Gender Equity Report

I currently serve as the Executive Director of the Office of Policy Analysis for the University System of New Hampshire. In addition to my office's reporting obligations to the Board of Trustees, we conduct research on a wide range of topics relating economics to higher education, such as legislative and student demand for higher education, cost functions for colleges and universities, and graduate program ratings. I regularly publish findings from these studies in peer-reviewed academic journals, and have presented my work at conferences including the Allied Social Science Association and the Southern Economics Association. A list of my academic publications and professional qualifications is provided in the Appendix.

During the past ten years, I have developed a particular expertise in the area of faculty/staff compensation and work-related issues. I assumed the position in January of 1990 of Coordinator/Research Associate with the Management Information Division at the University of Minnesota. My primary responsibility was to conduct research on faculty/staff compensation issues. This work resulted in the development of alternative methods for adjusting faculty salaries and measuring faculty salary inequities. Subsequently, I have conducted statistical studies of faculty gender equity using national data licensed through the National Center for Education Statistics. More recently I have completed a gender equity study for non-faculty staff for the University of New Hampshire, and regularly teach professional development workshops on conducting gender equity studies for the Association for Institutional Research. My work on faculty compensation and work-related issues has been published in Research in Higher Education, The Review of Higher Education, Economics of Education Review, The Journal of Higher Education, and Quarterly Review of Economics and Finance.

I received my Ph.D. in economics from Indiana University, where I concentrated in econometrics. I have also taught statistical analysis, econometrics, and microeconomics at the undergraduate and graduate levels annually on an adjunct basis at Indiana University, the University of Minnesota, and the University of New Hampshire. During the past four years, I have been engaged as a statistical/testifying expert for the University of Minnesota. My billing rate is $130/hour for work performed as a statistical/testifying expert.

The analysis and opinions contained in this report are based upon my education, training, and experience in the areas of economics and econometrics, upon my knowledge of accepted practices in the field for analyzing questions of gender equity and faculty pay, and upon my review of the existing literature in the field.

_______________________________________

Robert K. Toutkoushian, Ph.D.

Executive Summary

The goal of this study was to examine the compensation and rank assignments for female faculty at the University of Cincinnati and determine if there is evidence that they are being treated differently than men in terms of their pay and rank. There is clear statistical evidence that female faculty at the University of Cincinnati currently earn significantly less than comparable male faculty in the 1999-00 academic year. After controlling for the effects of years of experience, years of seniority, educational attainment, academic department/college, academic rank, and race on earnings, I found that women in regular faculty positions currently earn on average 2.83% less than men using the single-equation method or 4.85% less than men using the Oaxaca multiple-equation method. These findings are not materially affected by whether or not academic rank is included in the salary model. Given the wide variability that I found in the salary inequities for women, the Oaxaca method is the approach that should be used for measuring and removing gender-based salary inequities at the University of Cincinnati.

I show that the total cost of removing the unexplained wage gap between men and women would be approximately $1.5 million in the Oaxaca method. This option, however, would leave many women at the University of Cincinnati with large gender-based salary inequities. The total cost of removing all current individual-specific salary inequities for those women who are underpaid would be $2.2 million. Finally, the cost to the University of removing all current and past gender-based salary inequities is conservatively estimated at $7.3 million, and can range as high as $23 million at the other extreme. Based on these findings and my professional judgment, I recommend that the University of Cincinnati implement a salary adjustment plan for female faculty that will properly compensate them for both their current and past gender-based salary inequities experienced while employed by the University of Cincinnati, and address individual-specific differences in salary inequities.

At the same time, the data do not provide clear evidence that female faculty who are currently employed by the University of Cincinnati are less likely than comparable males to occupy the ranks of Full or Associate Professor, after controlling for the effects of the aforementioned variables on rank. Since these data do not include information on those female faculty members who have left the University, it is not known whether or not they faced inequitable treatment during the promotion process. Likewise, while women in non-regular academic positions earn significantly less than comparable women in regular positions, I found no statistical evidence that women in non-regular academic positions are underpaid by more than their male counterparts in similar positions. Finally, neither the salary nor rank models revealed any evidence that non-white faculty were treated differently than comparable white faculty.

An Analysis of Gender Inequities in Salary and Rank for Faculty

at the University of Cincinnati, Academic Year 1999-00

Introduction

Following the passage of the Civil Rights Act in 1964, and subsequent application to the academic labor market through Executive Orders 11246 and 11375, numerous studies have been initiated at colleges and universities across the country to determine if female faculty faced gender discrimination with respect to pay and rank. Ransom and Megdal (1993) summarize the findings from a number of gender equity studies, and a list of selected institutional and national studies that have been published in peer-reviewed academic journals are provided in Table 1.

What these and other studies have clearly documented over the years is that on average, female faculty earn significantly less than their male counterparts. These earnings differences persist even after taking into account differences between male and female faculty in their levels of factors such as experience, educational attainment, and field/discipline that arguably should influence compensation. Barbezat (1991), for example, found that after controlling for a standard set of variables, women faculty in 1989 earned approximately seven percent less than comparable men. Furthermore, studies using national data have shown that the current estimates of gender-based pay disparities are comparable to the levels found in the mid 1970s. More recently, I applied Barbezat's salary models to faculty in the most recent national survey conducted by the National Center for Education Statistics (NCES), and obtained estimates of the unexplained wage gap between male and female faculty that were very similar to those reported by Barbezat for earlier years (Toutkoushian, 1998).

In addition to being paid less than comparable males, female faculty often face another form of discrimination in the academic labor market in that they are often promoted at lower rates than males. Given the importance of academic promotion to a faculty member's career, both in terms of salary increases and the awarding of lifetime employment (tenure), discrimination in promotional opportunities for women is an equally serious concern as pay discrimination. A handful of analysts have investigated claims of gender discrimination in rank, and have generally found evidence that women have lower probabilities of receiving tenure and/or being promoted than male colleagues possessing similar characteristics (see Riggs et al., 1986; Weiler, 1990; Ransom and Megdal, 1993; Toutkoushian, 1999).

In this study, I investigate whether there is evidence of gender-based inequities in rank and pay for female faculty at the University of Cincinnati. To accomplish this, I apply multivariate statistical techniques to data obtained from the University to investigate these issues. The approaches and model specifications that I use here follow the accepted practices of leading analysts in the field. I begin with a description of the statistical procedures used to investigate gender discrimination in rank and salary, and a review of the literature to show the types of controls that are most frequently used in these statistical procedures. The second section contains a description of the data and subsequent variables that I use in the study.

In the next section, I provide the results from the statistical analyses and use this information to answer the questions of whether or not there is evidence of gender discrimination in rank and salary for female faculty at the University of Cincinnati. In short, I find that while there is no evidence of overall gender discrimination in rank attainment for female faculty in tenured and tenure-track positions, there is convincing evidence that female faculty are currently paid between three and five percent less than male faculty possessing similar measured characteristics, depending on the method used to measure the unexplained wage gap. I also investigate whether female faculty who are in non-regular academic positions are paid significantly less than male faculty members in similar academic positions at the University of Cincinnati. In the last section I provide recommendations as to how these gender-based salary inequities can be removed, and the subsequent cost of doing so. This section will highlight the importance of making salary adjustments to compensate women for both their past and current salary deficiencies due to unequal treatment by gender.

Background Information

Measuring Gender-Based Salary Inequities

Most inquiries into possible salary discrimination are initiated by observing a difference in average salaries between male and female faculty. The media often report statistics such as "On average, a woman earns eighty cents for every dollar earned by a man." Figures 1 through 3 provide information for each rank on the amounts by which average salaries for male faculty exceed the average salaries for female faculty at the University of Cincinnati from 1978-79 through 1998-99. Throughout this period at the Full Professor level, male faculty have earned between ten and sixteen percent more than female faculty, with the average difference being close to twelve percent. Smaller, but persistent, average salary differences were also found between male and female faculty at the Associate and Assistant professor ranks.

