Earnings profiles of department heads: comparing cross-section and panel models.
The focus on department heads is interesting for at least two reasons. First, in contrast to the typical upward career path in most sectors of the economy, those who serve as university department head generally are chosen from the ranks of the faculty, and, excluding those who retire, most department heads return to their regular faculty position after serving as head. Second, and of particular interest to the present study, during the period spent as head, administrative activities tend to be substituted for research. Because faculty salaries depend on one's research skills, skills lost or forgone during the period spent as department head are likely to result in slower salary growth after a person steps down as department head.
Salary growth may also slow during the term as department head itself if those serving as head tend to be chosen from the ranks of faculty having above-average productivity and if salary growth of department heads is limited to the salary growth of regular faculty, as is the case at Kansas State University. At least partially offsetting the slower wage growth of present and former department heads, however, is the administrative premium they are paid.
Because wage patterns are affected by service as department head, several interesting issues arise for those contemplating such service. When is the optimal time during a career to accept a term as department head and, from a financial perspective, when is it time to step down? Is the reduced wage growth associated with headship more severe in disciplines for which research skills depreciate rapidly?
After providing a theoretical framework for examining the effect of headship on earnings profiles, we present both cross-section and panel models of wage determination. Despite the advantage of a panel study, which permits us to follow the earnings path of faculty as they move into and out of headship, no panel study of department heads has previously been conducted. Following our presentation of the empirical results for the basic model, we investigate whether service as department head more adversely affects earnings profiles in the sciences, which are characterized by rapid obsolescence of research skills, than in other disciplines.
Prior Research and
Motivation of the Present Study
In most studies examining how service as department head affects earnings, administrative experience has been measured with a dummy variable set equal to one for either present department heads or present and former heads.(1) Invariably, the estimated coefficient is found to be positive and statistically significant, and in most cases the premium is estimated to be 15-30% of salary. An obvious drawback of the dummy specification is that the administrative premium is constrained to be constant over one's career. To be fair, we should point out that many of the studies focused on issues other than the headship premium, but the fact remains that an alternative specification is necessary to investigate the effect of administrative service on wage growth.
Two studies have allowed the administrative premium to vary with length of time served as department head, but neither study provided a specification rich enough to fully capture the wage growth of department heads. Using a quadratic specification, Sauer (1988) found positive but diminishing returns to years of administrative service. But Sauer did not allow for an upward shift in salary when a faculty member initially becomes head, even though department heads receive an immediate premium upon assuming their administrative responsibilities. Because of specification error, Sauer's estimates of the salary premium for administrative service are biased.
The most sophisticated specification incorporated by any prior study was that of Saks (1977), which allowed for an upward shift in salary once a person becomes department head and further permitted the size of the administrative premium to change with time on the job. But Saks's specification failed to allow for a downward shift in salary once a person relinquishes headship of a department. This omission is likely to bias estimates of the premium for administration, especially during the post-headship period, and casts doubt on Saks's findings - that the administrative premium rises by almost 1% per year during the headship period and falls at a roughly comparable rate during the post-headship period. Theoretically, serving as department head may generate an initial wage premium, at least part of which is returned once headship ends, and also affect wage growth during the headship and post-headship periods. To date, no one has tested for all these possibilities.
A further shortcoming of prior studies is that their estimates of administrative premiums are likely biased upward. As Saks (1977:537) explained,
There may be an unobserved selection bias for department chairpersons. Even if the salary differential is calculated after statistically controlling for some measures of quality, there may be unobserved quality differences between department chairpersons and other faculty. If, for example, only the "best' faculty get selected to head departments, then their salary differential would always be positive if salaries are at all related to quality.(2)
Although this problem frequently is acknowledged, prior research, which has focused on cross-section estimation, has not addressed the issue empirically.(3)
After laying out the framework for our analysis, we present cross-section estimates of the effect of headship on earnings of present and former department heads. The cross-section analysis provides a basis for comparison with prior studies and permits us to use an unusually rich array of explanatory variables. This analysis is followed by a panel study that controls for unobserved personal characteristics of those who serve as department head. If those who become department head tend to have unusual drive, intelligence, or other personal characteristics that are correlated with productivity, cross-section estimates of the rewards for administrative service will be biased upward, and panel estimates will provide an indication of the magnitude of this bias.
The Effect of Service as
Department Head on Earnings Profiles
Earnings in the academic labor market, as elsewhere, tend to increase with experience. But earnings, both present and future, are also affected by administrative service. Across universities, a faculty member who becomes department head can expect an administrative supplement and may also have leverage to bargain for an increase in base pay. The size of the initial administrative supplement varies from university to university, as does the procedure for selecting a department head.
At Kansas State University, faculty participate in the selection process, but authority to appoint a department head ultimately rests with the dean of the college. The dean appoints a search committee consisting of faculty from the department in which the position is open. Faculty have the right to recommend candidates to the committee, to assess competing candidates, and to meet privately with members of the committee and with the dean. After obtaining input from the faculty, the search committee forwards to the dean a short list of candidates acceptable to the majority of faculty in the department. The dean chooses among candidates on the list.(4)
Traditionally, department heads at Kansas State University (KSU) have received a 10% increase in monthly pay that is forfeited once they leave the post, but certain faculty are able to negotiate an agreement allowing them to retain a portion of the 10% supplement after returning to a regular faculty position. Because faculty at KSU are not covered by a collective bargaining agreement, those asked by their dean to serve as department head often have room to negotiate over terms of compensation. In addition to the administrative supplement, department heads who have been on nine-month appointments also receive compensation for administrative activities conducted over the summer. But with this added pay comes a new set of responsibilities.
When a faculty member becomes department head, he or she allocates time to acquiring administrative skills at the cost of research skills. According to Moses and Roe (1990), 86% of department heads surveyed reported that administrative responsibilities caused them to reduce their scholarly activities. (See also Dressel et al. 1970.) Not only are fewer research skills acquired during the period as department head, but reduced research activity is also likely to accelerate depreciation of existing skills. In that event, a department head who returns to a regular faculty line will have fewer research skills as a consequence of having served as head. Because non-administrative faculty are rewarded to a large extent on the basis of research output, an implication is that past service as department head is likely to adversely affect research output and therefore to result in a flatter earnings profile than would have occurred if headship had been avoided.
