Printer Friendly

Will changing times change the allocation of faculty time?

I. Introduction

Faculty time is a key university resource, and how professors use their time has come under increasing public scrutiny. The malleability of faculty time allocations is an important issue because the combination of increasing fiscal constraints and public pressures to enhance teaching may require major adjustments in the future. Although general time allocation issues have received much attention in the economics literature (Juster and Stafford 1991), analyses of the time allocation decisions of faculty have been few and largely descriptive (Yuker 1984, Bunnell 1960). Previous studies of faculty employment have focused on measures of faculty output such as publications and citations, departmental rankings, collective bargaining, and compensation (Ransom 1993, Stephan and Levin 1991, Barbezat 1987).

In this paper we study faculty inputs, specifically time allocation decisions, and build on previous descriptive studies by providing a behavioral context for these decisions. Universities provide a useful setting to study time allocation decisions for several reasons. First, faculty perform several tasks (for example, teaching, research, and service) that, although related, are nevertheless distinct. Second, institutional missions and incentive structures vary across institution types. These characteristics of academia permit us to model and examine empirically how individual and institutional differences affect time allocation decisions.

To motivate our empirical approach, we develop a random-utility model for academics who allocate their time among four alternatives: teaching, research, service, and nonuniversity activities. Consistent with hedonic life-cycle models of human-capital investment (Rosen 1974), time is allocated to the various alternatives based upon their relative expected returns. This approach yields four share equations that describe the percentage of time spent in each activity. A key advantage of our formulation is that it permits substitution among alternatives in the faculty choice set, including substitutions between teaching and research as documented in prior work (Paul and Rubin 1984).

The data for the analysis include a large number of faculty from public and private universities which are categorized by four distinct educational missions. For each institution type, four share equations are estimated simultaneously using grouped-data, multinomial logit. The estimates are used to conduct two separate analyses. First, the marginal time allocation probabilities and their standard errors are calculated for changes in personal and institutional attributes. The results support prior expectations of how such attributes might affect the allocation of time, and they also suggest possible time allocation differences by gender, nationality, ethnicity, discipline, and union status that are not obvious, a priori. Second, differences in the total allocation probabilities among the four institution types are decomposed into components that reflect coefficient and characteristic differences between institutions and their faculty. The results suggest that the incentive structure of a university plays a primary role in time allocation decisions and that faculty attributes tend to reinforce an institution's mission. In other words, the institutional settings under which faculty are hired significantly condition the time allocation response of faculty to policies for change.

II. The Model

Individual faculty members face differing rewards from the allocation of time to teaching (T), research (R), service (S), and leisure (L) because personal attributes and institutional constraints differ and because faculty preferences for each pursuit vary.(1) The utility derived from pursuit K by faculty member i at time t ([U.sub.Kit]) is thus assumed to depend contemporaneously on both personal and institutional characteristics ([X.sub.it], [I.sub.it]) and salary ([V.sub.it]), which is a proxy for innate ability and past time allocation decisions:

(1) [U.sub.Kit] = [u.sub.K]([X.sub.it], [I.sub.it], [V.sub.it]) + [[Epsilon].sub.Kit]

K = (S, T, R, L)

where [u.sub.K] represents observable returns to pursuit K and [[Epsilon].sub.Kit] is a random variable measuring unobserved returns.

We assume that the utility derived from each pursuit depends linearly on current salary and vectors of personal and institutional attributes:

(2) [u.sub.K]([X.sub.it], [I.sub.it], [V.sub.it]) = [X.sub.it][[Alpha].sub.K] + [I.sub.it][[Beta].sub.K] + [V.sub.it][[Gamma].sub.K].

The coefficients [[Alpha].sub.K], [[Beta].sub.K], and [[Gamma].sub.K] vary by activity (K) but not by individual (i) or time period (t). This is appropriate for academics who work in a competitive labor market and where in equilibrium they would earn the same observed return for a given combination of attributes in a particular pursuit. The observed returns across pursuits need not be equal, however, because attributes are unlikely to contribute equally to academic output in each choice. Moreover, universities differ in their demands for these traits, so institutional factors affect these returns.

