Unraveling the academic productivity of economists: the opportunity costs of teaching and service.

1. Introduction

The old adage that one must "publish or perish" poses interesting challenges to members of the academic community. The profession requires some degree of balance between the lofty goals of research productivity by which many are judged successful or not and the daily requirements of academic employment, namely teaching and service. Although many previous studies have examined the degree to which departments or individual academics are successful or productive in their research, few have painted a comprehensive picture of the determinants of research productivity. The current study bridges that gap by investigating the determinants of research productivity and by examining the opportunity costs incurred when academics focus on teaching and service.

This study explores the relations among research productivity, teaching, and service on the basis of individual-specific information obtained for approximately 715 academic economists. Responding to an online survey, these economists provided significant information regarding their teaching and service commitments as well as personal and institutional information. The publication record of each respondent was then obtained from EconLit, completing a rich data set for the systematic study of research productivity.

Results of a Tobit analysis reveal much about the nature of research productivity, underscoring, for instance, the importance of gender, coauthorship, presentations at conferences, and peers who publish. Among the more important findings from this analysis is that both teaching and service commitments significantly affect the research productivity of academic economists. These relations hold across types of academic employer, though to varying degrees. Taken together, the results provide interesting insight into the roles of academic scholars, teachers, and colleagues.

Section 2 of the paper places the current study in the context of the existing literature, whereas section 3 describes the data collection process and the data available for analysis. Section 4 presents the empirical results. The summary is presented in section 5.

2. Context

Given that economists are concerned with production and productivity, it is hardly surprising that they focus so intently on the research productivity of the academic members of the profession. Much of this activity has used various measures of productivity to rank economics departments. Siegfreid (1972) produced one of the earliest departmental rankings followed by Graves, Marchand, and Thompson (1982) and Davis and Papenek (1984). More recently, Conroy and Dusansky (1995), Scott and Mitias (1996), and Dusansky and Vernon (1998) have tackled the subject. Others, including Bodenhorn (1997) and Hartley and Robinson (1997), have focused exclusively on departments at liberal arts institutions. Rankings are variously based on surveys, on citations in journals or texts, and on quality-adjusted publications by departmental faculty.

Individual economists are also rated on the basis of productivity, though often as a byproduct of departmental rankings as in the case of Scott and Mitias (1996) and Bodenhorn (1997). The primary intent of Medoff (1989) and Liner (2001) is to order top economists by scholarly output. Hutchinson and Zivney (1995) take an alternative approach, tracing the publication histories of all economists (not just the "big guys") receiving their Ph.D.s from 1969 through 1988. Davis and Patterson (2000) go a step further and provide regression analysis to determine the impact of source of Ph.D., current employer, dissertation field, gender, and number of coauthors on productivity, again for all economists receiving the Ph.D. from 1982 through 1983.

What is missing from most of these works is a rigorous analysis of the opportunity costs economists face with the use of their time. Maske, Durden, and Gaynor (2003) develop a useful theoretical model recognizing the competing uses of time on the basis of a utility maximization model presented by McDowell and Melvin (1983). The resulting theoretical model is applicable in this case: "Article production is dependent on time devoted to teaching, research, administrative duties ..." and other control variables (Maske, Durden, and Gaynor 2003, p. 557).

The lack of person-specific data has meant that other empirical research has not explicitly examined the impact of teaching and service commitments on research productivity either. Bodenhorn (1997) seeks to explain the productivity of individual faculty at top liberal arts institutions, but lacking individual-specific data on teaching and service, he assumes both are constant. Fox and Milbourne (1999) address this issue for the Australian market, but no comprehensive study exists for economists at American academic institutions. The current research provides a unique opportunity to evaluate the impact of human capital and workplace variables on scholarly productivity of academic economists in the United States through (i) its examination of individual-specific data on teaching load and service commitments and (ii) its control for type of academic employer.

3. The Data

To explain variations in economists' research productivity, person-specific information was collected through a two-step process. First, a web-based survey was used to obtain workplace and human capital information from academic economists. Second, a search of EconLit generated publication data for survey respondents.

Web-Based Survey

A web-based survey was conducted to collect information about economists. An e-mail was sent to 4864 economists at American colleges and universities in February 2003 inviting them to go to a secure web page that posted the survey. (1) A $10 donation to a charity of the respondent's choice was given as an incentive to respond. (2) Questionnaires were completed by 907 economists, yielding a participation rate of roughly 19%. Tabulations of the respondents' rank, gender, and experience indicate that this sample is reflective of economists at U.S. academic institutions. (3)

Of the 907 responses received, several were excluded from the study. Approximately 50 responses did not include adequate information to link the individual to his or her publishing information. Additional observations were dropped if they represented individuals with unusually heavy administrative burdens (e.g., university presidents, directors of research centers). Persons with two years of experience or less were omitted. These individuals have had little time to push research through the pipeline, making it difficult to separate very productive researchers from unproductive ones. Finally, persons classified as instructors were omitted. The majority (about 70%) of instructors did not have a doctorate. Ultimately 714 usable responses were available for analysis.

