SUMMARY OF FINDINGS
The analyses in this chapter reveal a number of important findings about the application, recruitment, interview, and hiring process.
 Shirley Tilghman, 2004, “Ensuring the Future Participation of Women in Science, Mathematics, and Engineering,” in National Research Council, The Markey Scholars Conference: Proceedings, Washington, DC: National Academies Press, pp. 7-12.
 Because we performed a large number of t-tests on our faculty survey data, we will only report as significant those results with p < .05 in order to protect ourselves from false positives. Results near p < .05 will be reported as approaching significance. For the regression analyses on our survey data, reported in the final outcomes section of this chapter, we will report any results with p < .05 as significant. The reader will want to note that there are some instances in which the differences are statistically significant, but the absolute differences are quite small.
 We also performed a large number of t-tests on the NSOPF:04 data, so we followed the rule for reporting significance in these data that is described in the previous footnote.
 Data was created using the Department of Education’s Data Analysis System (DAS) available online at http://www. nces. ed. gov/dasol/. Gender was used as the row variable. The column variable was average total hours per week worked. Filters were only Research I institutions; full-time employed; with faculty status; assistant, associate, or full professors; with instructional duties for credit; and with principal fields of teaching as engineering, biological sciences, physical sciences, mathematics, and computer sciences.
 See previous footnote on how the DAS analysis was conducted.
 The committee acknowledges that the p-values for all the data presented for its faculty and departmental surveys are unadjusted and the fact that many of the data presented are interconnected.
 Kramer, C. Y., 1956, Extension of multiple range tests to group means with unequal numbers of replication, Biometrics, 12, 307-310.
 The comparisons between men and women overall, and by discipline, in terms of the number of thesis committees a faculty member served on are not reliable, due both to small sample sizes and to the longtailed distribution of this response; a few large values in response can strongly affect the comparison.
 Note that the definition the NSOPF uses is different from the definition used in the faculty survey.
 The 2002 Cornell Higher Education Research Institute (CHERI) Survey on Start-up Costs and Laboratory Allocation Rules: Summary of the Findings is available at http://www. ilr. cornell. edu/cheri/ surveys/2002surveyResults. html, accessed October 7, 2008. See also the presentation by Ronald G. Ehrenberg, Michael J. Rizzo, and George H. Jakubson, “Who Bears the Growing Cost of Science at Universities?” presented at the 2003 Conference. See also Ronald G. Ehrenberg, Michael J. Rizzo, and Scott S. Condie, “Start-up Costs in American Research Universities,” CHERI working paper, WP-33, March 2003, Cornell University.
 Sara Rimer, “For Women in Science, Slow Progress in Academia,” New York Times, April 15, 2005.
 See, for example, a thorough assessment conducted by New Mexico State University in 2003, “Space Allocation Survey,” available at http://www. advance. nmsu. edu/Documents/PDF/ann-rpt-03.pdf.
 Available at http://news-service. stanford. edu/news/2003/may21/womenfaculty-521.html.
 Note that the University of Pennsylvania’s research used an unusual metric of research space per grant dollar.
 University of Pennsylvania Gender Equity Committee, “The Gender Equality Report, Executive Summary, Almanac, Vol. 48, No. 14, December 4, 2001, available at http://www. upenn. edu/almanac/ v48/n14/GenderEquity. html. See the full report at: http://www. upenn. edu/almanac/v48pdf/011204/ GenderEquity. pdf.
 CWRU Equity Study Committee, “Resource Equity at Case Western Reserve University: Results of Faculty Focus Groups,” March 3, 2003, pp. 46-47. Available at http://www. case. edu/president/ aaction/resourcequity2003 .doc.
 Purdue conducted a survey in 2001, which asked female and male faculty whether they were satisfied with the amount of lab space. Women were less satisfied. (This is different from how much lab
space each gender has.) Available at http://www. cyto. purdue. edu/facsurvey/faculty/survey/http://www. cyto. purdue. edu/facsurvey/faculty/survey/results/intro. htm.
 The medians for men and women faculty in civil engineering were quite similar, while the means were significantly different.
