DeutschlandTREND: Die wirtschaftliche Lage

1
Exit Polls in Germany
Richard Hilmer Managing Director Infratest
dimap 3MC 2008 Session 28 Exit Poll Berlin,
26th of June 2008
2
History of Exit Polls in Germany

• 1978, June First Exit Poll by Infas for the
  ARD (September 1978 first Exit Poll by
  Forschungsgruppe Wahlen (FGW) for the ZDF)

• 1979 Abandonment because of political debate

• 1990 Restart Exit Polls by Infas and FGW for
  all national and state elections

• 1996 Infratest dimap took over responsibility of
  Exit Polls for the ARD. Exit Polls realized
  since 1997

• 3 for national elections
• 2 for european elections
• 40 for state elections
• 30 for local elections
• _______________________
• 75 Total

3
Targets of Exit Polls and how to reach them

• External Factors
• Voting System
• Official Support
• Party-System
• Turnout
• Absentee voting
• Competitive Environment

• Internal Factors
• Sampling
• Field work
• Selection process
• Questionnaire
• Modelling

• Main Targets
• measure voting behaviour of electorate and
  selected subgroups
• identify major motives
• analyse the voter flows
• precise prediction of the results in real time
• sideeffect showcase for capacity and efficiency
  of MR

4
Targets of Exit Polls and how to reach them

• External Factors
• Voting System
• Official Support
• Party-System
• Turnout
• Absentee voting
• Competitive Environment

• Internal Factors
• Sampling
• Questionnaire
• Field work
• Selection process
• Modelling

• Main Targets
• measure voting behaviour of electorate and
  selected subgroups
• identify major motives
• analyse the voter flows
• precise prediction of the results in real time
• sideeffect showcase for capacity and efficiency
  of MR

5
Optimizing Exit Polls field work and data
processing

• adequate tools Questionnaire fitting to the
  voting system

6
Optimizing Exit Polls field work and data
processing
7
Optimizing Exit Polls field work and data
processing
8
Optimizing Exit Polls field work and data
processing

• adequate tools (fitting to the voting
  system)
• intensive interviewer training and support
• trouble avoiding/trouble shooting program
  (precontacting officials / back up teams for
  election day)
• maximum of transparancy and anonymity in
  surveys

9
Optimizing Exit Polls field work and data
processing
ARD-Election-Box
10
Optimizing Exit Polls field work and data
processing

• adequate tools (fitting to the voting system)
• intensive interviewer training and support
• trouble avoiding/trouble shooting program
  (precontacting officials/ back up teams for
  election day)
• maximum of transparancy and anonymity in
  surveys
• data delivery-processing
• absolute reliability
• automatic checking procedures

11
Optimizing Accuracy of Prognoses
12
Exit Polls New Challenges

• Declining Turn Outs
• Increasing number of absentee voters
• Increasing number of late deciders
• Especially problematic for pre-election surveys
• but also for comparison of precinct and absentee
  voting in Exit Polls.

• Voting systems become more and more complicate
• Structural bias in sampling despite of a
  response rate of 70 to 80

13
Optimizing Exit Polls - Sampling
1. Level Precincts (national election appr.
80.000)

• Random sampling
• Regional stratification
• Representation of past election result
• Differentiated coverage of strongholds weak
  spots of major parties

(s)
14
Optimizing Exit Polls Selection of Voters
2. Level Selection of Voters
old approach

• sample every K-th voter

15
Optimizing Exit Polls Selection of Voters
2. Level Selection of Voters
New approach

• full voters survey (per time slice)
• Avoids structural bias
• Increase number of interviews (BTW 100.000
  instead of 20.000)
• Delivers more precise results and structures
• Better representation of precinct results
• Minimizes need for weighting/modelling

16
Optimizing Accuracy of Prognoses
17
Optimizing Accuracy of Prognoses
State Election in Hamburg 2008




18
Thank you!