The survey asked for a variety of information related to economists' human capital and workplace characteristics. Many of the time-varying characteristics were limited to the five-year period from 1998 through 2002. For example, questions about research support and service were couched in terms of the most recent five-year period. Although this time period is somewhat arbitrary, it is a sufficient period to reflect conditions over time without being so long as to make recall difficult and thus reduce the survey response rate. The explanatory variables available from the survey are listed in the Appendix.

Explanatory Data and Hypotheses

Column 1 of Table 1 presents descriptive statistics for the entire respondent pool categorized as personal characteristics, institutional attributes, and teaching and service commitments. The sample is an experienced group with average experience of 19 years. Gains in human capital related to time on the job should generate higher levels of productivity.

In terms of gender, 21% of the sample is female. Although Bodenhorn (1997) and Davis and Patterson (2000) find no gender impact on research productivity, numerous previous studies find that women publish less. Fish and Gibbons (1989), Barbezat (1992), McDowell and Smith (1992), Broder (1993), and Maske, Durden, and Gaynor (2003) all find some evidence that women are less likely to publish than men.

The literature on coauthorship indicates productivity gains from coauthorship, largely based on a division of labor argument. McDowell and Melvin (1983), Barnett, Ault, and Kaserman (1988), Davis and Patterson (2000), and Maske, Durden, and Gaynor (2003) all find positive returns to coauthorship. Individuals in the current sample have an average of 0.57 coauthors. This suggests that the average scholar coauthors some works and works alone on some projects.

The number of years required to complete the Ph.D. and the number of manuscripts under review before acceptance of the first job may be predictors of future productivity. In this sample, the respondents took an average of 5.8 years to complete the doctoral program. A large portion of respondents, 45%, had no publications under review before completion of the Ph.D. Submissions early in one's career should be a positive predictor of future publication success, whereas years to Ph.D. is expected to be inversely related.

Finally, many economists choose to present their work at professional conferences. These presentations are assumed to positively affect publication as they provide a valuable mechanism for presubmission feedback and impose deadlines for completion of manuscripts. Individuals in this sample averaged roughly two presentations per year.

With respect to employment, fully 40% are employed in departments that offer the terminal degree in economics, 18% in departments whose highest degree is the master's, and 42% at institutions who offer only the baccalaureate degree. With the resources available to and the research expectations of faculty at doctoral-granting departments, one would expect faculty at research universities to publish more than at undergraduate departments. Similar logic applies to faculty at master's-granting institutions.

The departmental research climate is expected to influence individual publication rates as well. This effect is proxied by two variables, the first of which is the granting of summer stipends to support research. Such grants reduce the need to teach in the summer for supplemental income and are an indication of departmental research support.

Another indicator of a department's research environment is the number of peers who publish. On average, respondents indicated that 73% of their departmental peers were active scholars. The presence of scholarly active peers is expected to increase productivity as it enhances both formal and informal collaboration and may produce a competitive environment of "keeping up" with one's colleagues.

Although there may be some synergies between teaching and scholarship, the two also compete for the individual's scarce time. Thus research productivity and hours spent teaching are expected to be negatively related. By the same rationale, summer teaching reduces energies devoted to scholarship and thus is expected to reduce productivity. For many, summer is viewed as a block of time that can be dedicated exclusively to scholarship.

Although there may arguably be some synergies between teaching and scholarship, one is hard pressed to find any way in which service to the department or institution could be anything other than a drain on research efforts. Indeed, the more time-consuming the service commitment, the greater the expected decrease in research output.

Although the aggregate data in column 1 is descriptive, it may mask differences across type of academic departments. Columns 2, 3, and 4 give data disaggregated by highest degree in economics awarded by the respondent's department. The three groups have fairly similar experience and gender profiles. Individuals at Ph.D.-granting institutions on average coauthor more and make more presentations at professional meetings than their peers at institutions that do not offer the terminal degree. In addition, they appear to have been faster out of the starting block, taking less time to complete the Ph.D. and having more work under review at peer-reviewed journals before accepting their first job. Economists at Ph.D.-granting departments teach less, both during the regular academic session and in summer school, though their service commitments appear to be similar. (4)

Measure of Research Productivity

Refereed journal articles are commonly accepted as the primary outlet for scholarly research in the field. Accordingly, information regarding peer-reviewed journal articles (including notes, but excluding comments and replies) for each respondent from 1998 through 2002 was gathered from the EconLit database.(5) In total, the individuals in this sample published 2010 articles in 378 different journals during this time period. Of these, 212 articles were published in the top 10 economics journals, (6) with more articles (73) appearing in The American Economic Review (AER) than any other journal. The number of journal article pages was used as the baseline to create an index of each respondent's research productivity.

