Prediction Markets and the Ballot of March 11: A First Evaluation

A guest post by Oliver Strijbis, Sveinung Arnesen, Kjetil Thuen, and Lucas Rachow. Oliver Strijbis is a postdoctoral researcher at the University of Hamburg, Sveinung Arnesen is a postdoctoral researcher at the University of Bergen. Together with Kjetil Thuen and Lucas Rachow they are founders of politikprognosen.ch.

****

Shortly before the voters went to the polls on March 11 we published on this blog the predictions taken from our prediction market. In short, the accuracy of our forecasts were mixed. Whether the predictions were accurate or not largely depended on the proposal. But let’s look at it in detail.

The good news is that the predictions for the ballot on the “Buchpreisbindung” and the “Bauspar-Initiative” were strikes. The prediction from March 7 (the latest prediction we published) was that the share of Yes-votes for the Buchpreisbindung would be 45.5% while it was 43.9% (1.6% deviation). The share of Yes-votes for the Bauspar-Initiative was expected to be 46.7% while it was 44.2% (2.5%).

Our predictions for the other proposals were less accurate. For the ballot on the “Zweitwohnungen” we predicted 45.5% of Yes-votes – deviating 5.1% from the final result of 50.6%. Hence, we did not predict the right winner in this instance. Our predictions for the referendum on the law on “Geldspiele” and the “6 Wochen Ferien für alle” initiative have foreseen the winner but were rather far off from the final results: while our prediction was 70.5% of “Yes” for the Geldspiele the result was 87.1%, and while we foresaw the share of Yes-votes for “6 Wochen Ferien für alle” to be 39.9% it was only 33.5%.

Why have our predictions concerning some ballots been far more accurate than others? It is generally argued that the accuracy of the predictions depend on what kind of relevant information the traders possess about the topic, and to which degree the traders are motivated and able to express their beliefs in the prediction market framework (e.g. Sunstein 2006). Based on these theoretical assumptions we would like to discuss a couple of points that might explain variance in the accuracy of the predictions of the March 11 ballots.

The proposals. Predictions are easier if there is abundant information on a proposal and if the campaign follows a well-known pattern. The fact that not less than five proposals were voted on at the same day had the effect that media coverage on some of the proposals was scarce. Furthermore, the initiative on the “Zweitwohnungen” and to some degree also the referendum on the law on “Geldspiele” did not produce constellations of political alliances which are typical for direct democratic decisions in Switzerland. Hence, experience helped only partially for the formation of well-informed expectations. However, in the case of the initiative on “6 Wochen Ferien für alle” the pattern followed a clear left-right division that has been experienced in many cases in the past. This means that also features concerning the topic and the number of the proposals can only partly explain the variance in the accuracy of our predictions.

The participants. Prediction markets work on the basis of the relevant information accumulated by the participants. We have tried to guarantee that our market profits from well-informed participants by recruiting mainly among social scientists in general and political scientists in particular. However, for many of them the mechanisms of a prediction market might be new. Furthermore, some of them might participate more out of goodwill than intrinsic motivation.

While some of our predictions have pointed to the potential of prediction markets for forecasting the results of direct democratic decisions in Switzerland, there remains room for improvement. Although, due to the often very short and spiritless campaigns, predictions on Swiss direct democratic decisions might always be fraught with risk, we think that this improvement is possible. Since the participants are undergoing a learning process and since in the future participants can be substituted we are confident that we can achieve more accurate predictions in a not too distant future. Even more, competent traders will also over time accumulate wealth in the market and thus become more powerful than those making wrong trades. In other words, we expect more influence to be transferred to those with good judgment, and less to those with bad judgment. Much will now depend on our ability to gain the most informed and motivated participants.

One thought on “Prediction Markets and the Ballot of March 11: A First Evaluation

  1. Pingback: Zoon Politicon

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s