Assuming that wage growth is reduced in the post-headship period, the magnitude of this effect is likely to vary by academic discipline. In particular, one would expect wage growth to be affected most severely in those disciplines in which knowledge depreciates most rapidly.
Service as department head may also lead to lower wage growth during the headship period if the returns to administrative skills are lower than the returns to the forgone research skills. At KSU, the average wage growth of department heads is limited to the average wage growth of faculty.(5) If department heads tend to be selected from the ranks of faculty with above-average productivity, the result will be that those serving as head received above-average pay increases prior to headship but only average pay increases while serving as head. In other words, those serving as department head could have realized higher wage growth had they remained in regular faculty positions.
A simplified graphical representation may be helpful in illustrating a department head's career path. In Figure 1, earnings profiles for regular faculty and for the department head are given by the lines JKLM and ABCDEF, respectively. The line ABE'F' depicts the earnings profile that the person serving as head would have experienced in the absence of headship.
For simplicity, we assume that (1) the earnings profiles are log-linear(6) and (2) the department head, after stepping down, does not later return as head. We may distinguish three phases in the earnings profile of a department head: a pre-headship period (AB), a period as department head (CD), and a post-headship period (EF). In the first phase, AB is the earnings profile of the faculty member before becoming department head. Because most individuals who lead an academic department are elected or appointed on the basis of their scholarly competence (Scott 1980), we expect that the salary profile will be steeper for those who ultimately become department head than for other faculty members (AB will be steeper than JK). We also recognize the possibility that a department head's earnings profile begins at a higher point (A rather than J) because of above-average expected productivity.
In the second phase, BC is the upward salary adjustment (pay premium) of a faculty member when he or she becomes department head. The gap BC depends in part on the bargaining power of the department head while negotiating the salary with higher administration. The earnings profile during headship is represented by the segment CD. Because CD lies above BE', the financial rewards for serving as department head can be depicted by the trapezoid BCDE'.
The slope of CD is drawn to be flatter than the slope of BE' to be consistent with the hypothesis that headship adversely affects earnings growth. Whether the data are consistent with this hypothesis is, of course, an empirical issue.
The third phase begins when the department head re-enters the ranks of the faculty after serving as head. Two changes in that faculty member's profile are expected: a drop in salary (DE) and a change in the slope of the earnings profile. Indeed, if the headship period is lengthy and if research skills have atrophied sufficiently, the department head may step down to a salary below that which would have prevailed if he or she had not served as head (E < E'). On the other hand, if department heads experience only a modest pay cut when they step down and if the headship period is short, so that the cumulative effect of slower wage growth is modest, E is likely to lie above E'.
The second change is in the slope of the earnings profile. A flatter slope (EF) is anticipated because of the interruption in research activities during service as department head. If forgone research limits future pay increases, the financial cost of serving as head is depicted by the area EE'F'F.
In summary, a faculty member who becomes department head can expect an upward shift in the earnings profile, but because of a loss of research skills, service as department head is also likely to result in a flatter profile (slower earnings growth). If earnings growth slows sufficiently, headship eventually may lead to earnings that are lower than would be expected in the absence of administrative service.
A Cross-Section Model
of Academic Salaries
Academic salaries depend on teaching and research skills and, for those who become department head, on administrative activities. Analytically, faculty can be viewed as offering distinct bundles of attributes ([X.sub.1], ..., [X.sub.n]) that are rewarded on the basis of the university's assessment of the value of these attributes:
(1) Earnings = f ([X.sub.1], ...,[X.sub.n)
In this section we list variables capturing key faculty attributes, starting with past and present administrative service; describe the Mincer-type earnings equation to be estimated with cross-section data; and present empirical results for Kansas State University. We then modify the model in the next section so that it can be estimated using panel data.
Based on the preceding theoretical discussion, administrative activity can be measured by a set of five variables: (1) whether or not a faculty member is currently serving as department head (HEAD), (2) the length of the headship period for current heads (HEADSHIP PERIOD), (3) whether or not a faculty member formerly served as head (FORMER HEAD), (4) the length of headship for former heads (FORMER HEADSHIP PERIOD), and (5) the number of years since completing headship for former department heads (POST-HEADSHIP PERIOD).
HEAD captures the initial upward shift in salary for a person serving as department head (equal to BC in Figure 1). HEADSHIP PERIOD measures the rate at which the initial headship premium is eroded. In terms of Figure 1, HEADSHIP PERIOD measures how much more slowly earnings rise during headship than would be expected in the absence of headship - how much flatter CD is than BE'.
FORMER HEAD captures the downward shift in the earnings profile (equal to DE in Figure 1) when a department head steps down. If former department heads are allowed to keep a portion of the administrative premium for serving as head, the downward shift in the earnings profile when headship ends (DE) will be less than the upward shift associated with headship (BC).
In addition to the direct loss in earnings, as at least part of the initial department head premium is removed, former department heads may also experience an indirect loss as a result of reduced wage growth during the headship period. In that event, a person's earnings upon returning to a regular faculty position will depend negatively on the duration of the headship period. To capture this latter effect, a variable (FORMER HEADSHIP PERIOD) is added to measure the length (in years) of a former department head's tenure as head.
The variable POST-HEADSHIP PERIOD indicates what happens to the earnings growth of a department head after he or she steps down. In terms of earnings profiles, the slope in the post-headship period (EF in Figure 1) is predicted to be lower as a consequence of past headship; that is, the coefficient Of POST-HEADSHIP PERIOD is expected to be negative.
Experience. Although we have represented the earnings profile, for simplicity, as a set of linear segments in Figure 1, prior research indicates that experience-earnings profiles tend to be concave. To allow for diminishing returns to experience, a faculty member's experience is measured in quadratic form. More specifically, EXPERIENCE refers to the number of years since attainment of a faculty member's highest degree or number of years employed at the university, whichever is greater.