Baseline article pages were adjusted for (i) coauthorship, (ii) AER page equivalency, (iii) quality, and (iv) consistency over time. To adjust for coauthorship, total pages per article were divided by A, where A is the number of authors. For example, a person with a 10-page article and one coauthor was given credit for five of those pages. Following methods used by Graves, Marchand, and Thompson (1982), Conroy and Dusansky (1995), Scott and Mitias (1996), and Dusansky and Vernon (1998), the second adjustment converted each article's pages to the number of AER equivalent pages. (7) This adjustment, resulting in "AER-adjusted pages" per article, accounts for variation in page size and font across journals.

In addition, each article's pages were also adjusted for quality on the basis of the quality of the journal in which they were published. Accounting for quality is both essential and slippery, as there is no commonly acknowledged scale of scholarly excellence. Fox and Milbourne (1999), for example, group journals into four categories and use a simple 4, 3, 2, 1 quality adjustment factor. Such a method in this application is not optimal, however, it assumes that all journals in the "top" group are quality equivalent and that an article in the top group is four times more important than an article in the lowest group, a factor that may or may not be appropriate. Because it provides a more refined ranking system, this study uses the quality weighting derived from Laband and Piette's ranking of 130 journals (1994, Table A2, Rankings Based on Impact-Adjusted Citations Per Character), calculated from articles indexed in the Social Science Citation Index. One adjustment was necessary to Laband and Piette's ratings because that system includes several 0 values. These 0 values essentially would render publications in these journals as equivalent to no publication at all. In addition, there are more than 300 journals in the data set that are not included in Laband and Piette's study and that would implicitly have received a 0 value, also rendering them equivalent to no publication. To recognize the publication value of these articles while conservatively weighting the quality of these journals, a value of 0.05 was assigned to these data points. On the basis of the arguments of Bodenhorn (1997), the quality weighting used in this study was the square root of the modified Laband and Piette adjustment factors. (8)

Finally, the sum of each individual's AER, quality, and coauthor-adjusted pages were averaged over the number of active years during the five-year period under review. Thus, a person working each year from 1998 through 2002 was considered to be active for five years. Someone working from 1999 through 2002 was considered to have four active years, and someone working from 2000 through 2002 has three active years. The final result is average AER, quality, coauthor adjusted pages per year from 1998 through 2002.

The productivity index (PI) is thus calculated as

P[I.sub.i] = {[n.summation over j=1][([p.sub.i,j][[alpha].sub.j][q.sub.j]/[A.sub.i,j])]}/[y.sub.i]

where p is the number of pages in each article, a is the weighting for AER equivalent pages, q is the square root of impact-adjusted citations per character, A is the number of authors per article, y is the number of active years, i is individual, and j is journal.

Figure 1 presents the distribution of the productivity index. The average annual productivity for the respondents in this sample is 7.91 AER, quality, and coauthored-adjusted pages with a standard error of 18.53. The figure clearly shows, however, that more than one-third of the sample published nothing over the five-year period under review. Another one-third produced more than zero but less than five pages per year, and only 10% of the sample published between five and 10 adjusted pages per year. (9) Corresponding productivity index averages (with standard deviations in parentheses) for the subgroups are as follows: doctoral granting 15.52 (25.97), master's 2.77 (6.44), and baccalaureate 2.77 (7.8). Excluding those in the sample who published nothing, the average annual productivity for those academics who published in the last five years is 12.83 adjusted pages.

[FIGURE 1 OMITTED]

4. Empirical Analysis

To isolate the impact of institution and person-specific factors on research productivity, multivariate statistical techniques are used. The dependent variable in this analysis (the index of research productivity) has a great number of observations equal to 0. To bit estimation is the appropriate technique for analyzing dependent variables that cannot take values below a particular limit (Tobin 1958) and is the technique used in this analysis. Ordinary least squares (OLS) produces biased coefficients if applied to a censored dependent variable. (10)

The Impact of Personal and Institutional Characteristics

Estimated coefficients for the entire sample are reported in column 1 of Table 2 with absolute statistics in parentheses below the coefficients. It is informative to examine the coefficients in Table 2 for sign and significance. All coefficients are statistically significant except the master's dummy. In addition, all are of the expected sign with the exception of experience, which is significantly negative. This is contrary to previous work (Bodenhom 1997; Maske, Durden, and Gaynor 2003) but is perhaps not surprising given the high average experience of the sample. (11) The coefficient for the master's dummy is insignificant, but an F-test indicates that master's and doctoral are jointly significant.

It is difficult to find an intuitive meaning for the Tobit coefficients in Table 2 because the dependent variable is measured as an index and because the Tobit coefficients capture two distinct effects on the dependent variable. A single Tobit coefficient in this research setting incorporates (i) the impact of the independent variable on research productivity for individuals who have published and (ii) the impact of the independent variable on the probability of publishing for individuals who have no publications. McDonald and Moffitt (1980) note that a common mistake made when interpreting Tobit coefficients is to treat them as effects of the independent variables on the dependent variable for cases that are not censored (i.e., above the 0 limit).