A faculty member allocates time to the pursuit over a time interval t, [Y.sub.it], which provides the greatest utility. For example, the probability that academic i chooses to spend an hour in teaching activities, Pr[[Y.sub.it] = T], equals Pr[[U.sub.Tit] [greater than] [U.sub.Kit]] for all K [not equal to] T. Substituting equations (1) and (2) into Pr[[Y.sub.it] = T] yields:

(3) Pr[[[Epsilon].sub.Tit] - [[Epsilon].sub.Kit] [greater than] [X.sub.it]([[Alpha].sub.K] - [[Alpha].sub.[Tau]]) + [I.sub.it]([[Beta].sub.K] - [[Beta].sub.[Tau]]) + [V.sub.it]([[Gamma].sub.K] - [[Gamma].sub.[Tau]])]

for all K [not equal to] T.

Thus, in this case, faculty member i is more likely to spend the next hour in a teaching capacity the greater i's relative productivity in, or preference for, teaching and the larger the relative rewards to teaching offered by i's institution.

The hour-by-hour time allocation decisions, while not actually observed, determine the proportion of a week spent in each pursuit, [P.sub.Kit], a variable included in our data. In the context of equation (3), [P.sub.Kit] represents the proportion of hours in a week for which the return to pursuit K is greater than its alternatives. If [[Epsilon].sub.Kit] is independent and identically distributed with a Weibull distribution, [P.sub.Kit] provides an estimate of Pr[[Y.sub.it] = K] in a multinomial, grouped-data logit model:

[Mathematical Expression Omitted].

The formulation of the four share equations in (4) is consistent with the data used in the empirical analysis to follow. While the time allocation choice specified in (4) depends linearly on personal and institutional attributes, possible nonlinear effects are accounted for by including quadratic or higher-order terms in the X, I, and V vectors.(2)

III. The Data

The data for the analysis are from the National Survey of Postsecondary Faculty (NSOPF). A self-reported questionnaire was administered to a random sample of 8,000 instructional faculty at 480 U.S. institutions in 1987 by the National Center for Education Statistics of the U.S. Department of Education. These data are unique in that they detail time allocated to teaching, research, service, and leisure and include comprehensive information on institutional characteristics ([I.sub.it]), salary ([V.sub.it]), and personal and employment characteristics ([X.sub.it]) for a large number of faculty. Other faculty surveys, which are summarized by Creswell, Chronister, and Brown (1991), are either (1) less comprehensive in their coverage (for example, Office of Educational Research Survey 1988), (2) sample faculty from a narrow set of institution types (for example, only liberal arts institutions are included in the Council of Colleges Survey 1986), or (3) contain significantly fewer observations (for example, the Carnegie Survey 1989). These elements limit controls for faculty heterogeneity and comparisons across institution types either directly by exclusion or indirectly by providing small numbers of observations of "similar" individuals and/or institutions. The NSOPF is thus relatively well suited for an analysis of faculty time allocation.

To minimize unobserved individual heterogeneity, we focus on a subsample of 1,409 full-time, tenure-track, and tenured arts and sciences faculty at four-year colleges and universities. These institutions are either "premier" (PREMIER), PhD-granting (DOCTORAL), comprehensive (COMP), or liberal arts (L. ARTS) colleges and universities.(3) This categorization is important because our empirical analysis requires the incentive structure for time allocation decisions to be "similar" within an institution type so that salary adequately reflects the rewards (or penalties) to innate skills and past time allocation decisions, while variation across institution types may suggest how important incentives are to current time allocation decisions.

The sample participants indicated their average number of hours worked per week and the proportion of work time spent in teaching, research, and other work-related activities (defined here as service).(4) Descriptive statistics in Table 1 demonstrate, not surprisingly, that the relative research emphasis is greatest at premier institutions. Converting the statistics in Table 1 from the percent of total time in an activity to percent of work time in an activity, we find that faculty at premier institutions spend nearly 33 percent of their work time in research activities and 46 percent in teaching activities during the academic year. This is in contrast with liberal arts institutions, where faculty spend 10 percent of work time in research and 68 percent in teaching. Even at research-oriented institutions, where promotion and tenure decisions and salary increases are heavily determined by success in research, faculty only spend a third of their time during the academic year engaged in research.(5) Percentage of time devoted to service activities is relatively constant across institution types and accounts for approximately 22 percent of faculty work time. It is also interesting to note that the total hours devoted to work is highest at public premier institutions, where faculty work an average of 51.6 hours per week.(6) This finding is consistent with those of Myers and Mager (1980) that identify a positive relationship between the percent of time devoted to research and the total number of hours worked per week. However, it should be pointed out that the faculty at liberal arts institutions report working more hours on average (50.1 hours) than faculty at either doctoral (48.7 hours) or comprehensive (47.7 hours) institutions.

