ABSTRACT
David Gold is correct in calling attention to the author virtual neglect of tests of significance in descriptive surveys. As Gold says, the tests have become a form of ritual behavior: Reciting the magic phrase “significant at the.01 level” is often a substitute for hard thinking about the quality of one’s data. If one is concerned simply with establishing the existence of a difference in political interest between urban and rural residents, then there may indeed be some value in testing statistically the null hypothesis that there is no difference. Again, there is the problem of interviewer bias: It may be that the reported difference between urban and rural residents stems from the behavior of a few interviewers. It is possible to control this source of error by randomizing the assignment of interviewers to respondents, so that the possible biases of a few interviewers are not concentrated in one group of respondents.
