Sponsored by The Universities at Shady Grove and Rockville Institute

The Real Scoop on Exit Polling:
What the Election Night Prediction Business Is Like from the Inside

Thursday, October 30, 2008 at 4:30 p.m.
Free and Open to the Public

The Universities at Shady Grove
The Camille Kendall Academic Center

Building III, Room 3241
9636 Gudelsky Drive
Rockville, MD 20850
Directions | Map

The Issue

Every 2 years exit polls achieve a brief place in the sun for the American public at large. To varying degrees of success, television journalists attempt to convey rudimentary statistical concepts to the public, but viewers are initiated to the details.

Back to top

What We Know

A Little History

For about 2 decades news organizations have funded the shared set of election polls that produce the basic data, but networks and other subscribers maintain editorial control over how they interpret the results.

The exit polls are not the only data source for predicting outcomes. Vote tallies in sample precincts and then preliminary tallies by state governments eventually dominate the predictive models.

Vulnerability of Polls

In the 2000 Presidential election, Voter News Services and all of the major networks first awarded Florida to Gore relatively early in the evening. Although the actual voting intention of the majority of Florida voters is still disputed by some, the call based on exit poll data was premature, particularly because an exit poll bias toward Gore appeared in this state and a number of others.

Exit poll bias has become an increasing issue, leading the networks to become increasingly wary of interpreting their results when the exit polls show a race to be relatively close.

Studying the Poll Data

Exit polls provide an interesting test case for survey research. While their value as election predictors fades in a matter of hours, their conclusions can be compared to the accepted standard soon thereafter. Few other surveys are subject to as quick and decisive a reality check. The possible sources of exit poll bias can be listed, but the degree to which each of them affects the size of the bias is poorly understood.

Back to top

Goals for the Presentation

The presentation will cover:

  • The role of exit polls as part of the larger statistical system used to make election night predictions;
  • Possible sources of bias and concern over whether bias is increasing;
  • Anecdotes about what the election night prediction business is like from the inside; and
  • Why, for a statistician, election night is a challenging, engaging, and fun opportunity.

Back to top


Blair Lee, CEO, LDG, Inc. Mr. Lee has worked on Capitol Hill and has managed political campaigns. From 1978 to 1984 he was Montgomery County's chief lobbyist in Annapolis. Mr. Lee also writes a weekly newspaper column for The Gazette newspapers and is a regular political commentator for WBAL radio.

Robert Fay, PhD, Senior Statistician, Westat. Dr. Fay retired in January after 34 years with the U.S. Census Bureau. He now works in Westat's Statistical Group. For about 98% of his career, Dr. Fay has worked on the statistical problems of survey research: survey design, estimation, problems of missing data, variance estimation, and the analysis of survey data. The remaining 2%, however, is of considerably more immediate interest to family, friends, and, hopefully, the audience of the seminar. Beginning in 1978, he has been involved in election night predictions every 2 years, working at CBS, Voter News Services, and now CNN.

Back to top


Flyer (PDF)

Back to top