The descriptive statistics in Table 1 indicate that the "typical" faculty member at all four types of institutions is a white, married, fifty-year-old male full professor in the social sciences. It is noteworthy, however, that premier institutions have twice as many minority faculty as liberal arts colleges. Faculty at premier universities appear more likely to have a degree from a top-ten institution in their field, to be foreign born, and to have a grant, and they are less likely to be planning retirement in the next five years or to be dissatisfied with their job.(7) Salary is also positively related to research orientation, suggesting that faculty at premier institutions have greater opportunity costs of retiring. In the empirical model, salary is assumed to be recursively determined by innate skills and previous time allocation decisions. Under this hypothesis, including salary in the model yields consistent estimates as long as the equation error is not autocorrelated. Although we are not able to examine the issue of autocorrelation due to a single [TABULAR DATA FOR TABLE 1 OMITTED] observations of each individual, no significant marginal effects change sign in specifications that exclude log salary. Hence, we include the log of salary in the model and discuss the potential biases if autocorrelation is present.(8)

IV. Empirical Results

A grouped-data, multinomial logit model including the explanatory variables in Table 1 is estimated for each institution type.(9) Likelihood ratio tests that make pairwise comparisons between the separate estimates and those from combining the institution with its closest comparator (for example, PREMIER and DOCTORAL separately versus jointly) indicate significant structural differences in faculty time allocation between institutions. The coefficient estimates are used to derive the marginal probabilities for each explanatory variable. We also decompose total probability differences between institution types into those arising from mean differences in the attributes of the faculty and those arising from differences in the coefficients (namely, institutional incentives).(10)

A. Marginal Probabilities

We use the logit estimates to predict how a discrete change in the characteristics of the "typical" faculty member at a given type of institution (namely, a 50-year-old faculty member with the average LNSALARY and TIME, and characteristics implied by the binary variables equaling 0) affect the time allocation probabilities. Discrete changes are used because partial derivatives do not have an exact interpretation for binary variables, which constitute the majority of our explanatory variables. The attributes of the "average" faculty member form a base vector that is used to calculate the four time allocation probabilities, [P.sub.Ki]. A characteristic is then changed (for example, FEMALE is changed from 0 to 1), and the new probabilities ([Mathematical Expression Omitted]) are calculated. The estimated probability changes ([Mathematical Expression Omitted]) and their asymptotic t-values are presented in Table 2.(11) While there are some clear differences in the marginal probabilities for the four institution types, the time allocation responses to changes in personal, employment, or university attributes tend to exhibit consistent patterns across the institution types.

Among the personal attributes, age has the most pronounced effect. The marginal probabilities for AGE suggest that faculty have fairly consistent life-cycle patterns of time allocation across institution types. Research time and service time of the typical faculty member expand until middle age and then fall, while teaching time changes in the opposite direction.(12) It is interesting to note that leisure time increases significantly as faculty members approach their 60s only at comprehensive universities and liberal arts colleges. This implies that faculty at more teaching-oriented institutions tend to increase time spent outside the university as they get older, while faculty at more research-oriented institutions redistribute time among the work categories over their career but do not reduce total work hours. Thus, older faculty at research-oriented institutions do not value their nonwork time more highly than their younger colleagues. If research time produces relatively more valuable human capital than other activities (or at least more marketable and more highly valued human capital in the academic labor market), these results support Ghez and Becker's (1975) findings that the acquisition of human capital increases both work time and the age of retirement.

The marginal probabilities for gender suggest that the time devoted to teaching and research differs for men and women both at premier and liberal arts institutions. At premier institutions, women faculty spend a greater percentage of their time teaching than men and a smaller percentage of their time in research and service activities.(13) This result may be due, in part, to unobserved gender differences in the choice of disciplines within the humanities, social sciences, and natural sciences, but it is also consistent with prior evidence that women tend to publish less than men (Johnson and Stafford 1974; Over 1982; Hansen, Weisbrod, and Strauss 1978). At liberal arts colleges, female faculty spend significantly less time teaching than their male counterparts. Thus, female faculty's time allocation decisions appear to match their institution's missions less well than those of male faculty. This, in turn, could result in less favorable reappointment, promotion, and tenure reviews for female faculty.

