departments are on average inviting more women to interview than would be expected if gender were not a factor, or women who apply to tenure-track or tenured positions in research-intensive institutions are, on average, well qualified. It is important to note that these higher rates of success do not imply favoritism, but may be explained by the possibility that only the strongest female candidates applied for Research I positions. This self-selection by female candidates would be consistent with the lower rates of application by women to these positions.

For tenured positions, the expected percentage of interview pools with no women are 18, 24, 35, 62, 44, and 35 percent for biology, chemistry, civil engineering, electrical engineering, mathematics, and physics, respectively. The situation for tenured positions is much less clear. Electrical engineering, mathematics, and physics have smaller all-male interview pools than their probability pools. This is particularly true for electrical engineering, which had male-only interview pools 42 percent of the time compared to a probability of 62 percent. However, civil engineering, chemistry, and biology had larger all-male interview pools than expected, with chemistry being the most notable. Fifty percent of the interview pools for tenured positions in chemistry were all-male, while the probability value was 24 percent. This finding highlights the importance of disaggregating survey data by discipline.

As with the analysis of applications, the analysis of interviews focused on departmental and institutional variables. Most of the factors in the applicant model are also used here: discipline; departmental climate, as measured by female faculty; female faculty on the search committee and family-friendly policies; public versus private universities; and prestige. Much of the literature on making hiring more equitable focuses on bringing actors with a broader view from outside the department into the decision making, so we expect intervention by a dean might also be positively related to the probability of interviewing a woman.

Because departments draw from the pool of applicants in deciding whom to interview, this analysis controls for the percentage of applications from women—the dependent variable from the last model. We expect a positive relationship between the percentage of applications from women and the percentage of interviewees who are women.

The percentage of women in the interview pool appears to exceed the percentage of female applicants in all areas. We now investigate whether the percentage of women in the interview pool is associated with the institutional, departmental, and position-level characteristics described earlier and with two additional predictors: the percentage of female applicants and an indicator of whether the composition