Citations. Also included in quadratic form is a variable measuring the number of citations to a faculty member's research. Because citations are commonly used as a measure of the quality and quantity of research, we expect that wages will rise with citations, though at a diminishing rate.
Research grants. We also anticipate that faculty will be rewarded on the basis of the volume of grants they generate. Not only are grants a direct measure of research, but they also help reduce overhead for the university.
Books. Another form of research consists of books published by faculty members. Because it is not clear that all categories of books should be weighted equally, we construct separate variables measuring number of textbooks and number of research books.
Graduate faculty. Membership in the graduate faculty is used as another proxy for research productivity. At KSU, graduate faculty status requires refereed publications on research beyond the dissertation, and must be supported by a majority of graduate faculty in the department, the department head, and the Graduate Council of the Graduate School. Therefore, membership in the graduate faculty implies that research productivity has reached some threshold.
Teaching awards. Although universities claim to reward quality teaching, most prior studies have found that outstanding teaching does not significantly affect faculty salaries (Katz 1973; Siegfried and White 1973; Tuckman 1976; Tuckman and Hagemann 1976; and Hansen 1981). Because no university-wide measure of teaching skills is available at KSU, we cannot test for the effect of teaching excellence in our full sample. But data are available on number of teaching awards received by faculty from the College of Arts and Sciences. Such data permit us to re-estimate the cross-section model for the Arts and Sciences subsample after adding a variable for the number of teaching awards received by each faculty member.
University Distinguished Professors. We also include the variable DISTINGUISHED PROF to identify those who have received the University Distinguished Professor Award. This award is limited to faculty recognized nationally and internationally for outstanding scholarly achievements and carries with it an increase in base pay, often $10,000 or more.
Supervision of dissertations. In addition to other measures of productivity, the number of doctoral dissertations supervised is included as a proxy for research ability and the national reputation of a faculty member. Because a Ph.D. dissertation represents original research, "the supervising of a dissertation requires the same ability which research and publication requires" (Katz 1973). Therefore, we expect the number of dissertations supervised to be positively related to salary.
Doctoral degree. We include a dummy variable that indicates whether a faculty member has earned a doctoral degree. We expect that earnings will be higher for those with this degree than for those without it.
Gender and minority status. Dummy variables are also added to control for gender and minority status of the faculty member, with nonwhites classified as minorities. In light of previous research, we expect pay to be lower for women than for men and lower for minorities than for nonminorities.
Model and Data Sources
Based on the preceding discussion, the following cross-section model was estimated for faculty at KSU:
(2) LOG SALARY = [beta.sub.0] + [beta.sub.1] EXPERIENCE
+ [beta.sub.2] [EXPERIENCE.SUP.2] + [beta.sub.3] CITATIONS
+ [beta.sub.44] [CITATIONS.SUP.2] + [beta.sub.5] GRANTS
+ [beta.sub.6] TEXTBOOKS + [beta.sub.7] RESEARCH BOOKS
+ [beta.sub.8] GRAD FACULTY
+ [beta.sub.9] DISTINGUISHED PROF
+ [beta.sub.10] DISSERTATIONS + [beta.sub.11] DEGREE
+ [beta.sub.12] FEMALE + [beta.sub.13] MINORITY + [beta.sub.14] HEAD
+ [beta.sub.15] HEADSHIP PERIOD
+ [beta.sub.16] FORMER HEAD
+ [beta.sub.17] FORMER HEADSHIP PERIOD
+ [beta.sub.18] POST-HEADSHIP PERIOD + [mu]
In addition, a set of departmental dummies was included to control for differences in salaries across disciplines. Summary statistics for the variables included in the cross-sectional specification are presented in Table 1.
For the cross-sectional study, the sample consists of 842 faculty members at or above the rank of assistant professor who were employed at the university during the academic year 1991-92. Because of the inherent difficulty of measuring research activity in art, music, and military sciences, faculty in these departments were excluded from the sample. Also excluded were part-time faculty, administrators other than department heads, heads with supplemental administrative responsibilities, and those working for the state extension service. (Those with extension appointments were excluded because their work tends to involve field activities rather than teaching and research.)
[TABULAR DATA OMITTED]
Sources for the citation variable are the Social Science Citation Index and Science Citation Index, published by the Institute for Scientific Information.1 Sources for all other data are university records. For a detailed description of the data, see Ragan and Rehman (1994).
Salary data exclude payments for summer teaching and research but include administrators' summer salary. At KSU, most faculty members are paid on an academic-year basis (nine months), while most department heads are paid on a full-year basis (eleven months). To make salaries comparable, we adjusted the academic year to a full-year basis by multiplying academic salaries by 11/9, analogous to the procedure adopted by Hamermesh, Johnson, and Weisbrod (1982), Hansen (1985), and Rees (1993). Rees, commenting on this adjustment, observed that
This may seem to be an overadjustment, since some faculty members would like summer employment at full salary and cannot find it. On the other hand, the adjustment does not take into account sources of earned income other than summer employment. Data from the U.S. Department of Education suggest that these factors are roughly offsetting on average. (p. 152)
Consistent with this claim, the Chronicle of Higher Education (February 7, 1990, p. Al 6) found that, on average, full-time faculty receive supplemental earned income outside income plus nonsalary income from their institution) equal to 23% of their base salary.
Results of Cross-Section Specification
Table 2 provides estimated coefficients and statistics for the administrative variables included in equation (2). Estimated coefficients of all other variables have the predicted signs, and all but three - race, textbooks, and (for the Arts and Sciences subsample) teaching awards - are statistically significant. Because the main interest of our research is to measure the effect of administrative duties on earnings profiles of department heads, in this paper we discuss only estimated results for the administrative variables.(8)
Table 2. The Effects on Earnings of Administrative Service, Kansas State University: Regression Coefficients of Department Head Variables, Cross-Section Data, 1991-92. (Dependent Variable = Log of Salary)
Variable Estimated Coefficient t-Value
HEAD 0.3095*** 11.04 HEADSHIP PERIOD -0.0128 -3.27 FORMER HEAD 0.2571*** 4.11 FORMER HEADSHIP PERIOD -0.0083 -1.41 POST-HEADSHIP PERIOD -0.0077 -1.41 Sample Size 842 Adjusted [R.sup.2] 0.7365 F-Value 31.52
Note: Also included in the regression are the other variables listed in equation (2) and a set of departmental dummies. For complete results of estimation, see Ragan and Rehman (1994). Sources: See source note to Table 1. *Statistically significant at the .10 level; **at the .05 level; ***at the .01 level (two-tailed tests).
