Nate Silver is a statistician, writer, and the founder and editor-in-chief of the well-known blog FiveThirtyEight, who rose to fame in the political area through his incredibly accurate data-driven predictions of U.S. Presidential election results. Before the 2012 U.S. Presidential election Mr. Silver correctly predicted the outcomes in all 50 states and the District of Columbia, leading to effusive praise such as “Silver definitively proved that geeks with mathematical models were superior to the gut feelings and pseudo-statistics of so-called political experts.”
The same tools used to forecast U.S. elections can be used to predict voting behavior in parliamentary elections, and thus the share of seats won by various parties. Given this, why haven’t political scientists or statisticians followed Mr. Silver’s path to fame by predicting the winners of parliamentary elections?
To understand this, it is helpful to break down a forecast of an “election winner” into two parts. First, how many votes will each candidate or party win? Second, how do the electoral rules and institutions of the country translate these votes into political power?
In forecasting the winners in U.S. Presidential elections forecasters really only have to deal with the first of these questions. Once forecasts of the popular vote in the 50 states and Washington D.C. have been generated, the answer to the second question is simply a matter of allocating the predicted Electoral College votes to each candidate and determining the winner.
However, in parliamentary democracies, the number of seats won by each of the parties is only the beginning of the process of determining the election winner. After the election, unless one party wins an outright majority of seats (a rare occurrence), those parties that won seats in parliament negotiate to form a government, sometimes with surprising results. Thus, in order to forecast the results of a parliamentary election we must make two predictions – how many votes (and thus seats) will each party win, and then which of those parties will successfully join together in a governing coalition.
For example, consider the recent attempt by Chris Hanretty, Ben Lauderdale and Nick Vivyan to predict the results of the 2015 U.K. Parliamentary election using the methods pioneered by Nate Silver. Perhaps because polling at the constituent level is still in its infancy, their model failed to predict that the Conservative Party would win an outright majority of seats and thus form a majority government. However, even if their prediction of a “hung parliament” where no party holds a majority was accurate, in order to predict the election winners they would need to predict which parties would join in a governing coalition.
Our research shows that predicting governing coalitions is a daunting task in parliamentary democracies. In our recent article in BJPolS, “A New Approach to the Study of Parties Entering Government,” we develop a new approach for estimating the probability of a party entering the government. Our model takes each party’s seats shares as given, and thus is focused on the second prediction that would be needed to forecast the outcome of a parliamentary election.
The key point in our study is that we cannot predict the probability of a party entering government in isolation. Whether a party enters the government depends not only on the characteristics of that party, but also on the characteristics of the potential coalitions of which it is a member. In our BJPolS article, we note that when we ask how party characteristics – such as legislative seat share – affect the probability that a party will join the government, we are really asking how these party characteristics will affect the bargaining process to form a government. The government formation literature acknowledges the fact that parties do not enter government independently – the identities and preferences of the other parties involved in the government formation process also matter. For example, the leaders of other parties might try to exclude a party from joining the government because the party defected from a previous governing coalition, or because conflicts among a party’s internal factions make it seem like an unreliable partner. Alternatively, parties with previous experience in government should be more welcome to join new governments if they have a reputation for being reliable coalition partners.
All of this makes predicting the ultimate “winners” in parliamentary elections much more difficult than simply predicting seat shares. In theory, the approach we describe in our BJPolS article could be combined with election forecasts to produce a workable data-driven approach to forecasting both seat shares and the resulting governing coalition. However, given the additional complexity and uncertainty inherent in this forecasting approach, it is doubtful that anyone will be able to replicate Nate Silver’s U.S. track record in parliamentary systems.
Garrett Glasgow and Sona N. Golder (2015). A New Approach to the Study of Parties Entering Government. British Journal of Political Science, 45, pp 739-754. doi:10.1017/S0007123414000015.