Other personal attributes also affect the time allocation decisions. The marginal probabilities for NOTMAR, while generally significant, indicate that marital status does not have a consistent effect on time allocation across the four institution types. For example, single faculty appear to research less at doctoral and comprehensive universities, but research more at liberal arts colleges. Ethnicity (NONWHT) has a significant effect on faculty time allocation at premier, comprehensive, and liberal arts institutions, but not at doctoral institutions. Specifically, [TABULAR DATA FOR TABLE 2 OMITTED] minority faculty appear to spend more time in activities outside the university and less time either at research, teaching, or service, depending on the institution type. In contrast, foreign-born faculty generally research more and teach less at all institution types. A degree from a top-ten institution also has a small but significant effect on the allocation of time for faculty at premier and liberal arts institutions. At premier institutions, faculty who receive a degree from a top institution spend less time in research activities and more time teaching. This suggests that faculty without prestigious academic credentials must emphasize research at the expense of time devoted to teaching to be hired and remain employed at a premier institution. At liberal arts colleges, the reverse is true: faculty with doctorates from the top schools spend more time on research and less time teaching compared with faculty without the prestigious credentials.

Most employment characteristics have marginal effects on time allocation that are consistent with expectations. With age held constant, assistant professors (ASSIST) generally spend more time on teaching and research and less time in leisure activities than full professors. There is little difference between the time allocations of associate and full professors, except that associates tend to spend less time in leisure activities and more time teaching than fulls. The time allocation differences across ranks seem to be somewhat more pronounced at the more research-oriented institutions (namely, the premier and doctoral institutions). It is noteworthy that at all institutions, faculty who plan to retire in the next five years (RETIRE) allocate more time to leisure and reduce their time spent in research, suggesting that retirement is a continuous process of shifting out of the labor force and into other activities.

As expected, department heads (CHAIR) spend significantly more time in service and less time in teaching, research, and leisure. Distinguished professors (DIST) at premier, doctoral, and comprehensive institutions spend less time teaching, while at liberal arts colleges they spend more time in the classroom. This finding underscores the importance of the teaching mission at liberal arts colleges. A consistent finding across all institution types is that humanities faculty (HUM) spend less time researching and more time teaching than those in the social sciences. Natural science faculty (NATSCI), on the other hand, have patterns that vary by type of institution. For example, at premier institutions natural scientists research more than social scientists, while at comprehensive and liberal arts institutions they research less and teach more. A possible explanation for this latter finding is the paucity of research support and the lack of graduate students at comprehensive and liberal arts schools; consequently, these science faculty are responsible for running their own teaching labs that are time-intensive. As expected, faculty who have grants (GRANT) spend more time in research activities and less time teaching. It is interesting to note that these grant-getting faculty also work more total hours than faculty without grants. This result provides additional support for Myers and Mager's finding of a positive relationship between percent of time devoted to research and total hours worked. It is also consistent with Stephan and Levin's 1992 study that found universities ". . . to reward fundable faculty and to punish those who do not get funding, giving the latter heftier teaching loads and smaller raises" (p. 169).

The expected marginal effects on time allocation of other employment characteristics [for example, percent of career at current job (TIME) and job dissatisfaction (NOTSAT)] are uncertain, a priori. In fact, the empirical results show that more recent hires do not have time allocation patterns that consistently differ from longer-tenure faculty. However, there are interesting and significant differences between satisfied and dissatisfied faculty. At all but liberal arts institutions, dissatisfied faculty spend more time working and less time at leisure activities than those who are satisfied. However, the time allocation among work activities differs across institution types. At premier institutions, dissatisfied faculty are those who spend less time on research; at comprehensive universities, dissatisfied faculty spend more time on research; and at liberal arts institutions dissatisfied faculty spend less time teaching. This result suggests that there is some job mis-match in the placement and objectives of the dissatisfied faculty.(14) At all institutions, dissatisfaction is correlated with a greater percent of time devoted to service activities. Although there is no consensus among researchers on how to interpret job satisfaction responses, Juster (1985) finds that job satisfaction ratings tend to be a proxy for the quality of social interactions, the amount of responsibility, and opportunities for learning on the job.