Based on the estimated coefficients of HEAD and HEADSHIP PERIOD, the average premium for headship (calculated for a department head with 5.4 years of headship experience) is 27.2% [exp (.3095 -.0128 x 5.4) - 1 = .272]. This is identical to the premium estimated by Siegfried and White (1973a) and within the range found by Hamermesh et al. (1982) in the seven universities they studied, but higher than the estimates of Katz (1973) and Saks (1977).
As predicted, department heads are penalized for serving as head over an extended period. The negative (and significant) coefficient of HEADSHIP PERIOD indicates that, for each additional year as department head, salary grows by about 1.3% less than would be predicted, other things equal, if the faculty member were not serving as department head. A faculty member who has served as department head for 5.4 years has already lost one-quarter of the initial headship premium (which was 36.3%).
What happens to salary for someone who steps down as department head? The estimated coefficients Of FORMER HEAD and FORMER HEADSHIP PERIOD indicate that the typical former department head (who served 7.5 years as head) initially earns 21.5% more than would be expected had he or she not served as head. As with the estimated headship premium, this figure overstates the true compensation for prior headship if headship is positively correlated with unmeasured productivity characteristics; we examine the extent of this bias in the next section. But cross-section results suggest that former department heads initially take much of their administrative premium with them when they leave the headship.
Of course, the carry-over headship premium may erode during the post-headship period. Indeed, this is what our model predicts as the consequence of lost research skills. Consistent with this view, the coefficient Of POST-HEADSHIP PERIOD is negative, though not statistically significant.
In summary, empirical results from the cross-section sample suggest that department heads are compensated for their administrative service during and after headship. These results are tempered, however, by the high likelihood that a portion of the estimated administrative premium is capturing unmeasured personal characteristics of department heads.
A Panel Study of Academic Salaries
Because of correlation between headship and unobserved personal characteristics, the headship and post-headship premiums estimated by the cross-section model are likely to be biased upward. To correct for these biases, we also estimated earnings of department heads with a panel study that incorporates personal fixed effects. Empirical estimates of the panel study are divided into two parts. In the first part, regression estimates are presented for the university at large (using the same disciplines covered in the cross-section study). In the second part, we examine whether experience-earnings profiles of department heads in the sciences differ from those of department heads in non-science disciplines.
Model and Data
For the panel study, we collected data for faculty who served as department head at any time during the period 1965-92. This procedure generated a pool of 170 department heads. Twenty-one department heads, those who joined the university before 1950 or completed their degree before 1950, were dropped from the pool. Because data on citations are not available for researchers prior to 1955, citations cannot be measured accurately for those who began their research in the period prior to 1950. This restriction left a sample of 149 department heads and a total of 2,259 observations.
The basic model estimated with panel data is similar to the cross-section model, with several important exceptions. Because data on books and research grants first became available in 1986 and 1989, respectively, these variables had to be deleted from the panel model. With the gender and minority status of each faculty member constant, variables measuring these characteristics were also excluded. The most important addition to the model is the set of dummy variables to measure the personal characteristics of individual faculty. Also included is real per-capita income of the state lagged one year (INCOME). This variable is designed to capture the relationship between economic conditions in the state and real salaries of the faculty. Higher per-capita income creates the potential for higher faculty salaries. This variable allows us to test whether that potential has been realized at this state university." Finally, to control for the effects of inflation, nominal salary was deflated by the Consumer Price Index.
With these changes, the earnings function of the fixed-effects panel model becomes
(3) LOG REAL [SALARY.sub.it] = [gamma.sub.oi] + [gamma.sub.1] [EXPERIENCE.sub.it] + [gamma.sub.2]] [EXPERIENCE.sup.2] + [CITATIONS.sub.it] + [gamma.sub.4] [CITATIONS.sup.2.sub.it] + [gamma.sub.3] [CITATIONS.sub.it] + [gamma.sub.6] DISTINGUISHED [PROF.sub.it] + [gamma.sub.7] [DISSERTATION.sub.it] + [gamma.sub.8] [DEGREE.sub.it] + [gamma.sub.9] [INCOME.sub.it] + [gamma.sub.10] [HEAD.sub.it] + [gamma.sub.11] HEADSHIP [PERIOD.sub.it] + [gamma.sub.12] FORMER [HEAD.sub.it] + [gamma.sub.13] FORMER HEADSHIP [PERIOD.sub.it] + [gamma.sub.14] POST-HEADSHIP [PERIOD.sub.it] + [epsilon.sub.it],
where LOG REAL [SALARY.sub.it] is the log of real annual salary of faculty member i in period t. Summary statistics of the explanatory variables are given in Table 3, and empirical results appear in Table 4.
[TABULAR DATA OMITTED]
Consistent with cross-section findings, the panel results show that faculty receive positive but diminishing returns to both experience and citations, and those with a doctoral degree earn more than faculty without the degree. But in contrast to the cross-section results, the estimated coefficients of GRAD FACULTY and DISSERTATIONS are negative, though only the latter is significantly less than zero.
We are inclined to take these latter results at face value and to argue that the reason estimated. coefficients were positive in the cross-section equation is that the cross-section equation lacked personal fixed effects. If graduate students are attracted to major professors of above-average drive, ability, reputation, and organization, DISSERTATIONS (and perhaps GRAD FACULTY) Will be proxies for such faculty characteristics. In other words, the coefficient for DISSERTATIONS in the cross-section equation will measure the return to ability and other personal characteristics rather than additional compensation for working with graduate students. Indeed, to the extent that serving as major professor leads to a smaller volume of publications and other output that is rewarded more highly than supervising dissertations, one would expect a negative return for serving as major professor - once personal characteristics of the faculty are taken into account.