Analysts recognized early on, however, that such comparisons of average salaries between men and women had limited value for determining whether comparable men and women were paid differently because they did not take into account possible gender differences in important labor market factors that should legitimately affect salaries. For example, male faculty typically have more years of experience in academe, on average, than their female counterparts. Given that employees with more experience are usually expected to earn more than employees with less experience, some of the observed difference in average salaries between males and females may be properly attributed to their different experience levels, rather than discrimination.

The standard procedure by which legitimate salary differences such as these can be measured and removed from the overall average salary difference is through multiple regression analysis. The primary advantage of multiple regression analysis is that it allows the analyst to control for the influences of other factors such as labor market experience, educational attainment, and field that may legitimately affect faculty compensation. The analysis begins with the specification of a salary equation of the form:

where lnY_i = salary for the i-th faculty member expressed in logarithms, X₁ to X_J = set of J independent variables that are deemed appropriate for differentiating salaries among faculty, ₀ to _J = set of coefficients to be estimated by multiple regression analysis, and _i = random error term. This semilogarithmic functional form was first popularized by Jacob Mincer (1974), and has become the standard model used in salary studies due to its appropriateness in situations where salaries are generated by a compounding process over time. The resulting salary model shows how each of the designated factors X₁ to X_J affects a faculty member's salary.

The salary model in equation (1) requires the analyst to select control variables X₁ to X_J that are both measurable and deemed relevant for varying salaries among all faculty. This issue is important because the final salary model specification can affect the estimated level of pay disparity between male and female faculty. Most variables are chosen on the basis of their having a connection to human capital theory, which asserts that a worker's level of compensation will be influenced by those attributes that contribute to his or her work productivity. Potential human capital measures for faculty include educational attainment and academic experience. While faculty research and teaching output are also interpretable as measures of an individual's human capital, they are not widely used in institution-specific salary models because they can be difficult to define and measure.⁽¹⁾ Aside from human capital theory, in studies where all female faculty are treated as a single group, it is common to add variables that control for meaningful differences across academic positions in the type of work being performed and the external markets in which they can find employment. These factors might include the length of appointment and an individual's academic field or discipline.

While measures of experience, education, and field enjoy near unanimous support among analysts, there is often disagreement as to the appropriateness of including other potential variables as controls in the salary model. It is argued that if women face discrimination in pay, then they may also face discrimination with regard to access to certain positions that also influence pay. The most common target of such disputes is with a faculty member's academic rank. To control for rank in a salary model involves an implicit assumption that equally-qualified male and female faculty will, on average, hold the same rank. If in fact female faculty have less access than males to higher academic ranks, then controlling for rank will mask some of the effects of gender discrimination because women are paid less in part because they are being held at lower ranks where the pay is also lower. Similar concerns have been raised with the use of variables controlling for whether a faculty member has held an administrative position within the institution.

The most objective approach for determining which variables should be included in a salary model is to review the types of variables used in faculty salary studies that have been conducted by other researchers and published in peer-reviewed academic journals. These studies can provide insight into the model specifications deemed most appropriate by professionals since they have passed through peer review, and help highlight important issues of contention in the field.

In Table 1, I provide a summary of the different salary model specifications used in twenty-four published studies of gender equity for faculty conducted since 1973. The studies are grouped according to whether they rely on data for one institution (17) or national samples of faculty (7) since differences in the model specifications arise due to the means by which data are collected (surveys versus personnel data) and the necessity in national surveys to also control for institution type and related factors. Within each group, data limitations also impact the range of regression model specifications that can be used. Most institution-specific studies rely on data found in institutional personnel databases. These transactional database systems usually do not contain information on faculty research and teaching productivity and detailed previous work history, and this constrains the degree to which analysts can control for these factors in a salary model, despite their possible importance.

The results show that virtually every institution-specific study listed here controlled for total experience, academic field/discipline, and educational attainment. At the other extreme, only two out of twenty-four studies shown in Table 1 controlled for time within current rank. Other factors, including research productivity measures, seniority, and administrative experience fall in between these extremes. National studies are more likely than institution-specific studies to include measures of research and teaching productivity in their salary models since this information can be collected through the same survey instrument used to collect other data on faculty, whereas an institution-specific study would require a separate survey of faculty or a mandate from the administration to provide publication and teaching histories to the institution.

After selecting the appropriate list of control variables for inclusion in the salary model, the results can be used to measure the unexplained wage gap between male and female faculty. Either a single-equation method or a multiple-equation method can be used for this purpose, with each having its advantages and disadvantages. In the single-equation method, a dummy variable for each faculty member's gender is added to the wage equation:

where G_i = 1 if female, 0 otherwise. The gender coefficient _J+1 represents the unexplained wage gap (UWG = _J+1), or the remaining portion of the difference in the average log of salaries between male and female faculty after taking into account differences in the levels of the independent variables in X for males and females. To facilitate interpretation, it is helpful to convert the unexplained difference in the log of salaries to the unexplained percentage difference in male/female salaries. Kennedy (1981) has shown that this can be done in the single-equation method by using the formula:

where 'exp' = exponential function, and Var(_J+1) = estimated variance of the gender coefficient. To illustrate, if the gender coefficient equals .05 and the variance equals .0001, then the .05 difference in log of salaries translates into a percent salary difference of exp(.05 - .00005) - 1 = 5.12%.

The single-equation approach has been criticized by a number of researchers, however, for imposing the restriction that each variable in X must have the same effect on wages for both male and female faculty. For example, if one of the independent variables in the salary model is years of experience, then this restriction means that an additional year in the labor force must have the same average impact on salary for both men and women. If these restrictions are inappropriate, then the estimated coefficient for the gender variable will lead to an incorrect measure of the unexplained wage gap. As an alternative, analysts such as Oaxaca (1973), Blinder (1973), Reimers (1983), Cotton (1988), Neumark (1988), and Oaxaca and Ransom (1994) have proposed using multiple-equation methods for measuring the unexplained wage gap. Rather than adding a gender dummy variable to a single wage equation, the multiple-equation methods involve estimating the wage equation (1) separately for males and females:

(4a) & (4b)

Equation (4a) is often referred to as the male wage equation, and equation (4b) the female wage equation, since each is estimated separately for the respective genders. In these two equations, each variable in the salary model is permitted to have a unique effect on salary for males and females (i.e., _fj does not have to equal _mj).

The most commonly-used alternative to the single-equation method is based on the work of Oaxaca (1973). Howard, Snyder, and McLaughlin (1992) refer to a variation of this procedure as the "best-White-male-model" because only the characteristics of white males are used to develop the baseline regression equation. Here, I use the characteristics of all males as the baseline since there is no preliminary evidence of unexplained earnings differences by race, and will simply refer to it as the Oaxaca method. Under the Oaxaca method, the values for the independent variables for each female faculty member are substituted into the male wage equation, and their predicted salaries are obtained:

where ln(woman as man)_i = predicted log of salary for the i-th woman if she were paid as a man. The difference between each woman's predicted salary as if male and her actual salary, both expressed in logarithms, is her individual gender-based salary inequity:

This quantity represents the amount by which each woman is underpaid, expressed in logarithms. To translate this into a percentage salary inequity for each individual, the following formula can be used:

where Y_i = actual salary. Unlike the single-equation method, the percent salary inequities in the Oaxaca method can vary across women. To derive a measure of the unexplained wage gap, sum the differences in equation (6) across all women and divide by the number of women:

Two methods have been presented here for measuring gender-based salary inequities and the unexplained wage gap. The literature on gender equity offers no clear guide as to which method is the most appropriate to use. It has become common practice for analysts to use both a single- and multiple-equation method when conducting studies of faculty compensation (see Barbezat, 1989; 1991; Ransom and Megdal, 1993; Toutkoushian, 1998). The choice between single- and multiple-equation methods depends on how much variability exists in the levels of pay disparity across women. By forcing each variable to have the same effect on the salaries of male and female faculty, the single-equation method restricts all women to have the same percent salary inequity. If in fact the levels of pay disparity for some women are significantly greater than the overall average, then a multiple-equation method would be needed to measure these differences and prescribe remedies for them.

