WAPOR GUIDELINES


WAPOR GUIDELINES FOR EXIT POLLS AND ELECTION FORECASTS

 Written by WAPOR Exit Poll Committee
 Approved by WAPOR Council, 10/12/06
 
Public opinion is a critical force in shaping and transforming society. Properly conducted and disseminated, survey research provides the public with information from the measurement of opinions and attitudes and allows its voices to be heard. This document provides guidelines for conducting exit polls and making election forecasts on election days.
 

EVALUATING EXIT POLLS AND ELECTION FORECASTS

Exit polls are polls of voters, interviewed after they have voted, and no later than election day. They may include the interviewing before election day of postal, absentee and other early voters. In some countries election day polls can not be conducted at the polling place, but in most cases, interviewing takes place at the polling location.

Exit polls can serve three different functions that are not mutually exclusive: predicting election results; describing patterns of voter support for parties, candidates, and issues; and supporting extensive academic research efforts.   The main difference between these may be the speed with which the results are formulated and disseminated.

Exit polls used for projections should be reported as soon after the polls close as practical. Any delay in disseminating the results will inevitably raise questions about the legitimacy of the effort, especially with regard to estimating the outcome of the election. If analysis is the only purpose of the exit poll, prompt release is less important.

In some countries, election laws prohibit the publication of exit poll data until after the polls have closed. WAPOR and ESOMAR oppose regulation of the conduct and reporting of polls in principle. However, no statement about the outcome of an election based on exit polls should be published before all the polls in the contest have closed. In national elections, this means polls relating to election results for elections in smaller voting units can be reported when all the polling places have closed in those locations, rather than waiting until all polling places used for voting that day have closed.  Descriptive information other than voting behaviour may be published before the polls have closed. 

ETHICAL PRINCIPLES

 Survey researchers in general and those conducting exit polls in particular need to follow certain broad principles in conducting their research:
 
  1. Exit polls conducted for public consumption should be impartial and non-partisan. Exit polls are scientific research designed to collect data and report information on electoral outcomes. They are not tools for partisan advocacy.
  1. Methods should be transparent, public, and well-documented. These goals can be achieved by publicly describing the methods prior to conducting the exit poll and by adhering to the standards of minimal disclosure delineated in this document. It is also recommended that when the exit poll is used for analysis, the data set (without individual identifiers) along with appropriate survey documentation be deposited in public archives and/or on web sites for general access.
  1. Data collectors must adopt study designs for their exit polls that are suitable for producing accurate and reliable results and that follow specific procedural and technical standards stipulated in this document.
  1. When reporting results from exit polls, data collectors and analysts must be careful to keep their interpretations and statements fully consistent with the data. Speculation and commentary should not be labeled as data-based reporting. Limitations and weaknesses in the design of an exit poll, its execution, and the results must be noted in all reports and analysis. Results should be released to the public and other interested parties through the general media and simultaneously made accessible to all.
  1. The identity of respondents in exit polls must be protected. No identifying information (e.g. name, address, or other IDs) should be maintained with the voter-level records, and the data set should not allow deductive disclosure of respondents’ identity. To limit the chances of deductive disclosure, small-area geographic details such as the specific polling place in which votes were cast should not be revealed.

EXIT POLL METHODS AND THEIR DISCLOSURE

Poll methods must be generally accepted as goodsurveypractice and must be disclosed in advance of the conduct of the exit poll, as well as with any projection or analysis or subsequent public release of the dataset.

Items for Minimal Disclosure: These items should be disclosed with any exit poll report or when any projection is made.  Good practice would be to disclose as much of the methodology in advance as possible, particularly those items marked with an asterisk, which should be disclosed before Election Day.

*Sponsor of the exit poll 
*Name of the polling company or principal researcher; prior experience (if any) in exit polling; and whether the data collector has any business or personal ties to political parties, candidates, political organizations or governmental bodies.
*Name of the organization responsible for analysis and projections, if different.
Number of interviews
*Number of sampling points
*Sampling frame 
*Geographic dispersion and coverage 
*How sampling points are selected 
*Where interviews are conducted: at polling places, in person at homes, by phone, etc.
*Any legal limits on data collection that might affect polling accuracy (e.g., minimum distance of interviewers from the polling place)
Time of day of interviewing
Whether interviewers are part of a permanent field staff or hired for the occasion
*How respondent anonymity is guaranteed (paper questionnaires, etc.)
The interview schedule or questionnaire and instructions
Which results are based on parts of the sample, rather than the whole sample
A description of the precision of the findings, including estimates of sampling error
Monitoring and validation procedures (if any)
Weighting procedures
Response rates (using one of the definitions in the AAPOR/WAPOR “Standard Definitions: Final Dispositions of Case Codes and Outcome Rates for Surveys”) and item non-response on vote questions
Any known nonresponse bias
General description of how estimates are made and the kinds of variables that are being used, and whether adjustments for non-response have been made
Known design effects 