The coefficient of INCOME is negative and highly significant, indicating that rising per-capita income in the state has not translated into higher faculty salaries. Although real faculty salaries have shown little trend (the correlation between TIME and log of real faculty salaries is 0.07), the quality of faculty has increased over time. Both experience and citations are positively and significantly correlated with TIME. An analysis that adjusts for the increasing quality of faculty shows that the trend in salaries has actually been downward. Given the extremely high correlation between TIME and real per-capita income (r = 0.97), our interpretation is that the income variable is essentially a time trend that captures the declining propensity of the state to fund faculty salaries.
Consistent with this interpretation, a report of the faculty senate found that faculty salaries at KSU decreased by 6% in real terms between 1970 and 1992 even though state real per-capita income advanced by 39%. The conclusion of both that report and this study is that increasing per-capita income in the state has not translated into higher faculty salaries. To the contrary, funding for salaries has eroded.
Results for headship variables. The estimated coefficient Of HEAD is highly significant but only half as large as in the cross-section equation (.156 versus .310). Because the cross-section equation does not account for the personal characteristics of those serving as department head, the implication is that roughly 50% of the initial headship premium as estimated by the cross-section model reflects above-average productivity of those who serve as department head; only 50% of the estimated premium is compensation for headship.
As predicted, the coefficient of number of years as department head, HEADSHIP period is negative and statistically significant. A person who has served as department head for 4.9 years (the average tenure of current department heads in the panel study) has already lost 5 percentage points of the headship premium. While the estimated premium is 16.9% for a new department head, it is only 12.2% for the average department head [exp (.1559-.0083x4.9) - 1=.122]. Table 4. Estimates of the Earnings Equation for Those Who Served as Department Head at Kansas State University:
Regression Coefficients, Panel Data. (Dependent Variable = Log of Real Salary) Variable Regression Coefficient t-value EXPERIENCE 0.0212*** 15.46 EXPERIENCE(2) -1.42x10-(4)*** -4.21 CITATIONS 0.98x10-(4)* 1.74 CITATIONS(2) -1.09X10-(7)*** -2.93 GRAD FACULTY 0.0111 -1.56 DISTINGUISHED PROF 0.1667*** 7.21 DISSERTATIONS -0.0031*** -4.24 DEGREE 0.0620*** 4.38 INCOME -0.0029*** -6.79 HEAD 0.1559*** 27.20 HEADSHIP PERIOD -0.0083*** -9.23 FORMER HEAD 0.1206*** 9.37 FORMER HEADSHIP PERIOD -0.0125*** -8.72 POST-HEADSHIP PERIOD -0.0089*** -6.10 Sample Size 2259 Adjusted R2 0.9006 F-Value 127.24
Also included in the equation is a set of faculty dummies.
In the absence of fixed effects, adjusted R(2) = 0.6318.
Sources: See source note to Table 3.
*Statistically significant at the .10 level; **at the .05 level; ***at the .01 level (two-tailed tests).
Consistent with findings for headship, a comparison of the coefficient Of FORMER HEAD in the panel and cross-section studies suggests that unobserved personal characteristics account for much of the estimated premium associated with former headship in the cross-section equation. Note that the estimated coefficient Of FORMER HEAD (0.1206) does not indicate the average wage premium of former department heads. The actual premium former department heads receive depends on the duration of headship. Consistent with our earlier findings that wages rise more slowly while a person is department head, the estimated coefficient Of FORMER HEADSHIP PERIOD indicates that each year of headship reduces a former department head's wages by approximately 1.3%. This means that a former department head whose headship lasted 7.3 years (the average among former heads in our sample) would initially receive a wage premium of 3.0% after stepping down as head.
The next question is whether this modest post-headship premium can be maintained. Our model suggests otherwise. If research skills depreciate during headship, one would expect less research output and therefore slower wage growth in the post-headship period. The highly significant coefficient OFPOST-HEADSHIP PERIOD provides evidence in support of this proposition. Wages rise 0.9% less per year, other things equal, during the post-headship period. After four years as former department head (the mean in our sample), wages are lower than would be predicted had the person not become head. In other words, the initial post-headship premium has completely disappeared for the average former department head in our sample.
Of course, the actual premium or discount a former department head receives depends on both tenure as head and length of time since the headship was relinquished. Table 5 presents the estimated post-headship premium or discount as a function of these two variables. For example, a person who served as department head for five years receives an estimated wage premium of 6.0% immediately after leaving headship but a wage discount of 3.9% after eleven years as former department head. An implication of Table 5 is that both the length of headship and its timing are important. Because those who serve early in their careers have a longer post-headship period during which wage growth slows, the present value of headship is lower for those who become department head at an early age than for those who wait until close to retirement.
Table 6 provides additional perspective on this issue. The top row of the table estimates the present value of the change in earnings attributable to a three-year headship, assuming a 5% discount rate.10 For someone becoming department head at age 40, the lifetime present-value amounts to 43.6% of pre-headship salary (salary in the year immediately prior to becoming department head). For a person who delays service as department head until age 60, the present value of headship is 62.2% of pre-headship salary. Other rows of Table 6 consider the present value of headships that last more than three years.
Just as one can from a financial perspective become department head too early in one's career, given the premium former department heads typically enjoy early in the post-headship period it is also possible to wait too long before accepting headship of a department. This is also evident from Table 6, which shows that the present value of a three-year headship is greater if that administrative service begins at age 50 than if it begins at age 60. More generally, the optimal time to become department head depends on the expected tenure as head. Based on estimated coefficients, department heads planning to serve for three years will maximize the expected value of headship by beginning their appointment thirteen years prior to retirement. Similarly, those planning to serve four to six years should begin twelve years prior to retirement, those planning to serve seven or eight years should begin eleven years prior to retirement, and those planning to serve nine or ten years should begin ten years prior to retirement.