To resolve the problems with interpreting Tobit coefficients, the percentage change in expected research productivity after shocking each independent variable is given in column 1 of Table 3. Consider the value calculated for years to Ph.D. (-4.19) as an example for interpretation. First, the expected value of research productivity was calculated at the mean for all independent variables. This expected value is defined as the product of the expected value of research productivity conditional upon being above the limit and the probability of being above the limit. (12) Next, the expected value of research productivity was calculated when increasing the time to receive a Ph.D. by one year (but leaving all other independent variables at their mean value). This increase in years to Ph.D. results in a decrease in expected average annual research productivity of 4.19%. Thus, the values given in Table 3 show the percentage change in the expected value of research productivity given the noted change in the corresponding independent variable. Standard errors calculated by bootstrapping are reported in parentheses.

The first panel of Tables 2 and 3 provides information for the personal characteristics of respondents. The first and perhaps one of the most interesting findings relates to gender. Consistent with the earlier findings of Fish and Gibbons (1989), Barbezat (1992), McDowell and Smith (1992), Broder (1993), and Maske, Durden, and Gaynor (2003), men publish significantly more than women, 41% more.

The coefficient estimates for the sample suggest that working with coauthors pays off in terms of significantly higher productivity though there is a diminishing impact. Increasing the number of coauthors by one during the time frame under review increases average annual research productivity by 22.5%.

Other personal attributes involving how aggressively one pursues one's scholarly career can also influence research productivity. Individuals who have manuscripts under review before accepting the first job have a significantly higher productivity index. The head start of an additional manuscript under review before accepting the first job increases average annual productivity by 10% for these economists. In addition, those in the habit of making presentations at professional conferences will see their productivity increase. Whether due to the external deadlines imposed by conferences or the feedback provided by conference participants, one additional paper presentation per year increases the productivity index by 16%.

Results shown in Tables 2 and 3 indicate that academic experience plays an interesting role in determining research productivity in this sample. A one-year increase in experience decreases productivity by 1.9%.

The departmental research gestalt influences the individual's scholarly activity. Referring to the second panel of Tables 2 and 3, economists in doctoral-granting programs publish 47% more relative to their peers at institutions that offer only the undergraduate economics major. Similarly, both summer stipends and the percentage of peers who publish significantly increase the productivity index. One additional summer stipend in the time frame under review increases research productivity by 11%. A 10% increase in the percentage of peers who publish will increase the individual's productivity index by 9.5%.

The Opportunity Costs of Teaching and Service

As anticipated, there is a significant opportunity cost of teaching and service commitments. Time devoted to teaching during both the academic year and the summer session significantly reduces research productivity. Teaching an additional 3-credit-hour class during the regular academic year reduces research productivity by 9.6%. An additional 3-hour class in the summer has an even greater impact, decreasing productivity by 17.7%.

The other primary workplace competitor for time is service to the department or institution, and all forms of service have a significantly negative impact on productivity. One additional committee assignment per year results in a 7.03% decrease in productivity. Those who add one committee chairmanship per year can expect to see their research productivity decrease by 16.9%. Finally, one additional year of service on average as department chair or program director will reduce the productivity index by 42%.

It is possible that teaching loads and service commitments are themselves endogenous. For example, prospective faculty members who wish to become accomplished scholars when seeking their first job are likely to seek positions where teaching and service loads are low. Bodenhom (1997) notes the possibility of the endogeneity of several variables but does not test for their impact. To address this issue, the sample was divided into three subgroups on the basis of the highest academic degree offered in economics by the individual's department: doctorate, master's, or undergraduate. Columns 2, 3, and 4 of Table 2 present the Tobit results of these estimations. (13) The corresponding columns of Table 3 give the percentage change in expected research productivity for these estimations.

The sign and significance of coefficients for academics at doctoral and baccalaureate institutions are largely consistent with the overall estimation results. Many of the coefficients in the master's specification do not achieve statistical significance, likely due to the small sample size. The doctorate and undergraduate sample sizes both approach 300, whereas there are less than half that number for the master's group, 128. (14) For this reason, the discussion that follows excludes the master's results.

Although distinctions in personal and institutional characteristics exist across the samples, for brevity this discussion focuses only on teaching and service. (15) The general conclusions regarding the negative impact of teaching and service on scholarly production drawn from the aggregate sample hold across types of academic employer, though the magnitudes vary.

The impact on individuals teaching in baccalaureate departments is far higher (roughly double) than in doctoral departments. The same additional 3-hour course reduces productivity by 18.5% for scholars at these institutions as compared with a reduction of 9.8% for their colleagues at doctoral granting schools. This relation holds for summer teaching as well; an additional 3-hour class in the summer reduces productivity by 27% for the undergraduate group and by 16.7% for the doctoral sample. The message is apparent; more time spent teaching clearly reduces research productivity.

The impact of service varies by type of academic employer as well. An additional committee assignment does not significantly affect economists in the doctoral group but significantly reduces productivity in the undergraduate group by 17%. Conversely, the role of department chair has no impact in the undergraduate sample but causes a 48% reduction for chairs at doctoral-granting institutions where the role is typically much more time consuming.