Measuring Gender-Based Rank Inequities

The manner in which academic rank is treated in gender equity studies requires a more detailed discussion. Dating back to 1974, professionals in the field have expressed concern over whether independent variables for current rank should be used as controls in the faculty salary equation. The argument against using rank in the salary model is that there is the possibility that rank assignment is gender biased, and hence controlling for current rank in the salary model would lead to an understatement of the level of pay disparity between men and women. ⁽²⁾

In their review of the literature, Ransom and Megdal (1993) highlight all of the studies that controlled for academic rank and stress that the "Inclusion of rank as an explanatory variable will likely understate the 'gender gap'" (Ransom and Megdal, 1993, p.27). For this reason, the official "salary kit" produced by the American Association of University Professors (AAUP) has recommended against using rank as an independent variable in faculty salary models (Scott, 1977). At the same time, advocates for using rank argue that it is desirable to control for rank in salary models since academic promotions typically involve increases in pay, and that academic rank serves as a good proxy for unobserved research and teaching quality. Rank has also been found to be a good predictor of faculty salaries (Raymond et al., 1988; Boudreau et al., 1997).

While the issue of whether to control for rank has yet to be resolved, researchers have adopted two general approaches to the problem. The first approach is to use logistic regression analysis or a comparable technique to conduct a statistical test to determine if men and women with similar characteristics have equal likelihoods of attaining higher ranks. As techniques such as logistic regression have become easier to use with advances in computer technology, the frequency with which they have been applied to various problems has increased considerably during the 1990s. This begins with the specification of a logistic regression model to explain a faculty member's rank, such as:

where P_i = probability that the i-th faculty member holds a specific rank, Z₁ to Z_K = set of K independent variables thought to influence rank assignments, ₀ to _K+1 = coefficients to be estimated, and v_i = random error term. Since rank and salary are viewed as human-capital based rewards for faculty, most of the same factors thought to influence salary are also typically used in the logistic regression model to study rank assignments. The results from this test can then be used to decide whether it is proper to include rank as a variable in the salary model (see Ervin, Thomas and Zey-Ferrell, 1984; Riggs et al., 1986; Weiler, 1990; Ransom and Megdal, 1993; Boudreau et al., 1997). If the coefficient _K+1 is statistically less than zero (for G=1 if female, 0 otherwise), then this would be evidence that female faculty have a lower likelihood than comparable male faculty of attaining a designated rank.

A number of studies that have examined faculty rank have concluded that females are less likely than their male peers to attain higher ranks within academia. Weiler (1990) found that for a national sample of faculty in 1968, after controlling for experience, race, educational attainment, type of institution, field, and publications, female faculty were less likely than comparable male faculty to hold higher ranks. In addition, Weiler showed that the gender disparity in rank accounted for fifteen to twenty percent of the overall gender-based pay disparity in his sample. Ransom and Megdal (1993), using national samples of faculty from 1969, 1973, 1977 and 1984, found that after controlling for educational attainment, experience, seniority, and publications, female faculty were significantly less likely than comparable male faculty to hold the rank of Full/Associate Professor. I also found that a similar gender difference in rank attainment persists for a sample of national faculty in 1993 at the Full and Full/Associate Professor levels after taking into account race, career publications, primary teaching field, educational attainment, type of institution, experience, and seniority (Toutkoushian, 1999).

A second approach is advocated by some researchers. Rather than use information on gender bias in rank to determine if it should be used in the salary model, these researchers present results from two different salary models, the first controlling for academic rank and the second without controlling for academic rank, and allow the reader to observe the sensitivity of the unexplained wage gap to the inclusion of rank. These types of studies do not argue in favor of one of the two salary model specifications since they do not offer evidence as to which model is more appropriate. But the studies make clear to the reader that this is an issue of contention among analysts and show how it affects the findings from the study. Analysts who have used this approach include Ferber (1974), Hoffman (1976), Raymond, Sesnowitz and Williams (1988), Barbezat (1987; 1989; 1991), and McNabb and Wass (1997).

I have provided specific details of how researchers have dealt with rank in Table 2. Note that only three of the twenty-four studies reviewed here, and only one after 1976, controlled for rank without either testing for possible gender bias in rank or showing the results from a similar salary model without rank. The most popular option has been to present findings from two salary models, with and without controlling for rank, although in the 1990s it has become common to conduct statistical tests for potential gender bias. It is very rare for a study to include rank as a variable in the salary model without taking either of these measures, and such a study would probably have a difficult time passing through peer review if submitted for publication in an academic journal due to this omission.

Analysis for the University of Cincinnati (Regular Positions)

Data Description

The data that I used in this study were obtained from the University of Cincinnati and consist of all faculty at the University for the academic year 1999-2000 who were on the payroll as of October 1999. The dataset consists of 1,967 individuals spanning all colleges at the University, including the main campus as well as the Clermont College and Raymond Walters College. To help ensure that the set of faculty used in the analyses is relatively homogeneous and that the analysis conducted here is comparable to accepted practices and standards in the field, I excluded several groups of faculty from the analysis. First, only faculty members who are employed in tenured or tenure-eligible positions are included in the primary salary models. Accordingly, individuals in "non-regular positions" (those with job titles beginning with either Adjunct, Research, Clinical, or Field Service) were omitted from the analysis in this section. Second, I omitted most of the faculty in the Medical School since the salaries that they receive from the University substantially underrepresent their total income due to the omission of income derived from clinical practices. The only exceptions that I made were for faculty members in basic science departments in the Medical School where individuals do not receive clinical income (Cell Biology, Molecular Genetics and Biochemistry, Molecular & Cellular Physiology, Pharmacology/Cell Biology). After excluding other individuals with missing data on one or more of the variables used in the following analyses, the final sample that I used consisted of 1,328 faculty.

Variable Description

From the personnel database for the University of Cincinnati, I created the following variables for each faculty member: logarithm of annual salary, three dummy variables for highest degree (doctorate, professional, other/not available), years employed at the University of Cincinnati, years since highest degree, four dummy variables for academic rank (full, associate, assistant, instructor), two dummy variables for race (white, nonwhite), and 113 dummy variables for departmental/college affiliation. For faculty with missing data on their highest degree, I used their age minus twenty-five as an approximation for years since highest degree. I squared the variables for years employed at the University of Cincinnati and years since highest degree to help capture possible non-linear relationships with salary. For faculty on less than 100% time appointments, I converted their salaries to a full-time equivalent (FTE) basis. Likewise, I increased the salaries of faculty who are currently on paid or unpaid leave from the University of Cincinnati by two percent to reflect the across-the-board salary increase that they would have received had they been on active status.

Table 3 provides descriptive statistics for these variables, broken down separately for male and female faculty members. It can be seen that on average male faculty are more likely than female faculty at the University of Cincinnati to have earned a doctorate degree (seventy-two percent for men versus fifty-three percent for women), and average about three more years of labor market experience both at the University and since attaining their highest degree. With regard to rank, men are more likely than women to hold the rank of Full Professor (forty-seven percent for men versus twenty-seven percent for women), whereas women are more prevalent than men at the Assistant Professor level (eighteen percent for men versus thirty percent for women).

