In class on Thursday, someone asked a question about footnote 11 on page 235 of Chapter 8 of Voice and Equality by Verba, Schlozman & Brady. I deferred because you all did not have copies of Appendix A, which describes the procedure. Here's the answer to the question regarding the meaning of that footnote:
Sometimes a sample does not reflect the underlying population. For instance, very slightly over half of people in the U.S. are female. But in many surveys, more women than men agree to be interviewed. (Why do you think this is the case?) If we count all the people who respond to such a survey equally, the results will reflect the answers of women more than a random sample of the U.S. population should. It would overcount women because men are underrepresented in the sample. To compensate, we might statistically given more weight to the answers of the men we were able to interview. In essence, this technique overcounts the answers of men to make up for the fact that men were under-counted in the sampling.
Footnote 11 tells you that the authors faced this problem with respect to racial groups. They *purposely* interviewed more Latinos and African-Americans than percentages in the population would suggest. They wanted to make sure they weren't using too few people in each group to represent the group. But then, when they wanted to make a complete picture of the US, there were too many interviewees from minority groups compared to the number of Caucasian people they interviewed for the whole sample to represent the proportions of each group that exist in the U.S. So they mathematically corrected for this.
Because the survey was done by a respected research facility (NORC), and the analysis was done by a respected group of authors, and the book and related articles have undergone peer review, I can tell you that we can trust these procedures. They are standard. However, since the book was written, better, or at least more complicated, statistical procedure have become more common. The method the authors used neglects some problems with variance, but this is outside the topic of this class. In short, merely over-counting the under-represented group doesn't address problems of estimating the spread of answers to the survey questions.
Subscribe to:
Post Comments (Atom)
With regard to the question of why woman are more willing to be interviewed, I have a few opinions. I think they both go back to traditional roles of women in America.
ReplyDelete1. I believe that, even though there has been a huge shift since the start of our country in the number of woman in professional fields, there are still many, many women who spend more time at home than their husbands, or men in general. As Chad said in class one day, yes there have been changes in trends, but a still prevalent trend involves traditional roles of women and men in the home. Perhaps the South gives a different perspective to these roles, but I have lived all over (Los Angeles, Colorado Springs, Columbia (SC)) and the traditional roles were still evident in all of those very different areas.
So, I believe that, considering a common survey technique is phone calls to people's homes, more women are reachable in this fashion and have the time on their hands available for this type of activity, whereas a smaller number of men are home or willing to spend the time they are at home on the phone with a stranger.
2. I also believe that the shorter history of women being allowed to participate fully in politics may also be a factor, albeit a much smaller factor in my mind. There may be some sentiment that, because women were not always even considered worthwhile voices in the political arena, that it would be wrong not to exercise that right when specifically offered to them (by, say, a telephone surveyor only requiring a few minutes of your time for a politically oriented survey). Possibly, some women feel more of a civic responsibility to participate because of past discriminations (the same methodology to describe why minority groups may be, volume-wise, more inclined to spend their efforts in politics).
Just my initial thoughts when posed with that question.
I agree with Kaitlin.
ReplyDeleteAlso, maybe women like to talk more.