5. Summary

The purpose of this study was to empirically investigate the relations among research productivity, human capital, and workplace variables, particularly the opportunity costs associated with teaching and service. Using individual-specific data obtained from 714 academic economists, this study created a uniquely rich data set from which to explore these relations, to examine the personal and institutional factors that affect productivity, and more importantly to understand objectively the trade-offs that must occur when the requirements of teacher and colleague run headlong into the requirements of scholar.

Using a Tobit analysis, the study finds that increasing teaching significantly reduces productivity in the aggregate sample and for economists at both doctoral-granting and baccalaureate departments. Service commitments are shown to negatively affect productivity as well. The largest time competitor, serving as department chair, reduces productivity in the aggregate and results in a larger decrease in productivity in doctoral-granting departments. Service on committees has the largest impact for economists in undergraduate departments. As the daily requirements of teaching and service increase, academics can and should expect to see corresponding decreases in their research productivity. Clearly, these results highlight the delicate balance between research productivity, teaching, and service.

Appendix

Definition of Variables

Personal characteristics

  Experience               The number of years since completion of the
                             Ph.D.
  Gender                   A dummy variable equal to 1 for males and
                             equal to 0 for females
  Years to Ph.D.           The number of years taken to obtain a Ph.D.
  Review                   The number of manuscripts under review at
                             refereed journals before the individual's
                              first academic position
  Presentations            The average number of presentations made at
                             professional conferences per year
  Coauthors                The average number of coauthors per
                             published article

Institutional attributes

  Doctoral                 A dummy variable equal to 1 if the
                             department offers the terminal degree in
                             economics
  Master's                 A dummy variable equal to 1 if the master's
                             degree in economics is the highest
                             degree offered by the department
  Undergraduate            A dummy variable equal to 1 if the
                             baccalaureate in economics is the highest
                             degree offered by the department
  Peers who publish        The percentage of full-time, tenure-track
                             economics faculty at an individual's
                             institution who have published in
                             peer-reviewed journals in the last
                             five years
  Summer stipend           The number of summer stipends received over
                             the last five years

Teaching commitment

  Teaching hours           The number of credit hours typically taught
                             during an academic year
  Summer hours             The number of credit hours typically taught
                             during the summer
Service commitment
  Committee                The average number of committees on which a
                             person serves in a typical year,
                             exclusive of the average number of
                             committees an individual chairs
  Committee chair          The average number of committees a person
                             chairs in a typical year
  Department chair         The number of years an individual has
                             served as department chair or program
                             director in the last 5 years

References

Barbezat, Debra A. 1992. The market for new Ph.D. economists. Journal of Economic Education 23:262-76.

Barnett, Andy H., Richard W. Ault, and David L. Kaserman. 1988. The rising incidence of coauthorship in economics: Further evidence. Review of Economics and Statistics 70:539-43.

Bodenhorn, Howard. 1997. Teachers, and scholars too: Economic scholarship at elite liberal arts colleges. Journal of Economic Education 28:323-536.

Broder, Ivy E. 1993. Professional achievements and gender differences among American economists. Economic Inquiry 31: 116-27.

Conroy, Michael E., and Richard Dusansky. 1995. The productivity of economics departments in the U.S.: Publications in core journals. Journal of Economic Literature 33:1966-71.

Davis, Joe C., and Debra Moore Patterson. 2000. Determinants of variations in journal publication rates of economists. The American Economist 45:86-91.

Davis, Paul, and Gustav Papanek. 1984. Faculty ratings of major economics departments by citations. American Economic Review 74:225-30.

Dusansky, Richard, and Clayton J. Vernon. i998. Rankings of U.S. economics departments. Journal of Economic Perspectives 12:157-70.

Fish, Mary, and Jean D. Gibbons. 1989. A comparison of the publications of female and male economists. Journal of Economic Education 20:93-105.

Fox, Kevin J., and Ross Milbourne. 1999. What determines research output of academic economists? The Economic Record 75:256-67.

Graves, Philip, James Marchand, and Randall Thompson. 1982. Economics departmental rankings: Research incentives, constraints, and efficiency. American Economic Review 72:1131-41.

Hartley, James E., and Michael D. Robinson. 1997. Economics research at national liberal arts colleges: School rankings. Journal of Economic Education 28:337-49.

Hasselback, James R. 2002. 2002 Prentice Hall guide to economics faculty. Upper Saddle River, NJ: Prentice Hall.

Hutchinson, E. Bruce, and Terry L. Zivney. 1995. The publication profile of economists. Journal of Economic Education 26:59-79.

Laband, David N., and Michael J. Piette. 1994. The relative impacts of economics journals: 1970-1990. Journal of Economic Literature 32:640-66.

Liner, Gaines H. 2001. Core authors and rankings in economics. Atlantic Economic Journal 29:459-68.

Maske, Kellie L., Garey C. Durden, and Patricia E. Gaynor. 2003. Determinants of scholarly productivity among male and female economists. Economic Inquiry 41:555-64.