 Mathematics was dropped from this analysis, as only 11 respondents in mathematics reported having lab space.
 Specifically, the observation for any respondent reporting any non-zero number, in practice from 0.5 to 19 postdocs, was changed to 1.
 See for example, Center for Research on Learning and Teaching (CRLT), The University of Michigan, “Resources on Faculty Mentoring.” Available at http://www. crlt. umich. edu/publinks/ facment. html.
 Other possible measures include original discoveries and patents. On gender differences in patenting, see Ding et al. (2006).
 In addition, the following changes to the data were made: There were about a dozen observations in which respondents reported numbers of less than $1,000. It was assumed that these numbers, such as $60 or $100 actually meant $60,000 or $100,000. It was also assumed that a single entry of $1.3 was in fact $1.3 million.
 Inspection of the data revealed that the survey results were highly influenced by a single senior female faculty in civil engineering who reported having no grant funding, and she was removed from
the survey results.
 Out of the 1,179 respondents, 4 responses were considered to be outliers and were removed.
 Note that one reason to get an outside offer is to put pressure on a faculty member’s current department to match a better offer. The faculty member might not actually want to leave his or her current department.
 The report, focusing on gender, is available at http://www. gseacademic. harvard. edu/~coache/ downloads7SNS_report_gender. pdf.
 Gender Equity Committee on Academic Climate, 2003. An Assessment of the Academic Climate for Faculty at UCLA, Los Angeles, CA: University of California at Los Angeles.
 In future studies, these two events should be separated, because male faculty tend to be older and are more likely to retire, while female faculty tend to be younger and are less likely to leave due to retirement.
 However, planning to leave or receiving outside offers are less than ideal proxies for job satisfaction. For example, faculty may plan to leave a position to retire.
 Some faculty remain associate professors and never come up for full professor status.
 For a discussion of issues and strategies related to bringing women into executive positions in academia, see NRC (2006).
 The committee acknowledges that the p-values for all the data presented for the study’s surveys of faculty and departments are unadjusted and that many of the data presented are interconnected.
 It may be that the only time the decision-making process becomes publicly visible is during litigation brought by faculty denied tenure or promotion.
 This general finding is commonly stated, even though individual institutions might have tenure or promotion rates that are comparable for men and women. As Nancy Hopkins (2006:18) notes in the case of the Massachusetts Institute of Technology (MIT), “Overall the tenure rates for men and women are almost identical in both the Schools of Science and Engineering.” Looking at a broader segment of academia is thus necessary to see if MIT, to continue the example, is representative of many institutions or is an outlier.
 This is done by subtracting the year an individual received a Ph. D. from the survey year.
See for instance Persell (1983) and McElrath (1992).
 In nine cases involving men who were up for tenure, the outcome was unknown.
 Anonymous, March 1, 1999, Women and Tenure at the Institute, MIT News Office, available at http://web. mit. edu/newsoffice/1999/trwomen. html. See also Hopkins (2006).
 See Appendix 5-2 and 5-3 for detailed tables.
 Alvarez, R. M. and J. Brehm, 1995, American ambivalence towards abortion policy: development of a heteroskedatic probit model of competing values, American Journal of Political Science, 39, 1055-1089.
 In the NSOPF data, there are many more men in the sample than women and the standard errors for women are much larger.
 Cox, D. R. and Oaks, D., 1984, Analysis of Survival Data, London: Chapman & Hall.
 Note that URLs may have changed between the preparation and release of this report.
 The sample was sent to the contractor. Once it was confirmed to have reached the contractor, the original file was deleted. Neither the committee nor the National Academies would know the names of potential respondents to the faculty survey.
 Fortunately, almost all e-mails were correct. “Bounce backs,” or non-working e-mails, were corrected. It is possible, though, that the wrong e-mail was collected and used, but that the contractor was not aware that this was an incorrect e-mail, and the respondent was never contacted.
C3. How were you made aware of your institution’s policy on tenure?
I was given a written policy.
