DE   |   EN
06.10.2023 from 
Research + Transfer
How reliable are forecasts in elections?

On Sunday, October 8, there will be another election in Germany: This time, it's about the state parliaments in Bavaria and Hesse. If you look at the media coverage, you regularly come across the first figures on how parties might perform. But what exactly do these figures actually say? How much should we rely on them? And are forecasts, projections and co all the same? Political scientist Dr. Benjamin Höhne talks about this in an interview.

As mentioned, state elections are coming up again at the moment - and the media are busy reporting on the possible percentages of the parties. How and where do these projections come about?

We should differentiate: There are projections, the Sunday polls and forecasts. What you're thinking of is probably a forecast, for example, of the parties' results in the state elections next Sunday in Bavaria and Hesse. A forecast is to some extent in a gray area. This is due to the fact that voting behavior has changed considerably in recent years, and people are making decisions about which party to vote for at increasingly short notice. Some only do so in the voting booth. For us as researchers, this means that it is becoming more difficult to make serious forecasts. In addition, there is the extrapolation that you mentioned. It is very precise. It is a statistical procedure, an extrapolation of the election result that is anticipated. Projections are known as the "6:00 p.m. forecast." On election day, there is the first election result at 6 p.m., and that is a projection. This is based on an election day poll. Teams from research institutes stand at selected polling places and randomly survey people coming out about how they voted. These results, which are summarized for various constituencies, are usually very close to the final election results. Unlike forecasting, this is a highly reputable thing. There are several institutes that offer this. For example, Infratest dimap for ARD's Deutschlandtrend and Forschungsgruppe Wahlen for ZDF. Both of Germany's main public broadcasters have an interest in delivering data that is as accurate as possible. In addition, there are providers who sometimes don't let themselves be looked at so closely in the cards and claim that in the next election, party X will arrive at a certain percentage value Y.

The ARD Deutschlandtrend now also said some time ago that the AfD had peaked at 21% and was the second strongest party - is this then a similar process?

This was a classic Sunday question. Voters are asked which party they would vote for if a state, federal or European election were held next Sunday. Sometimes you're pretty close, but some things can change again. In addition, these Sunday polls have a bit of a problem in that they only use very small sample sizes. As a rule, they operate with 1,000 people, most of whom are contacted by telephone. There is a heavy reliance on landlines. However, young people in particular increasingly rarely have a landline. They try to reach them via cell phones, which raises the problem of regional assignment. When you call a cell phone, you don't know at first: Are you in Magdeburg, in Halle or somewhere else entirely?Even if the institutes have come up with something to deal with this, the problem of the accessibility of certain groups still remains.Certain groups refuse to be interviewed.We know, for example, that there is a certain reluctance to vote for an extreme right-wing party.That's why there are sometimes surprises in this area. People are less willing to declare to a person on the phone that they vote for an extreme right-wing party.With the AfD, this pattern no longer seems to apply:Right-wing populism is currently experiencing a demoscopic and electoral upswing, which is even stronger in eastern Germany than in western Germany.

Do all three calculations, i.e. projection, forecast, classic Sunday poll, have the same implications and repercussions for politics and society?

First of all, they have an information function. They serve to provide information at an early stage. But of course they also have political implications. If I, as a voter, see that the party I favor is below the five-percent hurdle that usually applies to state elections, but also to federal elections, then that can lead me to perhaps not vote for that party after all and opt for the second-best option. Looking at Hesse, but also Bavaria, we see that the Left Party, for example, is well below the five-percent hurdle and is therefore highly unlikely to be part of the next state parliament. If this becomes apparent in advance, it certainly does not motivate some people to vote for the Left Party. In the projections, which present the first results on election Sunday with the 6:00 p.m. forecast, information is specifically passed on to certain circles in advance. Members of the media, as well as politicians, are already informed about this in the afternoon: What is the trend? Which result is emerging? In Bavaria, particularly interesting this time: How strong is the AfD? How strong are the Free Voters? Media people can plan their program accordingly: Who will be invited? Who gets how much space for an interview? Politicians prepare on this basis to go to the starting line with initial statements after the polls close. Then it's a question of interpretive sovereignty: How is the result to be interpreted? Who won, who lost?

What purpose do these preliminary calculations actually serve? For example, to motivate people to go to the polls? Isn't there a danger here of influencing voting behavior in advance?