Turning to compensation, male faculty earn on average $13,317 more per year than their female counterparts. This translates into a twenty-four percent difference in average salaries, or a 0.21 difference in the average logarithms of annual salary. To determine if this average salary difference is statistically significant, I applied a two-sample t-test to the data (Table 4). I first used Levene's test for the equality of variances to determine if the variances in male and female faculty are the same. Since this hypothesis was soundly rejected by the data, I used the version of the t-test that allows for unequal population variances. The results from the t-test show that the difference in average salaries is statistically significant at common levels of significance; therefore, male faculty clearly earn more than female faculty on average.

Empirical Results

As noted in the Introduction, the appearance of a large difference in average salaries could be attributable to differences across genders in other factors such as experience and education that should rightly influence a faculty member's compensation. A quick glance at Table 3, for example, shows that male faculty have more labor market experience than their female counterparts at the University of Cincinnati, and are more likely to have a doctorate degree as their terminal degree. It is expected, therefore, that these and similar differences in departmental/college affiliation and other factors may explain some portion of the average salary difference between male and female faculty.

Table 5 summarizes the results from two single-equation salary models for faculty at the University of Cincinnati. The log of annual (FTE-adjusted) salary is used as the dependent variable in each model. The first model controls for the following variables: female ("male" is the reference category), non-white ("white" is the reference category), highest degree doctorate, highest degree professional ("other or non-reported highest degree" is the reference category), Full Professor, Associate Professor, Instructor ("Assistant Professor" is the reference category), years since highest degree and squared years since highest degree, years at the University of Cincinnati and squared years at the University of Cincinnati, and 112 dummy variables for departmental/college affiliation ("Mathematics" is the reference category). The second model controls for all of the same factors, except that the three variables representing faculty rank were omitted from the model. These two model specifications are used so that the effects of controlling for rank on the unexplained wage gap can be readily observed. Each of the salary models controls for over one hundred and twenty variables. The complete results for each model are contained in the Appendix.

From the value of R-squared in the second model in Table 5, it can be seen that the variables for gender, race, highest degree, years since highest degree, years at the University of Cincinnati, and department/college affiliation collectively explain seventy-six percent of the variations in faculty salaries. Adding controls for academic rank to the salary model increases the explanatory power of the model to eighty-four percent. These values of R-squared compare very favorably with those obtained by other researchers who have modeled the salary determination process for faculty at other institutions. For example, the final salary model that was used as the basis for the settlement of a class-action suit at the University of Minnesota in 1986-87 explained less than sixty percent of the total variations in faculty salaries (Goodman, Hoenack, and Rasmussen, 1989; Toutkoushian, 1994). The key coefficients of interest for the present study, however, are for the variable "female" shown in the first row, since this represents the unexplained wage gap in the single-equation method. The fact that the estimated coefficient is negative and statistically significant at the 1% significance level when controlling for rank (and the 5% level without controlling for rank) is clear evidence that female faculty at the University of Cincinnati currently earn less than male faculty after taking into account their educational attainment, race, education, rank, experience, seniority, and department/college. Applying Kennedy's (1981) formula to the results of the regression model shows that according to the single-equation method, female faculty earn 2.83% less than comparable male faculty.

Interestingly, the estimated level of pay disparity for female faculty at the University of Cincinnati is not sensitive to whether or not controls for rank are included in the model, and thus the issue of whether rank should be included in the salary model is not very important in this application. I conducted a more formal test of this proposition in Table 6, where two logistic regression models are used to explain whether a faculty member holds the rank of either Full Professor or Associate Professor. The same set of independent variables used to explain a faculty member's salary are used to explain their current rank: highest degree, race, gender, years since highest degree, years at the University of Cincinnati, and departmental/college affiliation. In each model, gender is not a statistically significant determinant of rank after taking into account these other factors. Therefore, there is no evidence that female faculty who are currently employed by the University of Cincinnati occupy ranks that are lower than those held by male faculty with similar measurable characteristics. As a result, controlling for rank in the salary model will not create a downward bias in the estimated level of pay disparity for women, and therefore the set of independent variables used in the first salary model specification shown in Table 5 are the most appropriate for examining gender-based salary inequities for faculty.

As noted in the previous section, there is no universally-accepted best method for measuring the unexplained wage gap between male and female faculty. Accordingly, the next step in the analysis was to measure the unexplained wage gap using the Oaxaca method described earlier. I begin in Table 7 by presenting the results from applying the salary model separately to male and female faculty. In general, while I found the effects of many independent variables in the model on salary to be similar for both male and female faculty, some important exceptions are worth noting. First, the salary difference between Full and Assistant Professors is greater for males (+0.354) than for females (+0.321). Second, while male Instructors earn approximately ten percent less than male Assistant Professors, female instructors earn over twelve percent less than female Assistant Professors. Third, while the salaries for males increase with years since highest degree, the same does not hold for female faculty. Finally, the intercept of the all-male salary model (10.604) is approximately three percent higher than the intercept for the all-female equation (10.567).

Table 8 compares the results from applying the single-equation method and the Oaxaca method to faculty at the University of Cincinnati. The salary models used in each method control for the same independent variables. The total wage gap is decomposed into two quantities: the portion that is explained by differences in the independent variables for men and women, and the remainder, which is the unexplained wage gap. The values shown in square brackets below each quantity show the percentages of the total wage gap that are attributed to each quantity. In the Oaxaca method, the values for the independent variables for all women are substituted into the all-male wage equation, and the unexplained wage gap is computed as the average difference between each woman's actual and predicted logarithm of salary (equation 8). The second column repeats the findings from the single-equation method shown in Table 5 (column 1). The last column in Table 8 reports the calculated t-statistics for the unexplained wage gaps to determine if they are statistically significant.

Note that under each method, there is a significant unexplained wage gap between male and female faculty at the University of Cincinnati after taking into account the independent variables in the salary model. I found that the unexplained wage gap is equal to +0.028 using the single-equation method and +0.038 using the Oaxaca method. In percentage terms, these translate into a 2.83% average salary inequity in the single-equation method and a 4.85% average salary inequity using the Oaxaca method. Therefore, the conclusion that female faculty at the University of Cincinnati are currently paid less than comparable male faculty holds for both methods, and the unexplained wage gap is two percent larger in the Oaxaca method. As I will show later in this report, the Oaxaca method uncovers a significant amount of variation in the individual salary inequities faced by women. In this instance, it is preferable to use the Oaxaca method rather than the single-equation method to measure and correct salary inequities.

Analysis for the University of Cincinnati (Non-Regular Positions)

The preceding analysis only focused on faculty members who are employed in tenured or tenure-eligible positions. In addition to these individuals, there are one hundred and twenty-four faculty members who would meet the criteria for inclusion in the analysis, except that they are not employed in tenured or tenure-eligible positions. There would include faculty members with position titles beginning with "Adjunct," "Clinical," "Research," or "Field Service." To investigate whether female faculty in non-regular positions are paid less than male faculty in non-regular positions, the ideal procedure would be to use the same approach used for regular faculty members; that is, specify and estimate a separate salary model for all faculty in non-regular positions, and then measure the resulting unexplained wage gap using the single-equation and Oaxaca methods. However, this is not feasible in the present study due to the small number of individuals who are employed in these non-regular positions and the large number of variables in the salary model.

As an alternative, I use the following procedure to approximate the unexplained wage gap between male and female faculty in non-regular academic positions: First, I used the salary model estimated for male faculty in regular positions (Table 7, column 1) as the baseline salary model. Second, I substituted the characteristics of each of the one hundred and twenty-four faculty members employed in non-regular positions into this equation, and obtained their predicted log of salaries. Third, I calculated the average differences between the actual and predicted log of salaries separately for males and females. Finally, I compared these average differences to each other to determine if the average difference for females was greater than the average difference for males. Note that the objective of this exercise is not to determine if faculty in non-regular positions are paid differently than faculty in regular positions per se, but rather to determine if women in non-regular positions are being treated differently than comparable men with regard to compensation.

