Lab: Sampling Issues

Each election year, the media rely on exit polls to forecast and analyze election outcomes. Exit polls are surveys of voters immediately after they have cast their votes at their polling place. The polling places (precincts) are randomly selected to represent each state, and for national polls, the nation as a whole. Interviewers give selected voters a questionnaire that takes only a couple of minutes to complete. It asks for whom they voted, about important issues, and about demographics such as gender, age, and race. Participation is voluntary and anonymous. Interviewing starts when the polls open and continue throughout the day until about an hour before they close at night.

In one national election exit poll, voters were selected the following way. Within each state, a complete list of precincts was obtained. The precincts were divided up into counties, and within each county, precincts were selected randomly so that the odds of being selected were proportionate to the number of people who typically voted in that precinct. At each selected precinct, one or two interviewers stood outside and randomly selected roughly one hundred voters during the day as they exited from voting. The interviewers accomplished this task by counting voters as they left the polling place and selecting every voter at a specific interval (say, every third or fifth voter). The interval was chosen so that the approximately 100 interviews were spread evenly over the course of the day. (Note that the selection of voters at polling places approximates simple random sampling.)

1. What is the overall sampling design? Be specific. This is a complex design, involving more than one sampling method.
2. What sampling frame was used in the design?
3. Is stratification incorporated into the design? If so, what was the stratifying variable
4. Some journalists and commentators would like to obtain midday numbers or estimates from the exit polls. Why would midday numbers have a much larger margin of error than the numbers complied when the polls are closed at the end of the day? Give two possible reasons, one involving random sampling error¹ and the other involving sampling bias² .
5. Evaluate the sampling strategy with respect to coverage error³, assuming the target population is all voters in the national election.