Sample Surveys in the Real World

4

Sample Surveys in the Real World

63

image
Tetra Images/SuperStock

CASE STUDY An opinion poll talks to 1000 people chosen at random, announces its results, and announces a margin of error. Should we be happy? Perhaps not. Many polls don’t tell the whole truth about their samples. The Pew Research Center for the People and the Press did tell the whole truth about its sampling methods. Here it is.

Most polls are taken by telephone, dialing numbers at random to get a random sample of households. After eliminating fax and business numbers, Pew had to call 2879 residential numbers to get its sample of 1000 people. Here’s the breakdown:

Never answered phone 938
Answered but refused 678
Not eligible: no person aged 18 or older, or language barrier 221
Incomplete interview 42
Complete interview 1000
Total called 2879

Out of 2879 working residential phone numbers, 33% never answered. Of those who answered, 24% refused to talk. The overall rate of nonresponse (people who never answered, refused, or would not complete the interview) was 1658 out of 2879, or 58%. Pew called every number five times over a five-day period, at different times of day and on different days of the week. Many polls call only once, and it is usual to find that more than half of those who answer refuse to talk. Although Pew did obtain the desired sample of 1000 people, can we trust the results of this poll to make conclusions about people with residential phones? Can the results of the poll be extended to people who only have cell phones? By the end of this chapter, you will learn how to answer this question.