The results are provided in Table 9. It is shown that on average, faculty in non-regular positions earn over twenty percent less, on an FTE-adjusted basis, than their peers who are employed in regular academic positions. While the regular/non-regular pay differential for female faculty is approximately twenty-one percent, it is even larger -- approximately twenty-three percent -- for male faculty. Based on the results of the t-test, there is no clear evidence that female faculty in non-regular positions are paid less, on average, than comparable males. It should be noted that this analysis does not consider an equally challenging question; namely, are women more likely than equally qualified men to be employed in non-regular positions?

Removing Gender-Based Pay Disparities

Given the clear evidence of a significant unexplained wage gap between male and female faculty in regular positions at the University of Cincinnati, a procedure is needed for correcting these salary inequities. The literature on wage discrimination offers several alternative methods for removing gender-based salary inequities, and three such methods are described here.⁽³⁾ The first method is to give an across-the-board salary increase to all female faculty at the University of Cincinnati. While this method is relatively easy to implement and treats all women in the same manner, by definition it does not completely remove all of the current salary inequities for those women with larger than average levels of pay disparity. Alternatively, a second method could be used where each underpaid woman receives a salary adjustment equal to her current individual-specific level of pay disparity as calculated by the Oaxaca method.

Both of these methods focus on correcting gender bias in current salaries; however, it is crucial to note that if women at the University of Cincinnati have been underpaid during each of their years of employment, then only making adjustments to correct current salary inequities would still leave women with substantial gender-based pay inequities from their previous years of employment. To address this, a third alternative method for correcting salary inequities is described where each woman's salary adjustment takes into account both her length of employment at the University and her levels of gender-based pay disparity during her time of employment.

Method #1: Across-the-Board (Current Year)

With an across-the-board salary adjustment, all women in the sample would receive the same percentage salary increase. This is accomplished by increasing the log of salary for each woman by an amount equal to the unexplained wage gap. From Table 8, the unexplained wage gap was shown to be either 0.028 using the single-equation method or 0.038 using the Oaxaca method. Since theory offers no clear choice as to which of these methods is most appropriate to the problem at hand, I show cost calculations using both of these estimates. In each instance, the main objective of the Method #1 approach is to eliminate the unexplained wage gap.

Each woman's dollar salary adjustment is equal to the difference between her salary after adjustment and her current salary. The cost of implementing an across-the-board salary increase for all women is obtained by summing these salary adjustments across all women. Since fringe benefits would also accompany such salary increases (currently 30.5% of base salary for faculty at the University of Cincinnati), the total cost must also take into account fringe benefits. Using the single-equation method, the final cost formula can be written as follows:

and likewise using the Oaxaca method:

An across-the-board salary increase for all female faculty has some desirable properties: it is easy to implement, it treats all women faculty as a single, uniform class, and it would eliminate the unexplained wage gap between male and female faculty (depending on how it is measured).

Method #2: Individualized (Current Year)

While an across-the-board increase to all women would be successful at removing the unexplained wage gap between male and female faculty at the University of Cincinnati, in most situations such a plan would not fully remove all of the individual gender-based pay disparities at the institution. To see why this is true, think of the unexplained wage gap as being the average level of gender-based pay disparities for women. Almost by definition, some women will have larger-than-average pay disparities while others face smaller-than-average pay disparities. A uniform across-the-board salary increase for all women will shift the distribution of salary inequities to the right so that the new average is set equal to zero. However, women who originally had higher-than-average levels of pay disparity will still be underpaid relative to comparable men after receiving an across-the-board salary increase. In short, an across-the-board salary adjustment method is effective at correcting average salary discrimination whereas an individualized method is needed to correct all current inequities faced by women.

To measure individual-specific salary inequities, a multiple-equation method such as the Oaxaca method must be used. I substituted the characteristics for each woman into the all-male regression model shown in the first column of Table 7, and obtained her predicted log of salary if paid as a man (ln(woman as man)_i ). The difference between her predicted salary as if male and her actual salary represents the dollar amount by which each woman is underpaid relative to comparable men:

When the dollar salary inequity is positive, a woman would be predicted to earn more than she currently earns if she were paid as a man. If a woman's percentage salary inequity exceeds the average percentage salary inequity for all women, then she would still be left with a salary inequity if she only received an across-the-board salary adjustment equal to the unexplained wage gap.

Table 10 shows the distribution of salary inequities for female faculty at the University of Cincinnati, in both dollars and percentages, according to equation (11). While thirty percent of the women have actual salaries that exceeded their predicted salaries if paid as men, seventy percent of the women are deemed to be underpaid. At the higher end of the distribution, note that forty percent of the female faculty had salary inequities in excess of 6.4 percent, and about thirty percent of female faculty had pay disparities of ten percent or more. Clearly, there is a significant amount of variation in salary inequities faced by female faculty at the University of Cincinnati, and thus many female faculty would still be considerably underpaid if they received an across-the-board salary increase equal to the unexplained wage gap. In applications such as this, an individualized salary adjustment plan is an appealing alternative to the across-the-board method for correcting current salary inequities for faculty, and the Oaxaca method should be used to measure salary inequities.

Since it would be politically difficult to reduce the salaries of faculty (male or female) as part of any salary adjustment plan, only those women with actual salaries that fall below their predicted salaries if paid as if male would receive salary adjustments (272 women out of 439 in the full sample). I summed the dollar salary adjustments across all female faculty members with positive salary differentials to determine the total cost of removing all current salary inequities. As with an across-the-board salary adjustment plan, fringe benefits would also increase in response to any salary adjustment. Taken together, the cost of bringing the salaries of all underpaid female faculty in line with what they would currently be predicted to earn if paid as males would be:

Method #3: Correcting Current and Past Salary Inequities

An important limitation of the cost calculations shown in Methods #1 and #2 is that they only provide compensation to women for the salary inequity they experience in the current year. However, these methods do not compensate women for the salary inequities that they have experienced in years prior to the current year during which they were also employed at the University of Cincinnati. Women who have worked at the University of Cincinnati for more than one year and are currently underpaid have most likely been underpaid for each of their years at the institution, and are thus entitled to compensation to make up for their cumulative salary inequities. To illustrate this problem, Table 11 shows the annual and cumulative salary inequities for a hypothetical woman initially hired in 1975-76 at a salary of $20,000. Suppose that this woman should have earned three percent more -- $20,600 -- if she were paid as a man. In 1975-76, therefore, she was underpaid by the amount of $600. Now assume that over time salaries for both men and women at the University increased by five percent per year. In each year from 1975-76 through 1999-00, the woman would earn three percent less than she should earn if paid as a man. Note, however, that the cumulative amount by which she has been underpaid since being hired by the University grows dramatically. In the last year (1999-00), for example, her current salary would have to be increased by $1,935 to bring it up to the level she would receive if paid as a man. Taking into account her twenty-five years of employment, however, she was underpaid by a total of $28,636.

A woman's total salary inequity for her duration of employment at the University could thus be represented as:

where subscript t = year of employment, PCT_it = percent salary inequity for the i-th woman in year t, and Y_it = actual salary for the i-th woman in year t. There are two difficulties with applying the formula in equation (13) to faculty in this dataset. First, the dataset only includes each woman's current salary and not her previous salaries at the University of Cincinnati. Since these data are not readily available, I approximated previous salaries by subtracting from each woman's current salary an amount equal to the average salary increase given to faculty for the year in question. For example, if a woman earned $40,000 this year, and the collective bargaining agreement called for an average five percent salary increase (across-the-board plus average merit), then her predicted salary for the last year would be equal to $40,000 / 1.05 = $38,095. I performed these calculations for each year and each of the 439 female faculty members who are employed in the 1999-00 academic year.

