A Jeanne Site
California State University, Dominguez Hills
University of Wisconsin, Parkside
Latest update: May 6, 1999
Faculty on the Site.
You will need to find the whypoor variable in the DBH criminal justice book, if that is the number. The page numbers are different, but the data is the same.
It makes sense to recode the whypoor variable because the four values coded for the answer to the question on why respondents believe that people are poor are actually broken into two categories: one structural, one personal. If we say that people are poor because there are poor schools, that really says that there is a social structural problem in poverty. If we say that people are poor because they are lazy and won't work, that really says that poverty is a problem that resides in the individual.
Thus, we could look at Joan Kemp's theory on women working, and divide the explanations respondents give into structural and personal sources of the problem. We could also divide the solutions into personal and structural, except that this question wasn't asked in the GSS.SAV survey. That's one of the problems with secondary analysis of data. Sometimes we can't answer the new questions we'd like to ask because the old data didn't include them.
Note that this goes beyond recoding the values of one variable. This is putting two variables together to recode them as one new one. That is more complex. And our previous examples with CHATT do not cover it. Instead, we would need to create an index for whypoor, pp. 107 to 116 of DBH's criminal justice book. They create an index on ABORT. Practice with their example.
You are not required to recode whypoor for one of the basic exercises. But I do want you to e-mail me a brief (25 words or so) explanation for why you might want to do that in a re-analysis. Be sure to include that you recognize that sometimes combining variables will permit you to answer questions that you cannot answer from the original analysis, and that this process requires recoding. You cannot simply add percentages from the two variables you would like to combine to answer your question.
I would suggest recoding into povcause - one value for structural causes, one value for personal causes. Use Transform -> Recode -> New Variable to create POVCAUS
Simplest way to answer this question would be to recode whypoor into povcaus. Then, run the variables frequencies.
But you could just reinterpret the data in the frequencies of the whypoor tables for each value of the variable. You could discuss the percent of respondents who believe that poverty is caused by poor schools and the percentage who believe it is caused by a lack of jobs. In reinterpreting, you cannot just add the percentages together! You would need to recode to do that.
You could also reinterpret by adding together the two categories of Very Important and Somewhat Important. That's what the cumulative percent does.
Or you could decide to interpret just the Very Important category as compared to the Not Important category. You would essentially be throwing out some data that way, but then you would be sure that the two categories were distincly different. People who choose Very Important usually have very different opinions from those who chose Not Important, where we can't be too sure what Somewhat Important actually means. (This is called interpretation.)
Since one of the whypoor variables asks that specific question, you can just interpret the SPSS frequencies on that variable. Even if you don't want to include the Somewhat Important category, you don't need to recode. You could just read the Very Important category percentage from the table. If you want to talk about structural vs. personal views of poverty, then you would want to recode, but you would want to create a new variable, using the four whypoor variables as the values of the new variable.
Jeanne your answers.