Political parties may sometimes make claims about private data. These claims also require documentation. Any public statement referring to exit poll results should abide by these disclosure principles and requirements. 

Good Practices: Those conducting exit polls should always use generally accepted scientific methods.   However, there are a number of good practices that apply specifically to exit polls.

Exit polls typically employ clustering in their sample designs. Because of the possibilities that various groups might attempt to influence voters and/or exit poll respondents, exit poll researchers are not expected to disclose the actual sample points or locations.

Exit polls should collect information across the whole of the polling day.  Probability sampling (or full census) for interviews conducted at the polling place is the only acceptable selection method. Quotas are not appropriate for sampling at the polling place.

A national exit poll should represent the entire country, with 95% of the target population included in the sampling frame. If the sampling frame covers less than 95% of the target population, there should be an explanation for that decision.

Exit pollsters should keep in mind the relationship between small units for which the votes are tabulated and that can also serve as clusters for exit poll interviews. One way to evaluate an exit poll is to compare the actual election results and the estimates derived from the exit poll interviews for these same units. This comparison of small unit accuracy, typically at the precinct or polling place level is one of the best ways to understand the exit poll’s success.   But there are situations where this will not be possible, either because no tabulations are reported at the smallest voting unit level or because the sampling units do not coincide with voting units.
 

ELECTION PROJECTION METHODS AND THEIR DISCLOSURE 

Election projections can be made in other ways than by interviewing voters as they exit the polling place. While most projections are based on interviews with voters after they have voted at a polling place, other forecasting models may include:

-- interviews in person, by telephone, or by other means of communication with voters after or before having cast their votes

-- counts of official votes in a sample of precincts, often known as quick counts

-- a mix of methods 

A projection is an estimate that leads to a conclusion about the outcome of an election in a jurisdiction such as a nation, a state or a district. This may occur in two different situations:
·         If the winner is based on the popular vote for an office or a party, then a projection of the division of that vote is a projection of the outcome in the jurisdiction.
·         If the winner is based on the vote in multiple jurisdictions, such as election of a Parliament where votes are cast in districts or of a President where votes are accumulated based on victories won in many jurisdictions, a conclusion about which party has a plurality of seats in the new Parliament or which presidential candidate has a winning number of votes is a national projection.

The projection need not reach a conclusion about each sub-jurisdiction. It need only reach a conclusion about the jurisdiction outcome.

The objective of any projection is a conclusion about an election for some jurisdiction. A sample of that jurisdiction must be adequate to reach an unbiased conclusion with sufficient/appropriate confidence in the estimate. A national projection typically requires the coverage of the entire country, with at least 95% of the target population in the sampling frame.

 There will be times that a subset of the country will be used (for example, only competitive districts). But if a sampling frame is used that includes something less than the entire voting population of a jurisdiction then the pollster should define what is and is not included in the sampling frame in a disclosure statement. The pollster also must publish a rationale to justify the pollster’s ability to make an unbiased conclusion about the election outcome based upon collecting information from a subset of all jurisdictions.

WAPOR EXIT POLL COMMITTEE MEMBERS

Chair: Kathleen Frankovic, former WAPOR President, US
Director of Surveys, CBS News

Nick Moon, UK
Director, GfK NOP Social Research

Richard Hilmer, Germany
Managing Director, Infratest dimap

Mahar Mangahas, Philippines
President, Social Weather Stations

Alejandro Moreno, Mexico
Department of Public Opinion Polling Head, Reforma

Anna Andreenkova, Russia
Co-Director, CESSI

Warren Mitofsky, US (until his death in September)
President, Mitofsky International

Ex officio: Tom Smith, WAPOR Professional Standards Chair
Michael Traugott, WAPOR Vice President