A second difficulty with equation (13) is that estimates must also be used for the percent salary inequities faced by women in their previous years of employment at the University of Cincinnati. I assume throughout the following simulations that the levels of past salary inequity were equal to the current level of salary inequity. This assumption is justified on the basis of national studies of salary equity, as well as studies that have been conducted in the past on the salaries of faculty at the University of Cincinnati. National studies of faculty compensation have consistently shown that past levels of pay disparity were at least as large as the levels found in more recent data. Toutkoushian (1998), for example, found that in a national survey of faculty in 1993, women earned approximately eight percent less than comparable women, and that this unexplained wage gap was very similar to what other researchers found using survey data from 1977 and 1984. Other researchers, including Barbezat (1991) and Ransom and Megdal (1993) reached the same conclusions using different national surveys of faculty. Ransom and Megdal further report generally higher levels of pay disparity between men and women in earlier institution-specific studies than they found in more recent studies.

Information on the history of salaries at the University of Cincinnati provides strong evidence that previous levels of salary inequity at the institution were at least as large as current estimates. At the University of Cincinnati, collective bargaining has been in place since the 1975-76 academic year. The presence of collective bargaining should reduce the extent to which the institution could widen the unexplained wage gap over time. Across-the-board percentage salary increases will preserve the unexplained wage gap for continuing faculty, and across-the-board dollar salary increases will reduce the unexplained wage gap since the fixed dollar increases for women would, on average, translate into higher percentage salary increases. It is only through giving inequitable merit increases for continuing faculty, or inequitable salary increases to accompany promotions, or inequitable starting salaries for men and women, that the unexplained wage gap could possibly widen over time.

As can be seen from the history of negotiated salary increases at the University of Cincinnati in Table 12, across-the-board salary increases (in percentages and dollars) have been much more common than merit increases. From 1975-76 to 1999-00, continuing faculty received across-the-board percentage salary increases totaling 94.55%, across-the-board dollar salary increases of $12,863, and merit salary increases of only 16.5%. Clearly, merit increases have constituted a very small proportion of all salary increases given to continuing faculty over this period. Likewise, the average salary comparisons by rank shown in Figures 1 through 3 show that throughout this period, female faculty have earned less than their male counterparts within each rank.

In addition, two separate studies conducted at the University of Cincinnati strongly support the contention that gender-based pay disparities in the past were at least as large as the current estimates. A study of faculty salaries at the University of Cincinnati conducted by Johnson in 1995-96 revealed a similar level of pay disparity to that found here for 1999-00. Model 1 in Table 4 of Johnson's (1996) report showed that female faculty at the University of Cincinnati earned $1,676 less than comparable male faculty after controlling for rank, years in rank, years in career, years at the University of Cincinnati, and department/college. This regression model is the most similar to those shown in this report. While the salary difference is expressed in dollar figures and not percentages, if the average salary for female faculty in 1995-96 was $50,000, this would translate into a 3.4% unexplained wage gap.

More recently, Bellas, Ritchey, and Parmer (2000) conducted a statistical analysis of changes in salaries for continuing faculty at the University of Cincinnati from 1985 to 1995. They found that after controlling for race, highest degree, years since highest degree, rank at the time of hire, administrative experience, external labor market demand, and college affiliation, women at the University earned $2,085 less than comparable men in 1985 (or six percent, assuming an average salary for female faculty of $35,000 in 1984-85). Of particular relevance for this study is their finding that, for continuing faculty, the annual rate of salary growth for women (6.7%) was higher than the annual salary growth for men (5.4%) over this ten-year interval. Bellas, Ritchey, and Parmer also showed that there were no statistical differences in the average merit increases awarded to men and women, and that women were more likely to have received a promotion during this period. Taken together, this provides convincing evidence that the salary inequities faced by women at the University of Cincinnati in previous years were at least as high as the current salary inequities, and were in fact probably larger than the current estimates.

Table 14 provides five different estimates of what it would cost to correct both current and past salary inequities faced by women faculty at the University of Cincinnati. These salary inequities can be removed via an across-the-board or individualized approach, or a combination of these approaches. In the first two estimates (Method 3(a) and 3(b)), all 439 women in the dataset would receive a salary increase equal to the current unexplained wage gap for each year in which they were employed by the University:

where _it = predicted salary for the i-th woman in year t. Equation (14a) uses the unexplained wage gap from the single-equation method, whereas equation (14b) relies on the unexplained wage gap from the Oaxaca method.

In the third and fourth estimates, salary adjustments are only made for those 272 women who are currently underpaid by the University. They receive their full individualized salary adjustment in the current year, and receive either a 2.83% increase (using the single-equation method) or a 4.85% increase (using the Oaxaca method) for all previous years of employment at the University of Cincinnati:

where Y_iT = current salary for the i-th woman. Note that the only differences between Methods 3(a,b) and 3(c,d) are that Methods 3(c,d) use the individual-specific salary inequity for the current year, and apply the salary adjustments to only 272 out of 439 female faculty. Finally, in the last option, each underpaid woman receives her individual-specific salary adjustment in the current and previous years:

where PCT_iT = percent salary inequity in the current year, and is applied to the current and previous salaries for each underpaid woman. This method assumes that a woman's previous levels of salary inequity are the same as her current salary inequity.

Table 13 provides the cost calculations for each method for each year. The total cost estimates, including fringe benefits, range from a low of $7.3 million in Method 3(a) to over $23 million in Method 3(e). It is important to note that these calculations likely understate the true cumulative pay disparity faced by women at the University of Cincinnati for several reasons. First, they do not include any payments to women who are no longer employed as faculty by the University of Cincinnati. Second, these estimates do not reflect the fact that salary adjustments received in the past could have lost purchasing power over time due to the effects of inflation. Third, salary inequities faced by women prior to the 1975-76 academic year are not included in the totals. Finally, if the level of pay disparity at the University has generally fallen over time, then higher salary inequities should have been applied to previous years of employment and hence the cost estimates in Methods 3(a) through 3(e) will understate the total level of pay disparity for women. These calculations are useful for illustrating the true magnitude of the salary inequity problem faced by women at the University.

Main Conclusions

The goal of this study was to examine the compensation and rank assignments for female faculty at the University of Cincinnati and determine if there is evidence that women are being treated differently than men in terms of their pay and rank. The main conclusions of the study are the following:

There is clear statistical evidence that female faculty at the University of Cincinnati earn significantly less than comparable male faculty in the 1999-00 academic year. After controlling for the effects of years of experience, years of seniority, educational attainment, academic department/college, academic rank, and race on earnings, I found that women in regular faculty positions earn significantly less than men. Depending on whether a single-equation or multiple-equation method is preferred, women presently earn either 2.83% or 4.85% less than men. The high degree of variability in salary inequities that I found when applying the Oaxaca method lends support to the use of the Oaxaca method, rather than the single-equation method, to measure salary inequities and the unexplained wage gap. The regression model specification that I used for this analysis was shown to be very comparable to the accepted practices in the field, and the finding of a gender-based pay disparity is not materially affected by whether or not academic rank is included in the salary model.

At the same time, I found no clear evidence that female faculty who are presently employed by the University of Cincinnati are less likely than comparable male faculty to occupy the ranks of Full or Associate Professor, after controlling for the effects of the aforementioned variables on rank. Since these data do not include information on those female faculty members who have left the University, it is not known whether or not individuals who have left the University have faced inequitable treatment during the promotion process. Similarly, while women in non-regular academic positions earn significantly less than comparable women in regular positions, the same is generally true for male faculty, and thus there is no statistical evidence that the level of underpayment for women in non-regular academic positions is larger than for their male counterparts. It is noteworthy, however, that the compensation of all non-regular faculty is much lower on an FTE-adjusted basis than would be predicted from the all-male regular faculty salary model, given their relative qualifications. The present study does not address the appropriateness by which faculty are assigned to either regular or non-regular academic positions. The data also did not reveal any unexplained earnings or rank differences for faculty on the basis of their race.