Clearly, nonfinancial considerations are also relevant for those contemplating service as department head. Faculty for whom nonwage characteristics of the position are sufficiently onerous will maximize expected utility by avoiding headship. And those who become department head may tire of administrative responsibilities and choose to step down even though the present value of an additional year of service as department head is positive. Indeed, there is evidence of such resignations occurring at KSU. Even though the present value of service as department head increases with years served (see Table 6), the average term of a department head at KSU is seven years. This finding is consistent with the hypothesis that, on the margin, administrative premiums are necessary to compensate for the negative attributes of administration.
Table 6. Present Value of Headship
(Expressed as Percentage
of Pre-Headship Salary).
Age When Headship Begins: Length of Headship 40 Years 50 Years 60 Years 3 Years 43.6 75.1 62.2 5 Years 43.9 78.6 70.4 7 Years 46.1 84.0 10 Years 52.5 94.9
Assumptions: earnings rise with experience, which is assumed to equal age minus 30; coefficients for the experience and headship variables are taken from Table 4; faculty member retires after completing the year in which he or she turns 65 years of age; real discount rate is 5%.
Do returns to serving as department head differ by field? McDowell (1982) found that in the physical sciences research knowledge is less durable (research technology advances at a more rapid rate) than in the social sciences and humanities. One might also anticipate rapid advances in biochemistry, computer sciences, and various other disciplines. Although any partitioning of disciplines is subjective, we use Berger's (1992) science/engineering classification as a proxy for disciplines experiencing rapid advances and therefore rapid depreciation of research skills. Included in this "science" category are physical and life sciences, veterinary medicine, agriculture, engineering, architecture, computer sciences, mathematics, and statistics.
Because of more rapid depreciation of research skills in the sciences, we hypothesize that wage growth during the period following service as department head will be slower in the sciences than in other disciplines. In addition, if department heads' pay is based in part on research output relative to that of other faculty in the department, a more rapid depreciation of research skills should also lead to slower wage growth during the term as department head for those in the sciences. Finally, to compensate for this slower wage growth, we hypothesize that department heads in the sciences will negotiate a higher initial headship premium than department heads in other disciplines.
To test these hypotheses, we estimated a separate regression similar to equation (3), but with the addition of interaction terms to capture differences in returns to administrative service for department heads who come from the sciences. For example, FORMER HEAD SCIENCE in Table 7 is a dummy variable identifying those who are former heads of a department included in our science category. The coefficients for years as department head (HEADSHIP PERIOD SCIENCE) and total years since stepping down as department head (POST-HEADSHIP PERIOD SCIENCE) are expected to be negative because, as argued above, depreciation of research skills is more rapid in the sciences than in other disciplines. In turn, if the coefficient OF HEADSHIP PERIOD SCIENCE is negative, the coefficient Of FORMER HEADSHIP PERIOD SCIENCE (tenure as department head for former department heads) should also be negative - the greater the annual earnings loss during service as department head, the greater will be the cumulative penalty of a given headship in the post-headship period.
Finally, if department heads in the sciences are aware of the higher price they pay for headship relative to the price paid by department heads in non-science disciplines, they will demand a larger initial headship premium than do faculty in those disciplines. Thus, the coefficients Of HEAD SCIENCE and FORMER HEAD SCIENCE are expected to be positive.
As Table 7 indicates, coefficients of all variables associated with service as department head have the predicted signs and are significant at the .01 level. (Signs of all other coefficients are the same as in Table 4.) Department heads in the sciences initially receive a wage premium 3.5 % greater than department heads in other disciplines, but they are penalized more severely during their terms than department heads from other fields. In particular, service as department head reduces wage growth by 1.0% per year in the sciences compared to only 0.5% per year in other disciplines, a difference significant at the 0.01 level.
Also supporting the hypothesis that research skills depreciate more rapidly in the sciences are the negative coefficients of FORMER HEADSHIP PERIOD SCIENCE and POST-HEADSHIP PERIOD SCIENCE. In all disciplines, wages of former department heads are negatively related to tenure as head and to length of the post-headship period; but the penalty is more severe for department heads in the sciences. On the other hand, the slower wage growth of former department heads in the sciences is at least partially offset by the fact that these former heads tend to return to faculty lines with more of their headship premium intact. That is, the coefficient Of FORMER HEAD SCIENCE iS positive.
Does the initially greater premium received by department heads and former department heads in the sciences offset their slower wage growth? The answer depends both on the length of service as department head and on the time that has elapsed since stepping down from that position. What we can say is that, based on coefficients from Table 7, the "typical" former department head in the sciences, who served as head for 7.1 years and left that position 4.0 years ago, is earning wages that are virtually identical to (within 0.1% of) the wages that would have been received had headship been avoided; the typical" former department head in other disciplines, in comparison, is receiving wages 2.1% below the wages that would have been received had that position been avoided. Thus, for our sample, former department heads in the sciences do not appear to be at a disadvantage compared to their counterparts in other disciplines.
Table 7. The Differential Effects
on Earnings of Service as
the Sciences and the Non-Sciences:
Regression Coefficients, Panel Data.
(Dependent Variable = Log of Real Salary) Variable Regression Coefficient t-value EXPERIENCE 0.0205*** 14.81 EXPERIENCE(2) 1.20 x 10-(4)*** -3.48 CITATIONS 1.26 x 10-(4)*** 2.19 CITATIONS(2) -1.38 x 10-(7)*** -3.61 GRAD FACULTY -0.0112 -1.57 DISTINGUISHED PROF 0.1850*** 7.93 DISSERTATIONS -0.0033*** -4.48 DEGREE 0.0669*** 4.73 INCOME -0.0028 -6.72 HEAD 0.1396*** 18.38 HEADSHIP PERIOD -0.0053*** -4.51 FORMER HEAD 0.0655*** 3.31 FORMER HEADSHIP PERIOD -0.0084*** -3.93 POST-HEADSHIP PERIOD -0.0053*** -2.75 HEAD SCIENCE 0.0296*** 3.22 HEADSHIP PERIOD SCIENCE -0.0050*** -4.03 FORMER HEAD SCIENCE 0.0960*** 3.84 FORMER HEADSHIP PERIOD
SCIENCE -0.0072*** -2.84 POST-HEADSHIP PERIOD
SCIENCE 0.0072*** -2.99 Sample Size 2259 Adjusted R(2) 0.9019 F-Value 125.25
Also included in the equation is a set of faculty dummies.
