Same Work, Different Pay? Evidence from a US Public University

Abstract

This study examines detailed data for faculty at a typical public research university in the United States between 1995 and 2004 to explore whether gender wage differentials can be explained by productivity differences. The level of detail – including the number of courses taught, enrollment, grant dollars, and number and impact of publications – largely eliminates the problem of unmeasured productivity, and the restriction to one firm eliminates unmeasured work conditions that confound investigations of wider labor markets. The authors find that direct productivity measures reduce the gender wage penalty to about 3 percent, only 1 percentage point lower than estimates from national studies of many institutions and with fewer productivity controls. The wage structure for women faculty differs markedly from the wage structure for men. Interpreted against the institutional features of wage setting for this population, the paper concludes that penalties for women arise at the department level.

Keywords:

• Academic labor markets
• earnings differentials
• gender wage gap
• racial inequality
• wage determination

ACKNOWLEDGMENTS

We would like to thank Mark Chisholm, Carol Bernhard, Susan Carkeek, Denise Wallen, and Ann Powell for providing administrative data; the College of Arts and Sciences department chairs at UNM for providing or encouraging their faculties to supply current curriculum vitae; Arts and Sciences faculty for providing curriculum vitae; Justin Smith and Michael Milligan for research assistance; and then-Interim Dean Vera Norwood and then-Provost Reed Dasenbrock for initiating and supporting this study. Three anonymous reviewers provided helpful comments. Any remaining errors are our own. All personal information that would allow the identification of any person or person(s) described in the article has been removed.

Notes

¹ See, for example, George Psacharopoulos and Zafiris Tzannatos (1992) on gender gaps and George Psacharopoulos and Harry Anthony Patrinos (1994) on indigenous–non-indigenous gaps in Latin America.

² Quoted from the comments of an anonymous reviewer.

³ About 9 percent of instructional faculty and staff work at private research universities, 8 and 3 percent work at public and private doctoral universities (which offer fewer doctoral degrees), 16 and 7 percent work at public and private comprehensive colleges (where the highest degree is a masters), and 20 percent work at public two-year institutions.

⁴ Information retrieved from the Integrated Postsecondary Education Data System (IPEDS-2004) using the US Department of Education, National Center for Education Statistics Data Access System (DAS), http://nces.ed.gov/dasol/tables/ (accessed July 2008).

⁵ Percent of student body that is Hispanic is from IPEDS-2004 (2004); percent of faculty that is Hispanic is from the National Survey of Postsecondary Faculty (NSOPF-04; 2004). Both data sets were accessed through DAS.

⁶ In quotes because it is not the actual cost of living increase as measured by the Consumer Price Index.

⁷ We are grateful to an anonymous reviewer for raising this point.

⁸ The US National Research Council's website includes links to seventy-five recent institutional reports (see www7.nationalacademies.org/cwse/gender_faculty_links.html [accessed January 2008]).

⁹ The College of Arts and Sciences employed only one tenured or tenure-track African American faculty member in 2004 and five American Indians, seventeen Asians, and four who declined to state their ethnicity.

¹⁰ An exception is Zvi Griliches's (1986) study of firm-level data, noted in Deirdre N. McCloskey and Stephen T. Ziliak for “understand[ing] that populations should not be treated as samples” (1996: 106).

¹¹ For privacy, we eliminated the department name after we merged the publications and administrative records. After this modification, we are able to identify faculty members as working in the same department, and we can identify whether the department is in the humanities, social sciences, or natural sciences, but we are unable to name the department.

¹² In a few cases, years at UNM exceed years since degree. Where this occurs, we use years at UNM to measure experience.

¹³ We exclude years in which faculty members worked as college administrators.

¹⁴ Each principal and co-principal investigator is assigned the total grant amount.

¹⁵ Some mentoring activities are counted as course preps: students can register for independent studies supervised by a faculty member under one course number and for thesis credits under two distinct course numbers if they are working on an MA or PhD thesis.

¹⁶ Graduate student counts include all students enrolled in course numbers 500 and higher. Undergraduate student counts include all students enrolled in course numbers lower than 500. Because our teaching data are based on classroom counts, we may incorrectly identify graduate and undergraduate students in the small portion of cases where they share a classroom in cross-listed courses. Similarly, we may undercount preps when courses are co-taught or cross-listed, since there is only one professor of record for each set of classes taught in the same classroom and at the same time.

¹⁷ We followed the recommendation of department chairs contacted during the summer of 2005 to determine the best method for compiling publication records. Most chairs recommended using curriculum vitae, which they either solicited directly from faculty on our behalf or directed their administrators to provide. For several departments, chairs recommended a web search of publications. We used Sci-Search for chemistry, physics, and math and statistics and Econlit for economics.

¹⁸ For faculty with fewer than five years of experience, we use five times their average annual publication measures for the last five years measure.

¹⁹ See Bernt Bratsberg, James F. Ragan, and John T. Warren (2003) for a discussion of this issue.

²⁰ Relatively small numbers of Hispanics make it impractical to estimate wage structures by gender.

²¹ Thirteen faculty members did not hold a PhD in at least one year in the data, and eleven did not have a PhD in 2004 (the last year of data). Of these eleven, six are women, seven are in the humanities, and four are in the social sciences. Six had been at UNM for three or four years, and five had been at UNM for more than ten years. It is likely that the long-term faculty members without a PhD have distinguished themselves in other ways, for example, in creative writing. This would account for the variability in the association between PhD and earnings.

²² In specifications that omit article impact, the coefficient on both log articles variables is 0.001, which translates to 5 percent higher earnings for a faculty member with 50 percent more articles, about half the return to impact.

²³ In specifications that included interactions of age and productivity measures, older faculty appear to enjoy higher rewards for current and earlier books, current articles and earlier impact.Average returns for each research productivity category mask differences by broad discipline categories: humanities departments most reward books and impact; natural sciences departments most reward articles; and social sciences most reward impact but penalize articles.

²⁴ Given the university-mandated promotion raises, attributing the trend coefficient to university rules is not unreasonable.

²⁵ Ward (2001), however, finds that the wage structure for women in Scottish universities provides higher returns to high-wage departments than does the wage structure for men.