Are INSA's polling numbers politically intended?

The German polling institute INSA often has the most spectacular numbers. That's why it is cited frequently. But its political proximity to a right-wing party is obvious. And this connection seems to have an effect.

by Marcel Pauly  |  August 4, 2017

In March 2012 INSA started polling people's voting intentions in Germany. Since then the institute publishes new numbers every week and they get cited by the press quite often. The reason for that is INSA's numbers are interesting: ofter they are a little more extreme than the numbers by other polling institutes. For example, INSA was the only institute that rated the Social Democratic Party with less than 20 percent over weeks in 2016.

The institute does also have a special relationship to Chancellor Angela Merkel's Christian Democratic Union (CDU), its sister party CSU and to the right-wing populist party AfD: INSA's CEO Hermann Binkert used to be a CDU member and he worked for the federal state government in Thuringia. In September 2014 he resigned from the party and one month later he offered the just elected AfD group in the Thuringian state parliament to consult them. From that point INSA's polling numbers for the AfD started rising notably above the median of all institutes.

The deviation from the other institutes became permanent with the European refugee crisis. In September 2015 Merkel decided allowing refugees to cross the border from Hungary into Austria and onward to Germany. In the weeks and months after that the CDU/CSU ratings went down while the AfD's increased. And in INSA's numbers this development was even stronger: CDU/CSU dropped temporarily under 30 percent and the AfD reached more than 15 percent.

At the most extrem point the deviation between INSA's CDU/CSU rating and the CDU/CSU median by all institutes was 3.6 percentage points. For the AfD the highest deviation was 2.7 percentage points. These numbers are above the usual margin of error.

Exceeding the margin of error is not a problem per se. The polling institutes' goal is to be within the margin in 95 percent of the time. Assuming that the median of all polling institutes is quite close to the actual public opinion this means that a single institute's numbers for a certain party should not deviate from the median by more than the margin of error in more than 4 out of the past 65 months.

INSA is the only institute that produced an statistically unexpectable number of high deviations for as many as three parties. It significantly overrated the CDU/CSU in 17 out of 65 months and overrated the AfD in 10 months. Besides that the party Die Linke got overrated in 9 months.

All polling data was gathered by Wahlrecht.de. This analysis works with the institutes' monthly means for each party and with the medians of the monthly means of all seven German polling institutes listed by Wahlrecht.de. Please find all the data and the Python script that analysed it in this project's GitHub repository. And have a look at my other projects.