You can't deny that it has political implications. If we talk about Bavaria, for example, we see the CSU at around 36%. That's rather difficult for them. The CSU, as a people's party, is traditionally used to other numbers, but like other people's parties, it is struggling with the decline. In other words, people are turning away from the major parties. Also interesting in this context is the scandal involving the diatribe by Mr. Aiwanger, the chairman of the Free Voters. He has turned the accusations that were made against him around in such a way that he has placed himself in a victim role. You can now see demoscopically that the Free Voters have obviously benefited from this gambit, because they ended up at 16%, in other words, they gained ground. Since then, you can see how the parties are thinking about how to deal with this, how to react to it, what else they can do.

What factors influence such forecasts, projections and the like?

In election research, there are three main factors for voting behavior. The first is party identification. That is, people identify with a party, as a regular voter. This is a kind of "psychological party membership." Examples are the seasoned Social Democrat in the Ruhr area who votes for the SPD, or in Bavaria the Catholic churchgoer who votes for the CSU. These are typical core voter figures. Party identification is an important reason to go to the polls and vote for one's traditional party again. This happens more or less automatically. There is a great loyalty to this party. People always vote for the same party for years. Maybe their father, mother or grandparents voted for them. This is nothing unusual, especially in social democracy, which is the oldest party in Germany. But we see that this strong attachment to parties is weakening. It is less present among young people. In eastern Germany, it was less there in the first place, because it could not develop so quickly after 1990. In this respect, other factors are becoming more important.

The second factor is a party's platform, which is evaluated retrospectively and prospectively. You look at what a party actually implemented in parliament, what the core political issues were. Prospectively, the election programs are important: What do you promise? What does a party want to achieve?

Then we would be at the third factor. That is personality. These are the people running for election, especially the bandwagoners at the top, less so the candidates in the constituencies. The top personnel at the federal level are also important, as are those in the state in the case of state elections. In the last elections, the prime ministers always had a big incumbency bonus. These factors play a role when people are surveyed by a polling institute.

So do these figures actually reflect a real situation at a certain point in time? On September 28, the ARD-BayernTrend said that the CSU would reach 36% - will that then almost certainly be the result on October 8?

No. This is always just a snapshot on the day the poll takes place. A lot can still happen between now and the actual election. People with high party identification will probably actually vote on election Sunday as they indicated. For the others, it is uncertain. It is possible that they will still move in a different direction on the last steps to the polling station. This is interesting again because absentee voting is also on the rise.

With such statistics, there are often graphs that show that, for example, regular CSU voters have switched to the AfD. How is something like this reliably investigated? Where do such figures come from?

These are less likely to be regular voters. By definition, they stay with one party. But we see that party ties are becoming more fragile, both quantitatively and qualitatively. Quantitatively, the number of people who identify themselves as regular voters is decreasing. But we also see among those who have a strong attachment to a party that it can weaken. That's the qualitative aspect. Then we talk about the swing voter. They decide from election to election which party they will vote for, based on factors we've talked about. Specifically, the way this works through election day surveys is that when people come out of the voting booth, they are asked which party you voted for and which party you voted for in the election before last. From that, voter migration balances are obtained. Basically, there are such changes among all parties. We see, for example, that with regard to the AfD, it is not the case that voters only switch from the CDU, but also among the other parties. This also affects the Left Party, the SPD, but virtually not at all or only very little the Greens, because the Greens and the AfD are diametrically opposed. The voter migration balances also reflect that the AfD is able to mobilize from the non-voter spectrum. In other words, it is bringing people back into the system who did not vote last time or perhaps have not voted at all for some time.

Should we really rely on such forecasts as the ARD Deutschlandtrend?

I would generally recommend not relying on forecasts, trends and Sunday polls, but forming your own opinion. There are many ways to inform yourself. The parties are more active than ever in the election campaigns. There are booths where you can meet the politicians and perhaps pick up a pen or notepad. You can talk to the parties directly. There is also the Wahl-Kompass or Wahl-O-Mat from the Federal Agency for Civic Education. This is particularly interesting for young people. Users are asked various questions and then, based on their answers, a certain degree of fit with the various parties is indicated. This provides an orientation aid for people who don't have the leisure to take a close look at election programs.

What would be your tip: If you want to inform yourself and get a real impression of the current percentages of the parties, how should you proceed? Which figures do you look at?

I would say you can look at the Sunday poll, it's a sentiment poll. You can also look at the last election results. But I would advise you not to look at the Sunday polls at all, but to form your own opinion, to inform yourself well and, in the end, to vote on the basis of your own values and the promises that the parties make.


Author: Lisa Baaske