Share this:

George Deltas 80x108Helios Herrera 80x108Mattias Polborn 80x108Despite the current wall-to-wall coverage of the 2016 primary race, the primary elections themselves are not scheduled to begin until February 2016, and will last until June. This drawn-out primary cycle gives a great deal of influence to a small number of voters in early primary states, such as Iowa and New Hampshire. George Deltas, Helios Herrera and Mattias Polborn look at proposed alternatives to this system, in the form of a one-day national primary system or one where all states in the Northeast, Midwest, West and South would vote simultaneously. Using models of voter information, they argue that a sequential voting system performs much better than a one-day national primary, but that the parties could improve the system even further by encouraging lagging candidates to drop out more quickly.

As the 2016 Presidential primary race heats up, voters have a large number of potential candidates to choose from. On the Republican side, there are currently 16 declared candidates, none of whom currently garners more than 20 percent support on the RealClearPolitics Average of Polls.

Several of these candidates will not even make it on the ballots of the first caucuses and primaries in Iowa and New Hampshire in January 2016, because weak donor support and polling numbers will have already forced them out by that time. In fact, it is likely that several declared candidates recognize that this will be their ultimate fate, but they are staying in the race in order to gain free media exposure.

Still, even after several departures there will be a substantial number of candidates for voters to choose from, at least for the first primaries. At that stage, coordination will be an essential: For example, as long as Mike Huckabee, Ted Cruz and Rick Santorum all remain in the race, they will split the support of Evangelical Republicans among each other, and so will be less successful than if there was only one Evangelical candidate behind whom the socially conservative wing of the party would unite. Similarly, more moderate candidates such as Jeb Bush or Scott Walker are also likely to fish from one pond of voters that faces a similar coordination problem.

While like-minded voters benefit from coordination, which specific candidate that they should coordinate on is a difficult problem. The candidates are still mostly unknown to voters nationwide and their performance in national debates and on the campaign trail should play a major role for voters. Rushed coordination might eliminate early on a candidate who would have been great if he was ever seriously considered. Learning about candidate quality needs time.

In new research, we study the trade-off between voter coordination and learning about candidate quality, and how it is affected by different ways to organize the timing of the Presidential primaries.

In principle, the Presidential primaries do not have to be organized such that Iowa and New Hampshire vote first, followed sequentially by other states. The present system gives a lot of power to very few non-representative voters in those early primary states, and induces candidates to cater to their special interests.

Some therefore argue that we should move to a national one-day primary system where all states vote at the same time. Another system, proposed by the National Association of Secretaries of State (NASS), has four large rounds of voting in which all states in the Northeast, Midwest, West and South would vote simultaneously.

Simultaneous and sequential primaries can clearly yield different results: For example, the momentum from his Iowa victory helped Barack Obama in 2008, particularly because it eliminated the other anti-Hillary Clinton candidate, John Edwards.

Primary election featured

Credit: Adrian Black (Flickr, CC-BY-NC-2.0)

To better understand what the effects of different primary systems might be, we analyzed a model of coordination and learning over time. Coordination matters if there are two approximately equally strong voter groups (say, conservatives and moderates among Republican primary voters) and candidates in one camp outnumber those in the other — in this case, vote splitting may doom the candidates in the overcrowded camp. Sequential systems temper this effect because some candidates will eventually drop out.

To model learning about voter quality, we assume that voters in different states receive some information about candidate quality, for example by participating in campaign events and interacting with the candidates, or by watching local debates. Those who vote later are also indirectly influenced by information in earlier states: While they cannot directly observe the same information, if a candidate received many votes in an early primary, voters there likely received positive information about him. This gives rise to a “momentum effect”– winning, especially in early primaries, helps a candidate because it conveys positive information about him to voters in later states.

However, while momentum in our theory is a consequence of rational behavior rather than of an irrational psychological desire to “vote for winners”, it nevertheless can interfere with social learning about candidate quality. Suppose, for example, that voters in New Hampshire receive information about a candidate that is too good relative to the true value. In a sequential primary system, this lucky break has a much larger effect, because it also impresses — incorrectly — many voters in later states who discount their own, countervailing signals. If, instead, all states vote simultaneously, then all signals have the same weight, which statistically leads to a better chance that a high quality candidate wins more votes than a low quality candidate in the same ideological position.

So, from a theoretical perspective, the possibility to coordinate favors sequential elections, and the learning effect favors simultaneous elections. To decide which effect is stronger, we estimate the parameters of our model, using data from all post-2000 contested Democratic and Republican Presidential primaries, and find that both key features of the theory — voter learning about candidates, and the need to coordinate — are quantitatively important in these elections.

We then simulate the performance of different systems for the parameters found. Both forms of sequential elections –the current sequential system, and the four region NASS proposal — perform about equally well, and a lot better than simultaneous voting in all states.

Another interesting result is that lagging Democratic and Republican candidates in the data dropped out too late from the voters’ point of view. Intuitively, individual candidates do not internalize the negative effects they impose on their party and its voters by continuing their campaign.

Parties should therefore nudge candidates into withdrawing earlier than they currently do, for example by awarding the two leading candidates after a certain number of contests (say, 10 states) the same number of additional superdelegates for the convention. This would not distort the contest between the two front-runners, but a potential third candidate has a harder time catching up and might therefore drop out sooner.

These results do not just hold for the baseline parameter values estimated from the historic Democratic and Republican primaries, but are also very robust with respect to reasonable variations (such as doubling or halving the estimated parameter values). So, even though the estimates are based only on three contests each for Democrats and Republicans, we can be quite confident that the central results are very robust.

This article is based on the forthcoming paper ‘Learning and Coordination in the Presidential Primary System’ in the Review of Economic Studies.

Please read our comments policy before commenting.  

Note:  This article gives the views of the authors, and not the position of USApp– American Politics and Policy, nor of the London School of Economics.

Shortened URL for this post:


About the author

George Deltas 80x108George Deltas – University of Illinois at Urbana-Champaign
George Deltas is Professor of Economics and Associate Head in the Department of Economics at the University of Illinois at Urbana-Champaign. His research interests include industrial organization, political economy, and environmental economics.  

Helios Herrera 80x108Helios HerreraHEC Montreal
Helios Herrera is an associate professor in the Department of Applied Economics of HEC Montreal, Canada. His research interests are mostly in political economy. His main political economy contributions are on electoral systems and voter participation.

Mattias Polborn 80x108Mattias Polborn– University of Illinois at Urbana-Champaign
Mattias Polborn is Professor of Economics and Political Science at the University of Illinois at Urbana-Champaign. His research centers on the origins and effects of political polarization.

Print Friendly