University attributes, other than general institution type, are broadly measured by PRIVATE and UNION. Faculty at private schools spend more time in leisure activities and less time in service activities across all institution types. This is consistent with the notion that private universities provide more staff support for faculty. Only at premier institutions, however, do private and public schools have different missions with regard to teaching and research. Specifically, we find that faculty at private premier institutions teach significantly less than faculty at public premier schools.(15) Consistent with prior evidence of union effects in other industries (Freeman and Medoff 1984), faculty at unionized institutions (except for liberal arts colleges) have more leisure time than do faculty employed at nonunion institutions. Further, unionization is associated with reduced service time, but has no consistent effect on teaching and research time.

Finally, the marginal probabilities provide evidence that compensation is positively associated with time devoted to research and service, and negatively related to time devoted to teaching. However, the marginal effect of salary on faculty time allocation, while significant, is relatively small. For example, the largest predicted impact on time allocation of a $1,000 increase in the wage, ceteris paribus, is a 20 minute per week reallocation of time away from teaching by faculty at liberal arts institutions. In the context of our theoretical model, this suggests that past decisions, while significantly affecting future decisions, do not greatly condition future time allocations beyond the effect of observed personal and employment attributes.

B. Decomposition of Total Probabilities

Differences in the probability of spending time in a given pursuit between institution types reflect both differences in the attributes of faculty hired by the institutions and the institutional incentives given faculty once hired. For example, faculty who are good teachers are more likely to allocate time to teaching than those who are not, and all faculty are more likely to allocate time to teaching in a teaching-oriented institution. Thus, differences in the total probabilities between institutions can be decomposed into average differences in observed faculty attributes (namely, the means) and the estimated institutional incentives (namely, the coefficients).(16)

We focus on two-way comparisons of time allocation decisions between institution types with similar missions (namely, PREMIER versus DOCTORAL, DOCTORAL versus COMP, and COMP versus L. ARTS).(17) Because individual faculty members are likely to self-select into institutions that suit their preferred time allocations, this insures that faculty who have similar observed attributes also have similar unobserved attributes. Thus, a comparison between "comparable faculty" under different incentive structures provides a natural experiment permitting us to draw conclusions about the likely effects of efforts to modify faculty time allocations (for example, toward teaching).(18)

The total probabilities and their decompositions, along with asymptotic t-values, are presented in Table 3. The results confirm that each institution type has a distinct mission that recognizes the teaching-research tradeoff. All faculty are predicted to spend roughly three-quarters of their work time in teaching and research activities combined, but research time declines significantly and continuously from premier to liberal arts institutions while teaching time increases. Moreover, for each of the institution types, coefficient and mean differences in time allocated to teaching and research work in the same direction, indicating that universities hire faculty whose attributes reinforce their mission.

Time allocated to service does not differ significantly in the three comparisons. However, the decomposition of the totals indicates possible differences between institution types. Specifically, the mean and coefficient differences, when significant, have the opposite signs. This suggests that faculty attributes employed to satisfy the teaching/research mission of the institution are inconsistent with the institution's preferred time allocation to service. For example, the coefficient differences indicate that doctoral institutions prefer relatively more time in service than do comprehensive institutions, but the mean differences in the attributes of their faculty yield relatively less time devoted to service. This apparent tension between an institution's teaching-research mission and its service needs may help explain why the marginal probabilities indicate that faculty who spend more time in service are dissatisfied.

Overall, the coefficients account for most significant differences in faculty time allocation. This suggests that differences in institutional incentives account for much of the behavioral differences across institution types. It is possible that the coefficient differences do not arise from pressures placed on faculty subsequent to their arrival, but are due to the faculty members' attributes observed by the institution prior to being hired but not measured by the data. Nonetheless, given that observed differences in faculty attributes account for such a small portion of the time allocation differences and because we compare institutions with "similar" missions, unobserved differences in faculty attributes are unlikely to account for a majority of the time allocation differences across institutions. Thus, because incentives appear to play a primary role in faculty time allocation, our results suggest that institutions can change their missions. Given that the faculty are relatively suited to their institution's prevailing teaching-research mix, however, the current stock of faculty attributes limits the efficacy of policies promoting change.

V. Concluding Remarks

Universities in general, and public research universities in particular, are likely to face increasing fiscal and social pressures to divert resources from research toward teaching. This study examines the possible consequences of changing institutional missions by determining how faculty allocate time. While fundamental to the output mix of a university, this issue has not been considered in previous research on faculty outputs (for example, publication or salary studies) or in the general time allocation literature. Although colleges and universities can try to change their missions, our results suggest that institutions must take into account at least two factors: the career time allocation profiles of their faculty, [TABULAR DATA FOR TABLE 3 OMITTED] and the fact that faculty attributes reinforce the institution's mission under which they were hired or acquired.

