Multivariate Modeling of Time in Assistant Professor Rank

A Cox proportional hazards model[97] was fit to the measure of time in rank as assistant professor. A nearly identical model was fit to the data on time elapsed

TABLE 5-10 Mean Number of Months Between Receipt of Ph. D. and Promotion to Associate Professor Discipline Current Associate Professors Current Full Professors Men Women Men Women Biology 158(8)(13) 135 (5) (27) 102 (6) (13) 122 (48) (20) Chemistry 112 (4)(20) 127 (9) (23) 95 (3)(31) 88 (5) (17) Civil engineering 113 (6)(17) 80 (4)(9) 54 (4) (4) 68 (5)(4) Electrical engineering 110 (14)(4) 77 (2) (11) 78(10)(11) 85 (4)(10) Mathematics 66 (9) (16) 101 (10)(25) 80 (3) (32) 88 (4) (19) Physics 134 (5) (20) 113 (6) (20) 105 (3) (34) 104 (6) (14)

NOTES: The first set of parentheses indicates standard error of the mean, and the second set of paren­theses denotes number of observations used in the calculation.

There were only 418 faculty with current rank as associate or full who were hired at their current institu­tion as tenure-track assistant professors, who work full time, and who have a Ph. D. Only those faculty who were promoted to associate with tenure from assistant were used in the calculations. We omitted departments who did not provide information on gender of faculty, as well as one outlier who reported being unemployed for 27 years following graduation, three individuals with negative time to promo­tion (promotion happened before hire), and one person who spent 321 months as assistant professor. The numbers of cases used in Table 5-9 and here differ because we did not have reliable information about the time of graduation for some faculty. There were also several outliers. For example, a female math associate professor reported that 307 months (over 25 years) elapsed between obtaining her Ph. D. and her promotion to associate professor with tenure. A male math associate professor reported only 5 months elapsed between obtaining his Ph. D. and his promotion to associate professor with tenure. SOURCE: Faculty survey conducted by the Committee on Gender Differences in Careers of Science, Engineering, and Mathematics Faculty.

between receipt of Ph. D. and promotion to associate professor, except the second model could not include a measure of academic age, defined as time elapsed between Ph. D. and hire as tenure-track assistant professor. Other variables in both models included gender, discipline, current rank, an indicator for whether family leave was taken, type of institution (public or private), prestige, percentage of women among faculty in the department, and various two-way interactions with gender.

Results suggest that there is a complex interplay among the various factors in the model and time in rank as assistant professor. Only two of the factors—type of institution and the percentage of women among departmental faculty—appeared to have no significant association with time in rank as assistant professor. All other factors, including the interactions between gender and current rank, gender and academic age, and gender and prestige of the institution were significantly associated with time in rank as assistant professor.

Because of the presence of significant interactions, it is difficult to provide an interpretation of the effect of the main factors. In Table 5-12, we present some of

TABLE 5-11 Mean Number of Months Spent as an Associate Professor Discipline Men Women Mean Std. error n Mean Std. error n Biology 75 4 25 74 6 28 Chemistry 62 4 42 80 7 21 Civil engineering 72 7 16 75 9 18 Electrical engineering 61 8 17 70 4 20 Mathematics 59 4 35 68 5 22 Physics 77 5 41 79 7 25

NOTE: Table entries are computed using faculty who are currently full professors and who were pro­moted from associate to full and from assistant to associate at their current institution. We used only full-time faculty with a Ph. D. for whom we had complete information about the time of both promo­tions, resulting in a total of 311 observations. The proportion of nonrespondents was similar among men and women for all ranks.

SOURCE: Faculty survey conducted by the Committee on Gender Differences in Careers of Science, Engineering, and Mathematics Faculty.

the results obtained when fitting the Cox proportional hazards regression model and summarize findings via figures that show the probability of promotion to associate professor at each time point for men and women in different disciplines, who are of different current ranks and at universities of different prestige. Overall, it appears that women take significantly longer to achieve promotion to associ­ate professor with tenure, but this gender effect is confounded with current rank, discipline, and various other factors.

TABLE 5-12 Results Obtained from a Cox Proportional Hazards Regression Analysis of 351 Cases That Had Complete Promotion and Covariate Information Effect p-value Hazard Ratio Gender (1 = M, 2 = F) 0.007 0.360 Current rank (1 = F, 2 = A) <0.0001 0.166 Academic age <0.0001 1.046 Academic age x gender <0.0001 0.981 Rank x gender 0.001 2.282 Prestige 0.007 0.483 Prestige x gender 0.001 1.744 Family leave (0 = no) <0.0001 0.192 Biology vs. civil engineering 0.307 0.709 Chemistry vs. civil engineering 0.046 1.778 Mathematics vs. civil engineering <0.0001 4.126 Electrical engineering vs. civil engineering <0.0001 4.927 Physics vs. civil engineering 0.154 1.511

A hazard ratio below 1 indicates individuals in the category with the higher value of the explanatory variable “survive” longer. In this case, a faculty member “survives” in the rank of assistant professor if, in the next month, he or she does not get promoted to associate professor with tenure. For example, in the absence of interactions between gender and other variables, we would have concluded the average “risk” a female faculty will be promoted to associate professor with tenure at a given time point (given that she had not been promoted up until that time) is about 36 percent of that of a male. However, the presence of significant interactions prevents us from drawing conclusions about the effects of gender, rank, prestige, academic age, and others individually.