McDonald, John F., and Robert A. Moffitt. 1980. The uses of Tobit analysis. Review of Economics and Statistics 62:318-21.

McDowell, John M., and Michael Melvin. 1983. The determinants of co-authorship: An analysis of the economic literature. Review of Economics and Statistics 65:155-60.

McDowell, John M., and Janet Kiholm Smith. 1992. The effect of gender sorting on propensity to coauthor: Implications for academic promotion. Economic Inquiry 30:68-82.

Medoff, M. 1989. The ranking of economists. Journal of Economic Literature 20:405-15.

Scott, Loren C., and Peter M. Mitias. 1996. Trends in rankings of economics departments in the U.S.: An update. Economic Inquiry 36:378-400.

Siegfried, John J. 1972. The publishing of economics papers and its impact on graduate faculty ratings, 1960-1969. Journal of Economic Literature 10:31-49.

Sinha, Dipendra, and Joseph Macri. 2002. Rankings of Australian economics departments, 1988-2000. Economic Record 78:136-46.

Tobin, James. 1958. Estimation of relationships for limited dependent variables. Econometrica 26:24-36.

(1) E-mail addresses were obtained from Hasselback (2002). Although this guide omits some economists, there is no reason to suspect that certain types of economists (i.e., ones from smaller colleges or larger institutions) are systematically omitted. Approximately 20% of the original E-mails were returned as undeliverable, yielding the population of 4864.

(2) Survey respondents were given the choice of donating funds to one of five charities: the American Cancer Society, the American Society for the Prevention of Cruelty to Animals, Habitat for Humanity, Special Olympics, or UNICEF.

(3) The American Economic Association (AEA) does not tabulate comparable data for academics in the profession. However, a random sample of 385 academic economists providing detailed information in the AEA membership directory (http:// www.lbmchost.com/aea/search.asp) indicates that 84% of academic economists are male and 16% are female. Sixty-one-percent have achieved the rank of full professor, 22% are associates, and 16% are assistant professors. In addition, average years of experience (defined as years since completion of the Ph.D.) taken from the AEA sample is 19.39. These statistics suggest that the sample used in this paper is reflective of the AEA academic population. See the following section for a discussion of the demographics of the current sample.

(4) These data represent the number of service obligations rather than actual time commitments. The size of the department, for example, would obviously influence the time commitment of the departmental chair.

(5) The online EconLit search was conducted in May, 2003. The database covers all journals catalogued by Journal of Economic Literature. It is possible that articles in journals not included in EconLit have been omitted, but short of having a vitae for each respondent, this is the most comprehensive source of publication data available.

(6) The top 10 journals are defined as American Economic Review. Journal of Political Economy, Quarterly Journal of Economics, Review of Economics and Statistics, Review of Economic Studies, Econometrica, Economic Journal, Journal of Monetary, Economics, Journal of Economic Theory, and International Economic Review.

(7) AER page adjustment factors for over 400 journals were kindly provided by Dipendra Sinha. These unpublished factors were used in Sinha and Macri (2002). Remaining journals for which there were no adjustment factors were assigned the average value for the lowest quartile of Laband and Piette's (1994)journals, 0.68.

(8) Bodenhorn (1997) convincingly argues that the wide discrepancies in Laband and Piette's impact-adjusted values are undesirable as they give inordinately large weights to the highest ranked journals. He suggests taking the square root of Laband and Piette's weights. "The ordinal ranking of the journals is maintained when this is done, but cardinal weight differentials are considerably decreased" (Bodenhom, 1997, p. 325).

(9) Several alternative measures of research productivity were calculated; each method differed only in the quality weight used. In the first alternative, Laband and Piette's weights were taken as given from Table A2; the square root of this figure was not taken. The second alternative used quality weights from Laband and Piette's Table Al. The third alternative used the square root of this figure. The fourth alternative involved using an entirely different method to adjust for quality. In the spirit of Fox and Milbourne (1999), a weight of 4 was given to "A" journals, 3 to "B" journals, 2 to "C" journals, and 1 to "D" journals where A, B, C, and D journals were derived from rankings by Scott and Mitias (1996) and Dusansky and Vernon (1998). The resulting estimations using each quality measure were remarkably similar, indicating little sensitivity to the quality weighting used.

(10) The overall equation was also estimated using OLS. The OLS coefficients are smaller using this technique as expected since OLS does not account tot the censored dependent variable. The statistical significance of the variables is the same as in the To bit specification except for the number of committees on which an individual serves. The adjusted [R.sup.2] for the OLS specification is 30%.

(11) Several alternative functional forms were tested for the relationship between experience and productivity. Bodenhorn (1997) and Maske, Durden, and Gaynor (2003) use a quadratic. When used in this research setting, the quadratic also shows a negative relationship between experience and productivity, but the squared terms were insignificant. For this reason, and because experience and experience squared are highly correlated, the squared term was omitted. The findings for teaching and service are not dependent upon the functional form for experience.

(12) The expected value of research productivity is calculated according to McDonald and Moffitt (1980).