The chair or administration told me about the policy.
 learned about the policy from other faculty Other,…
 The results of analyses are not strictly comparable, as the earlier report used a different definition of S&E, among other differences.
 From the Survey of Earned Doctorates (SED) field list, this is equivalent to any field coded from 005 to 599.
 The one recent exception appears to be the medical or health sciences, where the proportion of women among Ph. D.s seemed to have leveled off.
 Recall that Long’s definition of S&E includes the social and behavioral sciences and is thus broader than the definition employed here.
 The committee’s charge did not include a focus on exploring the reasons for gender differences in labor force outcomes outside of academia. Readers should refer to Long (2001) and Xie and Shauman (2003) for a discussion of such factors.
 These data are for just the natural sciences and engineering.
 Other includes industry, government, and the nonprofit sector. Education in this table includes K-12 positions.
 See Alexander C. McCormick, “The 2000 Carnegie Classification: Background and Description (excerpt),” available at http://www. carnegiefoundation. org/dynamic/downloads/file_1_341.pdf [accessed on November 4, 2008]. The Carnegie Foundation updated its classification system in 2005 and is available at http://www. carnegiefoundation. org/classifications/.
 Reports for 80 of the 88 Research I institutions were collected and posted to the National Academies’ Committee on Women in Science and Engineering (CWSEM) homepage, located at http:// www7.nationalacademies. org/cwse/1gender_faculty_links. html.
 This is part of the reason why most of the statistical analyses carried out use regression. A few scholars have used event history or hazard models. See for example Weiss and Lillard (1982), Kahn (1993), and Ginther (2001). See Allison (1984) for an introductory description of the methodology.
 Conducted on odd numbered years until 2003, thereafter on even numbered years, beginning in 2006.
 The National Center for Education Statistics also conducted a survey of department chairs during the 1988 NSOPF, but the chairs survey was only done this one time.
 “Survey Methodology: Survey of Doctorate Recipients,” NSF Web site at http://www. nsf. gov/sbe/ srs/ssdr/sdrmeth. htm [accessed on March 17, 2004].
 Interestingly, research is adding care of older family members—for similar reasons as care of children (e. g., Sax et al., 2002).
 A review by the Women in Science & Engineering Leadership Institute (WISELI) at the University of Wisconsin-Madison titled, “Reviewing Applicants: Research on Bias and Assumptions” identified several studies suggesting that female candidates may have a tougher time. Available at http://wiseli. engr. wisc. edu/doc/BiasBrochure_2ndEd. pdf [accessed on October 7, 2008].
 Data for 1979 are from NRC (2001a) and were calculated by taking total number of male and female faculty at Research I institutions and subtracting male and female faculty at Research I institutions who were in social and behavior sciences. Data for 2003 are also from the Survey of Doctorate Recipients (SDR) as calculated by staff, using the same definition of S&E.
 This measure is deficient in two ways. First, the potential applicant pool includes postdocs, individuals with Ph. D.s from foreign institutions, individuals from outside academia, and individuals with current academic positions who are interested in switching to a new position (Ehrenberg, 1992). For example, in a study of physics hires in 2000, Kirby et al. (2001) found that 34 percent of new hires in doctorate-granting institutions had earned Ph. D.s outside of the United States. Likewise, in computer science (Zweben, 2005:10), for 2003-2004: “Thus, more than 75% of the faculty hires made this past year by Ph. D.-granting CS/CE [computer science/computer engineering] departments appear to have been new Ph. D.s, with the rest consisting of a combination of faculty who changed academic positions, persons joining academia from government and industry, new Ph. D.s from outside of North America and from disciplines outside of CS/CE, and non-PhD. holders (e. g., taking a teaching faculty appointment).” Second, it fails to account for the preferences of doctorates.
 Ellipses omitted.
 See also Bain and Cummings (2000).
 This is not a new problem. Stake et al. (1981) found letters of recommendation were more favorable when the letter writers and the job seekers were of the same gender.
 Fractional courses were rounded up to the nearest integral number of courses. Missing data was removed from the data prior to analysis. Finally, the data were from the committee’s survey of faculty.