Figures 5-3 (a-d) show the survival curves for men and women who are cur­rently associate or full professors in biology at high-prestige institutions (Figure 5-3a) or at medium-prestige institutions (Figure 5-3b). Figures 5-3c and 5-3d show the corresponding survival curves for faculty in electrical engineering. Male full professors are represented by a gray solid curve and male associate professors are represented by a gray dotted curve. Female full professors are represented by a black solid curve and female associate professors are represented by a black dotted curve.

The plots shown in Figures 5-3 (a-d) reflect some of the complexities in the relationship between time in rank as assistant professor, current rank, gender, and prestige of the institution. For example, consider first biology. We note that at high-prestige institutions and at any time point t + 1, a male who was currently a full professor had a higher chance of getting promoted and tenured than a male

Percent FIGURE 5-3(a) Survival curves in biology at highest prestige institutions.

Percent FIGURE 5-3(b) Survival curves in biology at medium-prestige institutions.

Percent Time (month) FIGURE 5-3(c) Survival curves in electrical engineering at highest prestige institutions.

who was currently an associate professor, and in turn, they both had a higher chance of promotion at month t + 1 (given that they had not been promoted until then) than a woman who was currently a full professor or a woman who was cur­rently an associate professor. These differences vanish, however, if we consider institutions of medium prestige. In that case, while a man who was currently a full

Percent FIGURE 5-3(d) Survival curves in electrical engineering at medium-prestige institutions.

professor still had a higher chance of getting promoted at any time, given he has not been promoted earlier, there were no differences between faculty who were currently associate professors or women who were currently full professors. In the case of electrical engineering, we observed a similar pattern even though the probability of promotion increased to one at a faster rate.

Academic age (time between receipt of Ph. D. and hire as tenure-track assis­tant professor) was negatively and significantly associated with time as assistant professor: The longer the time elapsed between Ph. D. and hire, the shorter the time spent in rank as assistant. This finding is consistent with the greater publica­tion record faculty who have spent time as postdocs can present at the time of a tenure review. Academic age may contribute to the gender differential seen in the simple means of time in rank by gender, since the effect of academic age is significantly stronger for men than for women (p < 0.0001). This greater impact may reflect that men may be more likely than women to spend time after receiving their doctorate and prior to taking their first academic job pursuing professional activities, such as postdoctoral research.

Another important factor affecting time in rank as assistant professor is the increasingly available option to take family leave and stop the tenure clock. Our results show a very significant effect of stopping the tenure clock (p-value < 0.0001; see Table 5-12). The “risk” of promotion of a faculty member who stopped the tenure clock is only about 80 percent of the “risk” of promotion of a faculty member who did not, given that neither had been already promoted at a given time. Consider, as an example, two faculty members with similarly impressive academic credentials so that their “risk” of promotion becomes one if enough time has elapsed since hiring. If one of them takes a 1-year leave at the beginning of his or her probationary period, then he or she will lag behind the person who did not take the leave with respect to promotion status, but the difference in the “risk” of promotion will get smaller and smaller as the overall probabilities of promotion for both of them become larger. The effect of this factor was similar for both men and women. However, our data confirm that women were more likely to take family leave. Table 5-13 shows that 10.2 percent of female and 6.4 percent of male associate professors stopped the tenure clock. Also, stopping the clock is becoming more common over time. Virtually no faculty who are currently full professors stopped the clock, but among assistant professors, 19.7 percent of women and 7.4 percent of men have already stopped the clock. These percentages are likely to continue growing in the future.

One question our survey does not permit addressing is whether a faculty member who stopped the tenure clock has a decreased probability of promotion. To answer that question we would need a longitudinal study where faculty can be followed from the time they were hired until the time they were promoted. Our survey, which collected a snapshot cross-sectional set of data, is not appropriate for this type of question.

TABLE 5-13 Number of Faculty by Gender and Rank Who Reported Stopping or Not Stopping the Tenure Clock or Who Did Not Respond to the Survey Question Gender and Rank Did Not Stopped Clock Stop Clock Nonrespondent Total Male full professor 1 52 261(83) 314 Female full Professor 2 46 18 (80) 237 Male associate professor 14 137 68 (31) 219 Female associate professor 29 184 71 (25) 284 Male assistant professor 17 211 2 (0.8) 230 Female assistant professor 56 226 2 (0.7) 284 NOTES: Numbers in parentheses are percentage of nonrespondents in each group. Only full-time faculty with a Ph. D. and with the rank of assistant, associate, or full professor were used in the calculation. There were 1,568 such faculty. Note that many of these individuals are missing information on other variables, and thus this table includes many more persons than most of the other tables in Chapter 5. SOURCE: Faculty survey conducted by the Committee on Gender Differences in Careers of Science, Engineering, and Mathematics Faculty.

Men who are full professors today spent the least time in rank as assistant professors. This is true across all disciplines, prestige of institution, and other fac­tors. Whether males who are currently associate professors have spent more or less time in rank as assistant than women who are currently full professors depends on the institution and discipline. It is probably fair to state that women who are currently associate professors have spent the longest time in rank as assistant professors in most cases.

Faculty in biology, physics, and civil engineering are similar in terms of time in rank as assistant professor. In chemistry, math, and electrical engineering, the time to promotion to associate professor was similar and significantly shorter. The difference between disciplines was similar for both genders. There were no significant differences between private and public institutions once all other effects were accounted for.

Results for the measure of time elapsed between award of Ph. D. and promo­tion to associate professor with tenure were different and easier to interpret from the results discussed above. Using this measure, the time in rank as assistant did not differ between men and women (although it took women slightly longer to be promoted to associate from the time of graduation with a Ph. D.), and it did not differ across institutions of different prestige. Time elapsed between Ph. D. and promotion to associate was highest for faculty who were currently associate professors (as before) and for faculty in biology relative to the other disciplines.