Regarding the first point, the marginal probabilities from a grouped-data, multinomial logit model indicate that faculty spend more time teaching and less time researching as they age. While this career time allocation profile is robust across types of institutions, the degree of substitution, the level of dissatisfaction, and the likelihood of retiring all decline with research orientation. In addition, the number of hours a faculty member works is relatively constant in research-oriented institutions over the life cycle, but exhibits the quadratic pattern found in studies of broader populations for faculty at teaching-oriented institutions. These findings suggest that, particularly for young faculty, increasing the current teaching load may not insure that more time is spent teaching over a career and may reduce the overall time commitment of faculty to the university. To the extent that research activities augment an individual's human capital, these investments of time in research endeavors tend to increase a faculty member's effective work life.

This paper also presents evidence that a number of other personal characteristics are significant determinants of faculty time allocation, suggesting that the responses to efforts to change the allocation of faculty time may vary widely across institutions even of a given type. Significant differences exist between public and private institutions, but even holding this factor constant, rank structure, fields of academic specialization, level of grant funding, and the proportion of faculty who are married or foreign-born or who earned degrees from top-ranked universities matter. Gender and ethnicity are also significant factors, and the results suggest that there is less congruence of institutional mission and faculty time allocation for these groups. All of these findings raise important issues not only about reallocating faculty efforts, but also about incentive structures that exist. For example, policies designed to eliminate discrimination may need to take into account the underlying causes of gender and race differences in time allocations.

With respect to the second issue, decompositions of the total probabilities indicate that universities and faculty are matched based on a double coincidence of time allocation preferences. Total time allocation probabilities are differenced between institutions with distinct missions and decomposed into differences attributable to differences in institutional incentives (namely, the coefficient estimates) and differences in mean faculty attributes. We find that most significant differences in faculty time allocations between institutions are due to incentive differences that, for teaching and research time, are reinforced by differences in faculty attributes. Thus, our results suggest that an institution can redefine its mission toward teaching by changing its incentive structure; however, the time allocation response of faculty is likely to be tempered by the faculty member's acquired attributes that reinforce the extent of the institution's mission under which they were hired or have been working. Moreover, resistance is likely to occur until faculty either retool or complete their career.

1. Leisure is used in the traditional sense of time spent in activities outside the labor market (for example, sleep, household production, recreation, etc.).

2. The time units for the empirical model, while predetermined by the data, have two advantages. First, the time interval of a week is sufficiently short that personal characteristics may be assumed to be constant. Second, although utility is updated continuously, the unit of an hour reasonably approximates a continuous time-allocation process. Even so, because the unit of analysis affects the standard errors in grouped-data analysis, some caution must be applied in the interpretation of statistical significance.

3. These classifications are provided by the NSOPF and originate from a Carnegie Commission study (1973). Premier institutions award substantial numbers of doctorates across many disciplines and include those institutions ranked among the 100 leading universities in federal research funds. Doctoral universities offer a full range of baccalaureate programs and Ph.D. degrees in at least three fields, but are less research-oriented and receive fewer federal research grants than premier universities. Comprehensive universities offer Master's degrees as the highest degree conferred. Liberal arts institutions are teaching-oriented and generally do not offer advanced degrees.

4. Some researchers conclude that respondents overestimate time spent in activities explicitly surveyed (that is, work activities). Given the focus on individual and institutional differences, marginal responses are not biased if faculty have an equal tendency to overreport. The tendency to overreport may be correlated with individual and institutional attributes: faculty at premier universities, where research has prominence, may overstate research time. Even so, this reinforcing bias would not mistake the direction of the effect. Thus, the signs of the differences are likely to be unaffected even in this case.

5. The Carnegie Foundation (1973) reports the percent of faculty at research and comprehensive universities responding that it is difficult to receive tenure without publishing was, respectively, 74 and 19 percent in 1969 and 94 and 65 percent in 1989.

6. Juster and Stafford (1991) found that in 1980 American men allocated, on average, 44 hours per week to work activities.

7. The likelihood of retiring is lowest at research-oriented schools even though the mean faculty age is approximately the same across institution types. This supports the Rees and Smith (1991) finding that elimination of mandatory retirement is a problem only at the premier institutions and is reflected in the higher average level of job satisfaction at research schools.