In the absence of fixed effects, adjusted R2 = 0.6495.
Sources: See source note to Table 3.
*Statistically significant at the .10 level; **at the .05 level; ***at the .01 level (two-tailed tests).
Summary and Implications
Applying human capital theory to the academic labor market, we derived various hypotheses about the consequences of research and administrative activities on the earnings of faculty and department heads. To test these hypotheses, we estimated earnings equations for faculty at Kansas State University using both cross-section and panel models of wage determination. Our results indicate that earnings are significantly related to professional experience, citations, volume of research grants, and publication of research books. Higher salaries are also received by those with a terminal degree and by those awarded a University Distinguished Professorship.
Consistent with prior research, we find that those serving as department head receive higher salaries than other faculty. According to the cross-section model, headship is associated with an average salary premium of 27%. But this figure captures both the compensation for serving as department head and the unmeasured productivity characteristics of those chosen to lead a department. To control for unmeasured productivity characteristics, we reestimated the earnings equation with a panel model incorporating personal fixed effects. The results indicate that the average premium for serving as department head at KSU is 12%, less than half the cross section estimate.
Because serving as department head entails the sacrifice of some research skills, it leads to slower salary growth in the period following such service. At KSU, salary growth is also reduced during the term as department head by an informal agreement that pay increases of department heads will not outpace those of regular faculty even though those serving as department head tend to be drawn from the ranks of high-productivity faculty. Although former department heads retain a portion of their headship premium when they initially step down, the longer they serve as head, the less of the premium they keep. For the average former department head in our sample, the premium for past administrative service has completely disappeared. And for those whose service as department head was extensive or occurred early in their career, salaries are lower than would be expected had they never performed that service.
The cost of serving as department head, in terms of reduced wage growth during and after the period as head, is greater for faculty in the sciences, where obsolescence of research skills occurs rapidly, than for faculty in other disciplines. As compensation for this slower wage growth, department heads in the sciences receive a larger initial headship premium than do those in other disciplines.
In summary, our findings indicate that faculty salaries depend on research output and on administrative service; that service as department head leads to an initial administrative premium that erodes over time; that cross-section estimates of this administrative premium are biased upward; and that salaries of former department heads depend (negatively) on both the tenure as department head and the number of years since that position was relinquished.
Although this paper's empirical findings are based on data from Kansas State University, certain propositions derived from our model can be generalized across universities. For example, the model predicts that, because of depreciation of research skills, wage growth in the period following service as department head will be adversely affected at universities in general. Whether wage growth also slows during the period spent as department head depends on the university's policy of rewarding administrators, but even here our model offers insights.
At such universities as KSU, where wages of faculty and department heads rise at the same rate, those serving as department head can anticipate lower pay increases than they would have received had they remained in regular faculty positions. But not all universities adopt the Kansas State model. At Michigan State University, for example, the pay of department heads rises faster on average than the pay of regular faculty. What we can predict, given department heads' higher wage growth at Michigan State, is that the initial administrative premium necessary for faculty to accept headship will be lower there than at KSU. Consistent with this claim, Saks (1977) estimated that the initial premium for department heads was 13% of salary at Michigan State, compared to the initial 18% premium we found at Kansas State; and given the cross-section nature of Saks's study, the 13% estimate is likely biased upward.
More generally, because different wage streams are consistent with a given present value of headship, we predict a trade-off across universities between the size of the inital premium received by department heads and the rate of wage growth during tenure as head. This is a testable proposition that can be examined by applying the panel model of the present paper to salary determination at other universities. (1) This is the approach adopted by Katz (1973), Siegfried and White (1973a, 1973b), Tuckman (1976), Tuckman and Hagemann (1976), Tuckman, Gapinski, and Hagemann (1977), Hansen (1981), Hamermesh, Johnson, and Weisbrod (1982), and Kaun (1984). (2) In contrast to KSU's department head system, many universities rely on a system of department chairpersons. Because differences between department chairpersons and department heads are subtle, however, we expect research results based on the alternative administrative systems to be similar. (3) The one exception was Hamermesh (1989), who gathered data on 100 professors at two points in time - 1979/80 and 1985/86 - and then estimated a wage-change equation. Although Hamermesh found that moving into administration was associated with a 13% increase in salary, confidence in these results is limited because few faculty became new administrators over this period and because information on length of service as department head and time elapsed since that service ended was not incorporated into the estimation. (4) For additional detail, see sections B120-124 in the Kansas State University Faculty Handbook. (5) The faculty senate passed a resolution calling for pay of administrators including department heads to rise no faster than the pay of faculty. The central administration has adhered to this request. (6) This assumption will be relaxed in the empirical part of the paper. (7) A limitation of citation indices is that they list only the first author of multiple-author articles, but Diamond (1986a, 1986b) found that the explanatory power of citations decreased when all authors of joint research were given full credit and that adding a variable on citations of non-first authorship failed to increase the explanatory power of the wage equation. Although these findings were based on a small sample, they suggest that the first-author citation count is a good proxy for total citations. (8) The full results of the estimation are reported in Ragan and Rehman (1994) and are available on request to the authors. (9) Data on nominal per-capita income of the state were obtained from the U.S. Department of Commerce, Survey of Current Business (I 992). This series was then deflated by the Consumer Price Index to obtain real per-capita income of the state.
(10) The present value of service as department head is the difference between the present value of earnings assuming a person becomes department head and the present value of earnings assuming the opposite. The effect of service as department head on earnings is calculated on the basis of the estimated coefficients of the administrative variables from Table 4). Real earnings are assumed to rise with experience as indicated by the estimated coefficients of EXPERIENCE and EXPERIENCE(2), and the faculty member is assumed to retire after completing the year in which he or she turns 65 years of age. The present value calculations abstract from the possible effects of service as department head on the values of other variables.
