The exit polls of 2018

Ipsos conducts exit polls during almost all elections. 2018, however, will be a special year, because on 21st of March we will be conducting (on behalf of the NOS) not one or two, but seven exit polls. Six of them concern the elections for the municipalities. Based on these exit polls, we are able to give an indication of the turnout and the results in the cities of Amsterdam, Rotterdam, Utrecht, Enschede, Weert and Emmen. The seventh exit poll concerns the referendum on the Wiv. This Referendum takes place on the same day as the elections for the municipalities.

Exit PollExit polls in 6 municipalities
The number of polling stations where the exit poll is conducted differs per municipality. In Amsterdam we are present at 16 polling stations. In Rotterdam, Utrecht and Enschede the number of polling stations is 15. When we add these numbers to the 12 polling stations in Emmen and the 10 polling stations in Weert, we arrive at a total of 83 polling stations. For these polling stations together about 1.6 million people entitled to vote for the municipality are invited. Together the polling stations where the exit poll is conducted form a good cross section of all the municipalities’ polling stations.

The exit polls for the municipalities result in an indication of the turnout and the election results (in percentages as well as in seats in the city council). Here the applicable research margins must be taken into account. Since the number of seats to be divided per city council is lower than for instance in the Tweede Kamer, the average margin in seats equals 1 seat. In actual practice, there may also prove to be a difference of 2 seats per party. For example, in Rotterdam and Amsterdam the chances of a difference of 2 seats are bigger due to the bigger variation in voting behaviour.

Country-wide exit poll on the Referendum
Alongside the six municipal exit polls, we also conduct a country-wide exit poll on the Referendum. For this exit poll, 21 polling stations across the country are selected. Together, these polling stations form a good cross section of all the polling stations in the Netherlands. The total number of people entitled to vote for the Referendum at these polling stations is about 35,000.

The results of the Referendum exit poll give an indication of the turnout and the election results for the referendum in percentages. An average risk level of 5% is applicable for the turnout and of 5% for the election results for the referendum. This means that if a measured turnout is 45% the actual turnout may prove to be 5% higher or lower. This same is true for the election results.

How does an exit poll work?
Voting once more
An exit poll is a shadow poll. Our interviewers ask voters when leaving the polling station if they’d like to ‘vote’ once more. The voter is given a ballot paper with the same choices as on the official ballot paper. Then the voters can anonymously cast their vote in the ‘ballot box’ of our interviewers.

Important: selecting the polling stations
A major difference with normal election surveys is how the base is built: for an exit poll we do not draw a sample of voters but of polling stations. Let’s take the General Elections of 2017 as an example. At that time, in the Netherlands, there were almost thirteen million people who were entitled to vote at about 9,300 polling stations. From these, we selected 43 polling stations. For these polling stations a total of about 67,000 people who were entitled to vote were invited, of whom about 82 per cent turned out and cast their vote. Thus, we realised the 2017 exit poll based on almost 55,000 voters. An impressive number. But still it doesn’t prove that much, as we should not forget that in fact it represents merely 43 of the 9,300 polling stations. We collected data among 55,000 people who are entitled to vote, but the pre-selection can have a huge impact.

Such a small sample is surely never enough?
That’s why a statistical analysis model offers a solution. By analysing the voting history of the polling station, we can still arrive at an accurate indication of the election results – despite the limited sample. How does our analysis model work?

The essence of the analysis is all about the comparison with the previous elections. We know for each polling station what people voted in those previous elections. By comparing the results of the exit poll with the results of then, we can determine what the trend is: which parties win and which ones lose. We project this trend on the Netherlands as a whole.

Something different than a final poll
An exit poll is based on actual voting behaviour. This is a crucial difference with the election survey that is held prior to the election day, the so-called final polls. This is because final polls are based on people’s voting intention, which does not always correspond well with their actual behaviour. There is a world of difference between ‘saying’ and ‘doing’ something.

Recent Ipsos exit polls
A more accurate polling survey than an exit poll doesn’t exist. The most recent exit poll Ipsos conducted was the one on the General Elections (15 March, 2017). That exit poll proved to offer a good indication of the turnout (which was exactly the same, namely 82%) and the results.

Exit poll vs. results, General Elections 2017

 

In percentages

In seats

 

Exit Poll

Result

Deviation

Exit Poll

Result

Deviation

VVD

20.3%

21.3%

1%

31

33

2

PVV

12.2%

13.1%

-0.3%

19

20

1

CDA

12.4%

12.5%

0.1%

19

19

0

D66

12.6%

12%

-0.6%

19

19

0

SP

9.4%

9.1%

-0.3%

14

14

0

GroenLinks

10.2%

8.9%

-1.3%

16

14

2

PvdA

6%

5.7%

-0.3%

9

9

0

ChristenUnie

3.7%

3.4%

-0.3%

6

5

1

50Plus

3%

3.1%

0.1%

4

4

0

Partij voor de Dieren

3.3%

3%

-0.3%

5

5

0

SGP

2%

2.1%

0.1%

3

3

0

DENK

2.1%

2%

-0.1%

3

3

0

Forum voor Democratie

1.4%

1.8%

0.4%

2

2

0

Overig

1.4%

2%

0.6%

0

0

0

Totaal

100%

100%

 

150

150

6

Also on the Referendum on the Ukraine-treaty in 2016, the exit poll proved to offer a good indication of the turnout and the results.

Exit poll vs. results, Referendum on the Ukraine-treaty 2016

Percentage no-voters exit poll

Actual percentage no-voters

Exit Poll turnout

Actual turnout

64%

61%

32%

32%

When can we be satisfied (again)?
Just like all the other Dutch, on election night we await the eventual results. Not just because we’re curious about the actual distribution of the seats, but especially to determine if the exit poll has done its job properly as it always does in other instances. Although – for several reasons - the exit poll is an extremely reliable instrument, it remains a sample-based survey.

 

Society