8. The small effect on coefficients when excluding LNSALARY may partially reflect that the small variance in salaries among faculty of the same rank is not a prime motivator for increased effort. Thus, current salary may largely capture unobserved individual attributes.

9. A parsimonious specification is selected over a more complex interactive one because there are no a priori theoretical reasons to prefer one model over another and because the cell size for interactions among discrete variables is generally small. The continuous variables AGE and AGE2 are interacted with HUM, NATSCI, and PRIVATE because empirical studies of academic productivity suggest different life-cycles across disciplines and institution types (for instance, see McDowell 1982). The results are generally insignificant and do not indicate consistent behavioral differences for these groups across institution types.

10. The multinomial logit results are excluded to save space and because they do not have a direct interpretation; the marginal effect of a variable can differ from the sign of its coefficient and each coefficient must be interpreted relative to the base choice. These estimates are available upon request.

11. The precision of the estimated changes depends on the accuracy of the logit model. Let [Mathematical Expression Omitted] be the estimated parameter vector from the logit model, and denote the covariance matrix estimate by [Mathematical Expression Omitted]. The variance of the probability change is then

[Mathematical Expression Omitted].

12. Levin and Stephan (1992) point out that time devoted to research may not only have a higher return early in a faculty member's career, but is relatively less costly than for an established scholar who has opportunities to consult, to give lectures, to run a research center, or to go into administration.

13. This effect is larger when LNSALARY is excluded from the model. The coefficient on a binary variable for gender in four separate wage regressions by institution type indicates that women earn significantly less than men, controlling for personal, employment, and university attributes. Thus, there appears to be a reinforcing (or self-fulfilling) time allocation response to the lower wages paid to female faculty.

14. This hypothesis is supported by those specifications that exclude the log of salary. The marginal probabilities for NOTSAT from these models indicate that dissatisfied faculty do not spend significantly less time in leisure activities at premier and doctoral institutions, but do spend less time researching and more time teaching. Such a time allocation decision is likely to be discouraged at research institutions and to result in lower pay.

15. The marginal probabilities for PRIVATE are somewhat larger and more significant in specifications that exclude LNSALARY. In particular, the results suggest that faculty at private premier institutions not only teach significantly less, but also research significantly more than faculty at public institutions. This reinforcing effect likely reflects the fact that private institutions pay more than their public counterparts, which a log salary equation predicts for each of the institution types.

16. Oaxaca (1973) demonstrates two methods of decomposing gender differences in wages into those arising from differences in means and estimated coefficients using OLS. We follow McMillen and Singell (1994), who develop an Oaxaca-type decomposition of the probabilities and their standard errors using a multinomial logit model in order to examine gender differences in the first job choice of Ph.D. economists.

17. As an example, consider the total probabilities for the average faculty members at premier and doctoral institutions. These probabilities can be expressed as [P.sub.Pi] = [P.sub.i]([X.sub.Pi][[Beta].sub.P]) and [P.sub.Di] = [P.sub.i]([X.sub.D][[Beta].sub.D]), where P(*) again represents the Weibull distribution, [X.sub.p] and [X.sub.D] are means for the vectors of explanatory variables, and [[Beta].sub.P] and [[Beta].sub.D] are the estimated coefficients. [P.sub.Pi] - [P.sub.Di] can then be decomposed into two parts: [P.sub.pi] - [P.sub.i]([X.sub.P][[Beta].sub.D]), which measures differences attributable to the coefficients; and [P.sub.i]([X.sub.P][[Beta].sub.D]) - [P.sub.Di], which measures differences attributable to the means. The precision of the estimated coefficients and mean differences depends on their variances,

[Mathematical Expression Omitted]

and

[Mathematical Expression Omitted].

An alternative decomposition for premier and doctoral institutions can be obtained by adding and subtracting [P.sub.i]([X.sub.D][[Beta].sub.P]) (rather than [P.sub.i]([X.sub.P][[Beta].sub.D]) as above) from [P.sub.Pi] - [P.sub.Di], and the variances calculated by a similar process.

18. The data do not classify liberal arts colleges as public or private. Thus, in the comprehensive and liberal arts comparison, the coefficient on PRIVATE is added to the constant term for the COMP estimates as if this information was unavailable; this insures that the coefficient and covariance matrices have the same dimensions for both institution types.