Berger, Mark C. 1992. "Private Returns to Specific College Majors." In W. E. Becker and D. R. Lewis, eds., The Economics of American Higher Education. Boston: Kluwer, pp. 141-71. Broder, Josef M., and Rod F. Ziemer. 1982. "Determinants of Agricultural Economics Faculty Salaries." American Journal of Agricultural Economics, Vol. 64, No. 2 (May), pp. 301-3. Center for the Study of the Department Chair. 1990. "Career Path of Department Chairs." Newsletter, Department of Educational Administration, Washington State University (Spring). DeLorme, Charles D., Jr., R. Carter Hill, and Norman J. Wood. 1979. "Analysis of a Quantitative Method of Determining Faculty Salaries." Journal of Economic Education, Vol. 11, No. 1 (Fall), pp. 20-25. Diamond, A.M., Jr. 1986a. "The life-Cycle Research Productivity of mathematicians and Scientists." Journal of Gerontology, Vol. 41, No. 4, pp. 520-25. _____. 1986b. "What Is a Citation Worth?" Journal of Human Resources, Vol. 21, No. 2 (Spring), pp. 200-215. Dressel, P. L., F. C. Johnson, and P. M. Marcus. 1970. The Confidence Crisis. San Francisco: Jossey Bass. Ferber, Marianne, and Carole Green. 1982. "Traditional or Reverse Sex Discrimination? A Case Study of a Large Public University." Industrial and Labor Relations Review, Vol. 35, No. 4 July), pp. 550-64. Hamermesh, Daniel S. 1989. "Why Do Individual-Effects Models Perform So Poorly? The Case of Academic Salaries." Southern Economic Journal, Vol. 56, No. 1 (July), pp. 39-45. Hamermesh, Daniel S., George E. Johnson, and Burton A. Weisbrod. 1982. "Scholarship, Citations, and Salaries: Economic Rewards in Economics." Southern Economic Journal, Vol. 49, No. 2 (October), pp. 472-81. Hansen, Richard B. 1981. "The Sensitivity of Variable Measurement in Faculty Salary Studies." Journal of Behavioral Economics, Vol. 10, No. 1 (Summer), pp. 1-12. _____. 1985. "A Test of the Hansen-Weisbrod-Strauss Model of Faculty Salaries." Atlantic Economic Journal, Vol. 13, No. 3 (September), pp. 33-40. Katz, David A. 1973. "Faculty Salaries, Promotions, and Productivity at a Large University." American Economic Review, Vol. 63, No. 3 June), pp. 469-77. Kaun, David E. 1984. "Faculty Advancement in a Nontraditional University Environment." Industrial and Labor Relations Review, Vol. 37, No. 4 July), pp. 592-606. McDowell, John M. 1982. "Obsolescence of Knowledge and Career Publication Profiles." American Economic Review, Vol. 72, No. 4 (September), pp. 752-68. Mincer, Jacob. 1974. Schooling, Experience, and Earnings. New York: National Bureau of Economic Research. Mincer, Jacob, and Haim Ofek. 1982. "Interrupted Work Careers: Depreciation and Restoration of Human Capital." Journal of Human Resources, Vol. 17, No. 1 (Winter), pp. 3-24. Mincer, Jacob, and Solomon Polachek. 1974. "Family Investments in Human Capital: Earnings of Women." Journal of Political Economy, Vol. 82, No. 2 (April), pp. S76-108. _____. 1978. "Women's Earnings Reexamined." Journal of Human Resources, Vol. 13, No. 1 (Winter), pp. 118-34. Moses, Ingrid, and Ernst Roe. 1990. Heads and Chairs: Managing Academic Departments. Queensland: University of Queensland Press. Ragan, James F., Jr., and Qazi Rehman. 1994. "Using Panel Data to Measure the Effect of Headship on Earnings and the Bias from Cross-section Estimation." Working Paper, KSU, December. Rees, Albert. 1993. "The Salaries of Ph.D.s in Academe and Elsewhere." Journal of Economic Perspectives, Vol. 7, No. 1 (Winter), pp. 151-58. Rehman, Qazi N. 1993. "Earnings Profiles of Department Heads." Doctoral dissertation, KSU, December. Saks, Daniel. 1977. "How Much Does a Department Chairman Cost?" Journal of Human Resources, Vol. 12, No. 4 (Fall), pp. 535-40. Sauer, Raymond D. 1988. "Estimates of the Returns to Quality and Coauthorship in Economic Academia." Journal of Political Economy, Vol. 96, No. 4 (August), pp. 855-66. Siegfried, John J., and Kenneth J. White. 1973a. "Financial Rewards to Research and Teaching: A Case Study of Academic Economists." American Economic Review, Vol. 63, No. 2 (May), pp. 309-15. _____. 1973b. "Teaching and Publishing as Determinants of Academic Salaries." Journal of Economic Education, Vol. 4, No. 2 (Spring), pp. 90-100. Scott, R. A. 1980. The Department Chairperson: Role, Function, and Future. New York: American Council on Education. Tuckman, Howard P. 19 76. Publication, Teaching, and the Academic Reward Structure. Toronto: Lexington Books. Tuckman, Howard P.,and Jack Leahey. 1975. "What Is an Article Worth?" Journal of Political Economy, Vol. 83, No. 5 (October), pp. 951-67. Tuckman, Howard P., and Robert P. Hagemann. 1976. "An Analysis of the Reward Structure in Two Disciplines." Journal of Higher Education, Vol. 47, No. 4 (July), pp. 447-64. Tuckman, Howard P., James H. Gapinski, and Robert P. Hagemann. 1977. "Faculty Skills and the Salary Structure in Academe: A Market Perspective." American Economic Review, Vol. 67, No. 4 (September), pp. 692-702.
|Printer friendly Cite/link Email Feedback|
|Author:||Ragan, James F., Jr.; Rehman, Qazi Najeeb|
|Date:||Jan 1, 1996|
|Previous Article:||The response of wages to protective labor legislation: evidence from Canada.|
|Next Article:||Racial differences in access to high-paying jobs and the wage gap between black and white women.|