Finally, the cost calculations that I provide in the previous section offer several alternatives for the University of Cincinnati to use to remedy the salary inequities for female faculty. If the goal is to simply remove the current unexplained wage gap, then this could be accomplished through giving all female faculty a salary increase equal to the unexplained wage gap. When combined with the accompanying fringe benefits, this would lead to a total cost of $1.5 million using the Oaxaca method. This approach, however, would leave many women at the University of Cincinnati with large gender-based salary inequities. The total cost of removing all individual-specific salary inequities in the current year for only those women who are underpaid would be $2.2 million. Finally, neither of these approaches would properly compensate women for all of the salary inequities they have faced while employed by the University of Cincinnati. The cost of removing all current and past gender-based salary inequities is conservatively estimated at $7.3 million, and can range as high as $23 million depending on the assumed levels of past salary inequities.

Based on these findings and my professional judgment, I recommend that the University of Cincinnati implement a salary adjustment plan for female faculty that will properly compensate them for both their current and past gender-based salary inequities. Furthermore, given the wide range of salary inequities faced by women at the University, an individualized salary adjustment plan is the preferred method for adjusting salaries.

Tables

Click here to view tables.

Appendix

Click here to view the appendix.

Curriculum Vitae for Robert K. Toutkoushian

PERSONAL: Business Address: Office of Policy Analysis

University System of New Hampshire

Myers Financial Center

27 Concord Road

Durham, NH 03824

(603) 862-0966

EDUCATION: Ph.D., Economics, Indiana University, 1991

M.A., Economics, Indiana University, 1986

B.A., Economics, Indiana University of Pennsylvania, 1984

Dissertation: "Strong Form and Semi-Strong Form Market Efficiency: A Theoretical, Empirical, and Experimental Analysis." (Arlington Williams and Roy Gardner, co-chairpersons)

FIELDS OF SPECIALIZATION:

Economics of Education

Econometrics

Labor Economics

Financial Economics

CURRENT POSITION: Executive Director, Office of Policy Analysis, University System of New Hampshire

Duties include responsibility for planing, conducting, and reporting studies related to all aspects of the University System, providing leadership in identifying relevant higher education issues, participating in the development of the planning and policy related agenda for the Board of Trustees, and conceptual design and implementation of decision support or related information systems. Current projects include (1) an analysis of performance indicators for higher education institutions, (2) a study of enrollment demand patterns of New Hampshire graduates, and (3) the creation of tools for strategic planning within the system.

PUBLICATIONS IN REFEREED JOURNALS:

"Addressing Gender Equity in Non-Faculty Salaries" forthcoming, Research in Higher Education, 2000.

"Do Parental Income and Educational Attainment Affect the Initial Choices of New Hampshire's College-bound Students?" forthcoming, Economics of Education Review, 2000.

PUBLICATIONS (cont'd):

"The Status of Academic Women in the 1990s: No Longer Outsiders, but Not Yet Equals,"

Quarterly Review of Economics and Finance 39(Special Issue), 1999, p.679-698.

"Faculty Time Allocations and Research Productivity: Gender, Race, and Family Effects,"

Review of Higher Education 22(4), 1999, p.367-390 [with Marcia Bellas].

"The Value of Cost Functions for Policymaking and Institutional Research," Research in Higher

Education 40(1), 1999, p.1-16.

"Racial and Marital Status Differences in Faculty Pay," The Journal of Higher Education 69(5),

1998, p. 513-541.

"Using Panel Data to Examine Legislative Demand for Higher Education," Education Economics 6(2), 1998, p. 141-157 [with Paula Hollis].

"Using Regression Analysis to Determine if Faculty Salaries Are Overly Compressed," Research

in Higher Education 39(1), 1998, p.87-100.

"Sex Matters Less for Younger Faculty: Evidence from the 1988 and 1993 NCES Surveys,"

Economics of Education Review 17(1), 1998, p.55-71.

"The National Research Council Graduate Program Ratings: What Are They Measuring?"

Review of Higher Education 21(4), 1998, p.427-443 [with Halil Dundar and William Becker].

"Determinants of Outsider Excess Returns From Insider Transactions and Semi-Strong Form

Efficiency," Applied Financial Economics 6, 1996, p.155-162.

"Testing the Fisher Effect as a Long Run Equilibrium Relationship," Applied Financial

Economics 6, 1996, p.115-120 [with Joseph Daniels and Farrokh Nourzad].

"The Measurement and Cost of Removing Unexplained Gender Differences in Faculty Salaries,"

Economics of Education Review, 14(3), 1995, p.209-220 [with William Becker].

"Using Citation Counts for Measuring Sex Discrimination in Faculty Salaries," The Review of

Higher Education 18(1), 1994, p.61-82.

"Issues in Choosing a Strategy for Achieving Salary Equity," Research in Higher Education

35(4), 1994, p.415-428.

SELECTED PAPERS PRESENTED:

"The Measurement and Cost of Removing Unexplained Gender Differences in Faculty

Salaries" paper presented at the Association of Institutional Researchers for the Upper Midwest (AIRUM), St. Paul, MN, October 1992, and at the annual meetings of the Midwest Economics Association, Indianapolis, IN, April 1993 [with William E. Becker].

"Issues in Choosing a Strategy for Achieving Salary Equity," paper presented at the HHH

Institute of Public Affairs, University of Minnesota, Minneapolis, MN, October 1993, and the AIR Forum, New Orleans, LA, May 1994.

"Citations and Their Marginal Values," paper presented at the annual meetings of the

Midwest Economics Association, Chicago, IL, March 1994.

"Using Citations for Measuring Sex Discrimination in Faculty Salaries," paper presented at the

AIR Forum, New Orleans, LA, May 1994, and the Department of Economics, University of South Carolina, September 1994.

"Determinants of Citations in Economics," paper presented to the Department of Economics,

Marquette University, Milwaukee, WI, May 1995.

"Article Placement, and Journal and Author Reputation, as Determinants of Citations,"

paper presented at the annual meetings of the Southern Economics Association, New Orleans, LA, November 1995 [with William Becker].

"Sex Still Matters in Academe: Evidence from the 1993 NCES Survey", paper presented

at the annual meetings of the Midwest Economics Association, Chicago, IL, April 1996.

"What Are the National Research Council Graduate Program Ratings Measuring?" paper

presented at the annual meetings of the Northeast Association for Institutional Researchers, Princeton, NJ, November 1996, the annual meetings of the Southern Economics Association, Washington, DC, November 1996, and the annual meetings of the Association for Institutional Research, Orlando, FL, May 1997.

"Using Regression Analysis to Determine if Faculty Salaries Are Overly Compressed,"

Paper presented at the annual meetings of the Association for Institutional Research, Orlando, FL, May 1997.

"Racial and Marital Status Differences in Faculty Pay," Paper presented at the annual

meetings of the Southern Economics Association, Atlanta, GA, November 1997 and the Association for Institutional Research, Minneapolis, MN, May 1998.

TEACHING EXPERIENCE:

Indiana University, 1985-1989

(students: undergraduates in economics)

Introduction to Statistics for Business and Economics

Introduction to Microeconomic Theory

University of Minnesota, 1990-1996

(students: graduate students in Public Affairs)

Intermediate Microeconomic Theory

Quantitative Methods I

Principles of Microeconomics

University of New Hampshire, 1997

(students: Ph.D. and Masters students in economics)

Econometrics II

PROFESSIONAL AWARDS:

Selected Teacher of the Year, Hubert H. Humphrey Institute of Public Affairs, 1994.

OTHER PROFESSIONAL ACTIVITIES:

Refereed Articles for:

Economics of Education Review

The Journal of Higher Education

Journal of the American Statistical Association (JASA)

Research in Higher Education

Professional Memberships/Committees/Activities:

Contributing Editor, Research in Higher Education

Member, AIR Professional Development Services Committee

Consulting Activities:

Testifying/Statistical Expert, Ian Maitland v. University of Minnesota, et al., Civil File No. 4-93-25, January-May 1999.

