off depends on how many voters are moderate. This is because moderate voters are hurt the most by an increase in campaign information, which always leads to more extreme policies than moderate voters prefer. Partisan voters, on the other hand, are hurt less by an increase in extremism, as it can lead some candidates to move closer to their ideal policy. More informative campaigns tend to make a sufficiently polarized electorate better off, and a sufficiently moderate electorate worse off.

Our findings have implications for anything that increases voter exposure to candidates before Election Day. Increasing the number of debates during a campaign, for example, would likely decrease the incentive for political candidates to run on moderate platforms. The findings are consistent with the observed upswing in the level of polarization over the last 25 years, when first cable news and then the Internet significantly increased voter exposure to candidates.

Over the next 18 months, the country will judge the candidates’ debate performance, public speaking, competence, trustworthiness, leadership skills, and even their looks. Some (but not all) of these things matter a great deal for the performance of a future president, and shouldn’t be ignored. But when candidates are evaluated based on these characteristics rather than their policy platforms, the less likely candidates are to choose the moderate policies that many voters want.

Christopher Cotton is Jarislowsky-Deutsch Chair of Economic and Financial Policy at Queen’s University.