(13) Peers who publish and give presentations may also be endogenous. To test for the sensitivity of the model specifications, the model was run omitting these potentially endogenous variables. The sign and significance of teaching and service coefficients did not change, making the model specification more convincing.

(14) The Chow test indicates that these three specifications may be combined.

(15) It is worth noting that the number of coauthors and presentations seems to be much more important for authors at nondoctoral granting schools. This result may reflect endogeneity--productive people do more of both. However, it may also indicate that interactions outside the department are more important for those in smaller departments.

Susan Washburn Taylor, * Blakely Fox Fender, ([dagger]) and Kimberly Gladden Burke ([double dagger])

* Department of Economics, Millsaps College, Jackson, MS 39210, USA; E-mail [e-mail]; corresponding author.

([dagger]) Department of Economics, Millsaps College, Jackson, MS 39210, USA; E-mail [e-mail].

([double dagger]) Department of Accounting, Millsaps College, Jackson, MS 39210, USA; E-mail [e-mail].

We thank Carl Brooking, Dan Hamermesh, Dek Terrell, and two anonymous referees for helpful comments and suggestions.

Received December 2003; accepted November 2004.

Table 1. Means for Respondents' Characteristics by Highest
Degree Offered (Standard Deviations in Parentheses)

                           Overall         Doctorate
                           n = 714         n = 288

                             (1)             (2)

Personal Characteristics
  Gender (male = 1)         0.79 (0.41)     0.81 (0.39)
  Average coauthors         0.57 (0.66)     0.79 (0.66)
  Years to Ph.D.            5.79 (2.34)     5.23 (2.04)
  Presentations             1.98 (2.04)     2.79 (2.63)
  Review                    0.55 (0.94)     0.79 (1.06)
  Experience               19.43 (9.61)    20.07 (10.61)

Institutional
  Characteristics
  Degree: masters           0.18 (0.39)       n/a (a)
  Degree: doctorate         0.40 (0.49)        n/a
  Peers who publish        72.49 (25.53)   82.66 (17.37)
  Summer stipend            1.17 (1.69)     1.72 (2.01)
Teaching Characteristics
  Teaching hours           13.89 (5.89)    11.58 (4.63)
  Summer hours              1.40 (2.55)     0.81 (1.89)
Service Characteristics
  Committee                 2.39 (1.47)     2.38 (1.46)
  Committee chair           0.82 (0.90)     0.87 (0.92)
  Department chair          1.35 (1.79)     1.17 (1.72)

                           Masters         Undergraduate
                           n = 128(3)      n = 298

                              (3)             (4)

Personal Characteristics
  Gender (male = 1)         0.80 (0.40)     0.76 (0.43)
  Average coauthors         0.53 (0.65)     0.37 (0.60)
  Years to Ph.D.            6.00 (2.85)     6.24 (2.26)
  Presentations             1.65 (1.37)     1.35 (1.19)
  Review                    0.49 (1.06)     0.35 (0.69)
  Experience               20.21 (8.49)    18.48 (8.97)

Institutional
  Characteristics
  Degree: masters               n/a             n/a
  Degree: doctorate             n/a             n/a
  Peers who publish        70.23 (26.49)   63.64 (28.14)
  Summer stipend            0.97 (1.45)     0.73 (1.23)
Teaching Characteristics
  Teaching hours           14.88 (6.30)    15.69 (6.28)
  Summer hours              2.16 (3.04)     1.65 (2.77)
Service Characteristics
  Committee                 2.71 (1.63)     2.27 (1.39)
  Committee chair           0.96 (0.98)     0.72 (0.84)
  Department chair          1.47 (1.91)     1.48 (1.80)

(a) n/a, not applicable.

Table 2. Tobit Coefficients by Highest Degree Offered

                              Overall             Doctorate
                              n = 714              n = 288
                                (1)                  (2)

Personal
  Gender                   5.06 ** (2.25) (a)   10.57 ** (2.49)
  Average coauthors       24.33 ** (9.12)       20.20 ** (4.50)
  Average coauthors
    squared               -6.63 ** (6.35)       -5.40 ** (3.59)
  Years to Ph.D.           -0.61 * (1.44)          -0.73 (0.93)
  Review                    1.50 * (1.58)           1.58 (1.03)
  Presentations            2.31 ** (5.25)        2.19 ** (3.56)
  Experience              -0.30 ** (2.90)       -0.38 ** (2.14)
Institutional
  Degree: masters            -2.74 (1.03)           n/a (c)
  Degree: doctorate         6.03 * (1.54)             n/a
  Peers who publish        0.14 ** (3.36)        0.22 ** (2.37)
  Summer stipend (b)       7.76 ** (3.07)        9.19 ** (2.28)
Teaching
  Teaching hours          -0.51 ** (2.87)       -0.83 ** (2.28)
  Summer hours            -0.96 ** (2.35)        -1.43 * (1.61)
Service
  Committee               -1.09 ** (1.80)           -1.31
  Committee chair         -2.75 ** (2.52)       -3.76 ** (2.01)
  Department chair (b)    -7.65 ** (2.83)     -13.9 5 ** (2.80)
Log likelihood                -2036.9              -1132.36

