The Mechanics and Problems of Sampling
- Self-selected polls are meaningless. People can "stuff the ballot box."
- Random sample: everyone in the population has the same chance of being in the study.
- At first, pollsters went door to door. During the 1970s, they switched to phones via random-digit dialing.
- But now most people have caller-ID, either through cellphones or landlines. And most people won't answer calls from unknown numbers.
- Internet surveys -- a partial remedy. But not everybody uses the Internet.
- Partisan nonresponse bias.
- "Weighting"
- Margin of Error/Confidence Interval. Example.
- A video
Election Polls
- Identifying likely voters is tricky.
- In 1948, pollsters stopped too early and missed last-minute shifts.
- Exit polls
- All respondents are voters;
- Respondents do not know the outcome.
- Huge sample size allows for analysis of subgroups.
- BUT more and more people vote early, so pollsters have to stake out early-voting sites and call mail voters.
Knowledge and Honesty
Focus Groups -- a supplement to surveys