References

Barbezat, Debora A. 1987. "Salary Differentials by Sex in the Academic Labor Market." Journal of Human Resources 22(4):422-28.

Bunnell, K. 1960. Faculty Workload. Washington: American Council on Education. Carnegie Commission on Higher Education. 1973. "Carnegie Commission on Higher Education." Berkeley.

Creswell, John W., J. L. Chronister, and Martha L. Brown. 1991. "The Characteristics and Utility of National Faculty Surveys." In Using National Data Bases, ed. Charles S. Lenth, 45-60. San Francisco: Jossey-Bass, Inc.

Freeman, Richard, and James Medoff. 1984. What Do Unions Do? New York: Basic Books.

Ghez, Gilbert R., and Gary S. Becker. 1975. The Allocation of Time and Goods over the Life Cycle. New York: Columbia University Press.

Hansen, W. Lee, Burton Weisbrod, and Robert Strauss. 1978. "Modeling the Earnings and Research Productivity of Academic Economists." Journal of Political Economy 86(4):729-41.

Johnson, George E., and Frank Stafford. 1974. "The Earnings and Promotion of Women Faculty." American Economic Review 64(6):888-903.

Juster, F. Thomas. 1985. "Preference for Work and Leisure." In Time, Goods, and Well Being, eds. F. Thomas Juster and Frank P. Stafford, 333-51. Ann Arbor: Institute for Social Research.

Juster, F. Thomas, and Frank Stafford. 1991. "The Allocation of Time: Empirical Findings, Behavioral Models, and Problems of Measurement." Journal of Economic Literature 29(4):471-522.

McDowell, John M. 1982. "Obsolescence of Knowledge and Career Publication Profiles: Some Evidence of Differences among Fields in Costs of Interrupted Careers." American Economic Review 72(4):753-86.

McMillin, Daniel P., and Larry D. Singell, Jr. 1994. "Gender Differences in First Jobs for Economists." Southern Economic Journal 60(3):701-14.

Myers, Betty, and Gerald Mager. 1980. "The Emerging Professorate: A Study of How New Professors Spend Their Time." ED 220(2):429-51.

Oaxaca, Ronald. 1973. "Male-Female Wage Differentials in Urban Labor Markets." International Economic Review 14(3):693-709.

Over, R. 1982. "Research Productivity and Impact of Male and Female Psychologists." American Psychologist 37(1):24-31.

Paul, Chris W., and Paul C. Rubin. 1984. "Teaching and Research: The Human Capital Paradigm." Journal of Economic Education 15(2):142-47.

Ransom, Michael R. 1993. "Seniority and Monopsony in the Academic Labor Market." American Economic Review 83(1):221-33.

Rees, Albert, and Sharon Smith. 1991. Faculty Retirement in the Arts and Sciences. Princeton: Princeton University Press.

Rosen, Sherwin. 1974. "Hedonic Prices and Implicit Markets: Product Differentiation in Pure Competition." Journal of Political Economy 82(1):34-55.

Stephan, Paula, and Sharon Levin. 1992. Striking the Mother Load in Science. New York: Oxford University Press.

Weiss, Y., and L. A. Lillard. 1978. "Experience, Vintage, and Time Effects in the Growth of Earnings: American Scientists, 1960-1970." Journal of Political Economy 86(3):427-47.

Yuker, Harold. 1984. Faculty Workload: Research, Theory, and Interpretation. Washington: ASHE-ERIC Higher Education Research Report Number 10.

Larry D. Singell, Jr., is a professor of economics at the University of Oregon; Jane H. Lillydahl and Larry D. Singell, Sr., are professors of economics at the University of Colorado, Boulder. Anyone interested in obtaining the National Survey of Postsecondary Faculty data should contact the Office of Educational Research and Improvement of the U.S. Department of Education. Permission from that department is required for use of the data.
COPYRIGHT 1996 University of Wisconsin Press
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1996 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Singell, Larry D., Jr.; Lillydahl, Jane H.; Singell, Larry D., Sr.
Publication:Journal of Human Resources
Date:Mar 22, 1996
Words:6512
Previous Article:The minimum wage and the employment of youth: evidence from the NLSY.
Next Article:The utilization of outpatient medical services in Japan.
Topics:

Terms of use | Privacy policy | Copyright © 2021 Farlex, Inc. | Feedback | For webmasters |