                              Masters           Undergraduate
                              n = 128              n = 298
                                (3)                  (4)

Personal
  Gender                      -0.08 (0.04)             1.76
  Average coauthors        25.49 ** (5.69)      18.38 ** (6.25)
  Average coauthors
    squared               -10.72 ** (4.48)     -5.99 ** (4.32)
  Years to Ph.D.               -0.31              -0.23 (0.63)
  Review                        0.65            1.97 ** (1.92)
  Presentations              0.85 * (1.43)      3.03 ** (4.57)
  Experience               -0.35 ** (3.15)        -0.11 (1.11)
Institutional
  Degree: masters               n/a                  n/a
  Degree: doctorate             n/a                  n/a
  Peers who publish         0.08 ** (1.95)      0.06 ** (1.83)
  Summer stipend (b)           -0.36            5.20 ** (1.86)
Teaching
  Teaching hours               -0.14           -0.43 ** (3.30)
  Summer hours              -0.51 * (1.41)     -0.65 ** (2.01)
Service
  Committee                    0.43 (0.72)     -1.21 ** (2.16)
  Committee chair             -0.16 (0.15)     -1.89 ** (1.71)
  Department chair (b)     -5.14 ** (1.84)        -2.10 (0.89)
Log likelihood                 -237.08              -559.99

(a) Absolute t-statistics shown in parentheses.

(b) These regressors have been averaged over the number of years
of active service.

(c) n/a, not applicable.

* denotes statistical significant at the 10% level, ** at the 5% level.

Table 3. Impact of Institution and Person-Specific Characteristics
on Research Productivity

                     Change in
                    Independent        Overall         Doctorate
                      Variable            (1)             (2)

Personal
  Gender          Male (relative     40.94 (22.24)   58.59 (33.31)
                    to female)
  Average         One additional     22.53 (4.38)     8.40 (4.71)
    coauthors       coauthor in
                    timeframe
                    under review
  Years           One additional     -4.19 (3.01)    -2.89 (3.84)
    to Ph.D.        year to
                    obtain Ph.D.
  Review          One additional     10.33 (7.35)     6.44 (6.92)
                    article under
                    review prior
                    to first
                    academic
                    position
  Presentations   One additional     15.93 (4.63)     9.15 (3.23)
                    presentation
                    per year
  Experience      One additional     -1.87 (0.69)    -1.46 (0.79)
                    year
Institutional
  Degree:         Masters            -17.7 (11.51)    n/a
    masters         (relative to
                    baccalaureate)
  Degree:         Doctorate          47.39 (18.82)    n/a
    doctorate       (relative to
                    baccalaureate)
  Peers who       An additional       9.47 (2.38)     9.21 (3.47)
    publish         10% of peers
                    publish
  Summer          One additional     10.62 (3.85)     3.31 (2.01)
    stipend         stipend in
                    timeframe
                    under review
  Teaching
    Teaching      Three additional   -9.61 (2.56)    -9.88 (4.20)
      hours         hours of
                    teaching
                    during the
                    academic year
    Summer        Three additional   -17.7 (5.26)    -16.7 (7.74)
      hours         hours of
                    teaching
                    during the
                    summer
Service
  Committee       One additional     -7.03 (4.00)    -5.29 (4.82)
                    committee
  Committee       One additional     -16.9 (6.14)    -14.7 (6.69)
    chair           committee
                    chair
  Department      One additional     -42.2 (9.35)    -47.8 (11.11)
    chair           year as
                    department
                    chair

                   Masters         Undergraduate
                       (3)             (4)

Personal
  Gender          -1.82 (32.48)   31.93 (49.19)
  Average         61.35 (44.77)   42.93 (12.03)
    coauthors
  Years           -6.13 (11.34)   -3.42 (4.53)
    to Ph.D.
  Review          13.50 (47.82)   34.15 (29.79)
  Presentations   18.40 (20.96)   35.12 (11.99)
  Experience      -6.75 (2.85)    -1.56 (11.35)
Institutional
  Degree:          n/a             n/a
    masters
  Degree:          n/a             n/a
    doctorate
  Peers who       17.79 (14.64)    8.78 (5.30)
    publish
  Summer           2.45 (30.25)   20.49 (16.58)
    stipend
  Teaching
    Teaching      -7.98 (12.52)   -18.5 (5.99)
      hours
    Summer        -27.6 (21.25)   -26.8 (13.97)
      hours
Service
  Committee        8.59 (16.88)   -17.6 (9.13)
  Committee       -3.07 (23.10)   -25.9 (14.62)
    chair
  Department      -69.9 (25.41)   -28.8 (24.90)
    chair

(a) Standard errors calculated by bootstrapping reported in
parentheses.

* Calculated as the percentage change in Y due to the attribute
(for dummy variables) or due to the specified change in the
independent variable (for quantitative variables).