We use cookies to help improve and maintain our site. More information.

31 October 2021

About forecasting

Any form of analysis is subject to cognitive biases, my own included. The more successful analysts are those who are aware of them. Last week, I wrote about a particular form of bias, the model bias. Once you have a mathematical model, like for the spread of a pandemic, you may be tempted to take the model more seriously than the reality it tries to explain.

There are many cognitive biases. For political analysts, the one to watch out for is the hindsight bias, or the I-knew-it-all-along bias. The human mind is not well equipped to deal with radical uncertainty, which is not subject to some known probability distribution. When people give you percentage probabilities on future political events, my advice is to walk away. Olaf Scholz's victory at the German election was unpredictable back in March or April. His own bet, as we recall him saying at the time, was that once the German public gets to know all of the candidates, they would choose him, the experienced finance minister. The bet succeeded. But back then, there was no way of knowing that Annalena Baerbock, the co-leader of the Greens, would implode due to a plagiarism scandal. And there was certainly no way of knowing that Armin Laschet, the CDU leader, would giggle during a meeting with flood victims. The fact that Scholz's bet paid off does not mean that it was bound to pay off. If it hadn’t paid off, the hindsight-biased analyst would have told us that it could not have conceivably worked. I concluded some time ago that the best way to look at the future is through scenarios, and events that might trigger them. But you can’t translate these into probabilities.

So what do we know? Before the German elections, we observed a long-running trend of declining share in the SPD vote. Early in the election, the SPD polled only 14%. Scholz gained some 10pp during the campaign. So did Gerhard Schröder in the 2002 and 2005 campaigns, when the SPD looked a likely loser. Then, and now, there is a volatile segment of the German electorate that changes its mind during the campaign. It tends to gyrate as a group. That group favoured Baerbock in April, Laschet in June, and switched to Scholz in August. That support is still holding up. But it could swing to Markus Söder in 2025. Or not.

We also know that the one unbroken trend in European politics is fragmentation. The most successful political parties nowadays score 20-25%, not 40-45% as they used to. Six national-level parties have seats in the German parliament. For the first time, Germany will have a three-party coalition. In the Netherlands, the trend to fragmentation is more advanced. There, elections in March sent 17 political parties to the Dutch parliament. It takes at least four parties to form a coalition. It is true that an overwhelming percentage of Germans want to see Scholz as chancellor. But only 25% voted SPD.

Cognitive biases do not exist in isolation. If you add the fallacy of composition to hindsight bias, you will spin the Scholz victory into a Social Democrat revival in Europe. The leaders of Spain, Portugal, and all of the Nordic countries are Social Democrats. With Scholz in the chancellery, they will be the strongest force in Europe. You may see a trend, maybe a secular trend. Be careful. Elections are zero-sum games. A 0.9% swing would have turned the election the other way. Narrow victories, like Donald Trump’s in 2016, the Brexit referendum, or indeed Scholz's triumph, are important. But they are almost never the beginning of a trend.

So where does this leave political observers? Most will continue to do what they did, extrapolating polls and story-telling. The smart analysts take a step back, in this spirit of this author, who admitted to be

confused as ever, but… confused on a higher level and about more important things.

If you would like us to notify you when a new column appears, please fill out this form.