Weighted Analyses Addendum

As discussed in the study report, Assessing the Academic Work Environment for Women Scientists and Engineers (September 26, 2002), we had only five possible indicators with which to evaluate the representativeness of the sample: track (tenure, research, clinical), college, rank, race-ethnicity and gender. The three faculty tracks were equivalently represented in the respondent sample and the pool of faculty included in the survey. Within the tenure and research tracks, there were no significant differences by race, rank or school between the respondent sample and the pool of faculty surveyed. This suggests that for tenure and research track faculty our survey sample is representative of the larger pool of faculty in terms of type of appointment held, college of appointment, rank and ethnicity. To assess the possibility that the sample of male respondents was less representative of all male scientists and engineers surveyed we compared male and female respondents to the overall sample pools of men and women separately. We found that for both groups, respondents on the tenure and research tracks did not differ from the pool as a whole; the male and female respondents on these two tracks appear to be equally represented.

However, across tracks, faculty of color responded at a lower rate (26%) than European American faculty (40%), as is often the case with social science surveys. Among clinical faculty, faculty of color and assistant professors responded at a lower rate than white faculty and those at higher ranks. There was also a gender difference on all tracks between survey respondents and the pool of faculty surveyed. Women of both academic groups (scientists/engineers and social scientists) responded at a higher rate than men: 50% female scientists and engineers, 47% female social scientists vs. 26 % male scientists and engineers.

The evidence we suggests that the respondent sample is representative of the larger pool of faculty surveyed. However, given the lower response rates for faculty of color male faculty, all analyses presented in the report were replicated using appropriate weights. Weighted data analyses adjust the raw survey data to represent the population from which the sample is drawn. In this case the data were weighted on the basis of race and gender demographic characteristics of the UM faculty population surveyed, as well as the response rates by race and gender. The weighted analyses continued to include the same controls previously used to correct for differences among the three core groups compared in the instructional track analyses. While the original analyses were done using SPSS, the weighted analyses were accomplished using the statistical package Stata, because SPSS significant tests for weighted analyses are based on incorrect error terms.

Summary

Overall, results from the weighted analyses confirm the findings presented in the survey report. All major findings (e.g., those using created scales and that were relatively robust and fit with a pattern of other findings in this survey and/or in other studies) included in the report were replicated. In some particulars (e.g., individual items) the weighted analyses findings differ slightly from the unweighted analyses. In some instances original findings were not replicated (e.g. results from the weighted analyses indicate that men scientists/engineers do not have more undergraduate students than female scientists/engineers and women scientists/engineers are not more likely than men scientists/engineers to be offered travel funding and course release in contract renegotiations). In other instances, new significant results were found that further supported key points in the report (e.g., weighted analyses indicate that men scientists/engineers are more likely than women scientists/engineers to be offered a partner/spouse position during contract renegotiation and women scientists/engineers are more likely to be single parents than men scientists/engineers). While these differences are important to note, it should be emphasized that the overall patterns of relationships found in the original analyses are sustained in the weighted analyses.

For your information, summary data tables of the weighted analyses, mirroring the tables presented in the original report, are included here.