EMPLOYMENT HISTORY:

Research Associate, Office of Planning and Analysis, University of Minnesota (January 1990 - June 1996).

Duties include the formulation and testing of policy relevant hypotheses relating to the economics of higher education. Responsibilities included developing research projects, maintaining large databases on faculty and staff, supervising staff in the completion of these studies, performing statistical analyses, and writing and presenting reports of our findings. Studies included analyses of pay equity for faculty and staff, developing enrollment forecasting models, identifying key determinants of citation behavior, examining the effects of alternative higher education financing formulas, and the determinants of student success.

Adjunct Faculty Member, Hubert H. Humphrey Institute of Public Affairs, University of Minnesota (September 1990 - June 1996).

Taught classes to Masters level Public Affairs students in statistical analysis, intermediate microeconomic theory, principles of microeconomics, and calculus.

Senior Applications/Software Programmer, University Computing Services (UCS), Indiana University (October 1988 - December 1989).

Served as a full-time statistical consultant to faculty, staff, and students. Provided assistance to users on issues such as research design, appropriate statistical software for specific applications, database construction and management, and the interpretation of statistical results. Duties also included teaching courses to faculty, staff, and students on a wide range of topics.

Materials Used in the Preparation of This Report

Published Studies Cited in Report

Ashraf, Javed (1996), "The Influence of Gender on Faculty Salaries in the United States,

1969-89," Applied Economics, volume 28, number 7, pp.857-864.

Balzer, William, Nancy Boudreau, Peter Hutchinson, Ann Marie Ryan, Todd Thorsteinson,

James Sullivan, Robert Yonker, and Deanne Snavely (1996), "Critical Modeling Principles When Testing for Gender Equity in Faculty Salaries," Research in Higher Education, volume 37, number 6, pp.633-658.

Barbezat, Debra (1987), "Salary Differentials by Sex in the Academic Labor Market," Journal

of Human Resources, volume 22, number 3, pp.422-428.

Barbezat, Debra (1989), "Affirmative Action in Higher Education: Have Two Decades Altered

Salary Differentials by Sex and Race?" Research in Labor Economics, volume 10, pp.107-156.

Barbezat, Debra (1991), "Updating Estimates of Male-Female Salary Differentials in the

Academic Labor Market," Economics Letters, volume 36, number 2, pp.191-195.

Boudreau, Nancy, James Sullivan, William Balzer, Ann Marie Ryan, Robert Yonker, Todd

Thorsteinson, and Peter Hutchinson (1997), "Should Faculty Rank be Included as a Predictor Variable in Studies of Gender Equity in University Faculty Salaries?" Research in Higher Education, volume 38, number 3, pp.297-312.

Ervin, D., B.J. Thomas, and M. Zey-Ferrell (1984), "Sex Discrimination and Rewards in a Public

Comprehensive University," Human Relations, volume 37, pp.1005-1025.

Ferber, Marianne (1974), "Professors, Performance, and Rewards," Industrial Relations, volume

13, number 1, pp.69-77.

Ferber, Marianne, Jane Loeb, and Helen Lowry (1978), "The Economic Status of Women

Faculty: A Reappraisal," Journal of Human Resources, volume 13, number 3, pp.385-401.

Ferber, Marianne, and Carole Green (1982), "Traditional or Reverse Sex Discrimination? A

Case Study of a Large Public University," Industrial and Labor Relations Review, volume 35, number 4, pp.550-564.

Ferree, Myra, and Julia McQuillan (1998), "Gender-Based Pay Gaps," Gender & Society, volume 12, number 1, pp.7-40.

Gordon, Nancy, Thomas Morton, and Ina Braden (1974), "Faculty Salaries: Is There

Discrimination by Sex, Race, and Discipline?" American Economic Review, volume 64, number 2, pp.419-427.

Hirsch, Barry, and Karen Leppel (1982), "Sex Discrimination in Faculty Salaries: Evidence from

a Historically Women's University," American Economic Review, volume 72 (September), pp.829-835.

Hoffman, Emily (1976), "Faculty Salaries: Is There Discrimination by Sex, Race, and

Discipline? Additional Evidence," American Economic Review, volume 66, number 1, pp.196-198.

Howard, Richard, Julie Snyder, and Gerald McLaughlin (1992), ""Faculty Salaries" in The Primer for Institutional Research, edited by Meredith Whiteley, John Porter, and Robert Fenske Tallahassee, FL: Association for Institutional Reseach, p.51-62.

Katz, David (1973), "Faculty Salaries, Promotions, and Productivity at a Large University,"

American Economic Review, volume 63, number 3, pp.469-477.

Koch, J., and J. Chizmar (1976), "Sex Discrimination and Affirmative Action in Faculty

Salaries," Economic Inquiry, volume 14, pp.16-24.

Johnson, George, and Frank Stafford (1974), "The Earnings and Promotion of Women Faculty,"

American Economic Review, volume 64, number 6, pp.888-903.

Kennedy, Peter (1981), "Estimation with Correctly Interpreted Dummy Variables in

Semilogarithmic Equations," American Economic Review, volume 71, number 4, pp.801.

Megdal, Sharon, and Michael Ransom (1985), "Longitudinal Changes in Salary at a Large

Public University: What Response to Equal Pay Legislation?" American Economic Review, volume 75, number 2, pp.271-274.

McNabb, Robert, and Victoria Wass (1997), "Male-Female Salary Differentials in British

Universities," Oxford Economic Papers, volume 49, number 3, pp.328-343.

Mincer, Jacob (1974). Schooling, Experience, and Earnings. New York and London: Columbia

University Press.

Neumark, David (1988), "Employers' Discriminatory Behavior and the Estimation of Wage

Discrimination," Journal of Human Resources, volume 23, pp.279-295.

Oaxaca, Ronald (1973), "Male-Female Wage Differentials in Urban Labor Markets,"

International Economic Review, volume 14, pp.693-709.

Oaxaca, Ronald, and Michael Ransom (1994), "On Discrimination and the Decomposition of

Wage Differentials," Journal of Econometrics, volume 61, pp.5-21.

Ransom, Michael (1993), "Seniority and Monopsony in the Academic Labor Market," American

Economic Review, volume 83, number 1, pp.221-233.

Ransom, Michael, and Sharon Megdal (1993), "Sex Differences in the Academic Labor Market

in the Affirmative Action Era," Economics of Education Review, volume 12, number 1, pp.21-43.

Raymond, Richard, Michael Sesnowitz, and Donald Williams (1988), "Does Sex Still Matter?

New Evidence from the 1980s," Economic Inquiry, volume 26, number 1, pp.43-58.

Riggs, M., R. Downey, P. McIntyre, and D. Hoyt (1986), "Using Discriminant Analysis to

Predict Faculty Rank," Research in Higher Education, volume 25, pp.365-376.

Scott, E. (1977). Higher Education Salary Evaluation Kit. American Association of

University Professors.

Toutkoushian, Robert (1994), "Using Citations to Measure Sex Discrimination in Faculty

Salaries," Review of Higher Education, volume 18, number 1, pp.61-82.

Toutkoushian, Robert (1998), "Sex Matters Less for Younger Faculty: Evidence of

Disaggregate Pay Disparities from the 1988 and 1993 NCES Surveys," Economics of Education Review, volume 17, number 1, pp.55-71.

Toutkoushian, Robert (1999), "The Status of Academic Women in the 1990s: No Longer

Outsiders, But Not Yet Equals," Quarterly Review of Economics and Finance, volume 39, Special Issue, p.679-698.

Weiler, William (1990), "Integrating Rank Differences into a Model of Male-Female Faculty

Salary Discrimination," Quarterly Review of Economics and Business, volume 30, number 1, pp.3-15.