DETERMINANTS OF VARIATIONS IN JOURNAL PUBLICATION RATES OF ECONOMISTS.

Davis, Joe C. ; Patterson, Debra Moore

Debra Moore Patterson [*]

Abstract

We track a cohort of economists over a 16-year period following receipt of the Ph.D. Using a multivariate model, we assess the influence on publishing success before and after tenure of graduate school attended, dissertation topic, type of employer, gender, and co-authorship.

I. Introduction

What factors explain why some economists publish more articles in scholarly journals than do others? Degree from or employment in a highly ranked graduate program? Employment in government or in the non-academic private sector? Field of specialization? Gender? Working with co-authors? In this study, we use a multivariate model to measure the effect of each of these potential correlates of scholarly productivity before and after tenure.

Our research builds upon an extensive literature on the scholarly productivity of economists, which consists of over 100 articles. Consider the following examples: publication rates of individuals are related to employer (Conroy, et al., 1995), to employer and field of specialization (Baumann, et al., 1987) to graduate school attended (Laband, 1986), to gender (Medoff and Skov, 1990) and to co-authorship (McDowell and Melvin, 1983). Isolation of the effects of more than one variable, however, requires multivariate analysis such as that used by Broder (1993), who predicted the effects of gender, employer, and school granting the doctorate on the number of articles published in top journals. In comparison to earlier research, our study incorporates a larger range of explanatory variables and journals.

II. Hypotheses and Empirical Results

The data we use to test hypotheses about variation in publication rates come from the following sources. The annual list of doctorates in economics in 1982-83 from US and Canadian institutions ("Eightieth List . . .," 1983) provides the name of the graduate, the school granting the doctorate, and the dissertation field of the 915 degree recipients. From EconLit, the database on CD-Rom containing records on articles listed in the Journal of Economic Literature, we compiled the publication records of the doctorates from 1982 through 1997. Where EconLit is silent on the doctorate's most-recent employment, we have consulted the membership directories of the American Economic Association (AEA) and the Committee on the Status of Women in the Economics Profession (CSWEP). We determined gender of the doctorates by examining first names and by consulting graduate departments and the membership directory of CSWEP. In cases where we were unable to determine gender, we deleted the individual from the sample. We have dat a on all variables used in our study for 642 individuals.

The members of the sample published some 44,425 pages over 16 years in 415 different scholarly journals. The 516 who published averaged 86 pages. Some 126 individuals in the sample did not publish in the journals included in our study, although they may have published in other outlets. The data further reveal a post-tenure decline in publication. Pages published by the sample rose from 1,154 in 1983 to a peak of 4,000 in 1988 and fell to 2,626 in 1996. All page counts are adjusted for co-authorship; for instance, two authors writing a 20-page paper together would get credit for 10 pages each.

Because our sample includes non-academics whose careers may not involve a probationary tenure period and because we use far more journals than other studies, we tested alternative specifications of our model by excluding non-academics and journals other than the top 24 [1]. The results were qualitatively the same with a few exceptions that we note in the text.

Quality of the Author's Graduate Education.

Do the top-rated graduate schools produce economists who are more productive in research? It seems clear that employers believe this hypothesis since graduates of top departments are far more likely to be hired by the top schools, according to research by Barbezat (1992). The best schools presumably equip students with superior research skills and foster an awareness of the frontier topics that are more likely to interest journal editors and referees. Moreover, the top schools are likely to attract the best students. Thus, we hypothesize that success in publishing is positively related to graduating from a highly rated doctoral program.

Table 1 shows that the economists in the sample who graduated from the top tier of departments averaged 47.7 published pages from 1982 to 1989 and 50.2 in the following eight years [2]. The graduates of departments in the middle tier averaged 32.2 and 33.3 pages in the same time periods, while graduates of the bottom tier averaged 18 and 19.2 pages. In order to isolate the effects of graduate preparation from other correlates of publishing success such as employer, dissertation field, gender and co-authorship, the results of multivariate tobit regressions are also shown in table 1. Tobit regressions are appropriate given that our dependent variables have a lower bound of zero [3].

After we account for other variables, the graduates of the top-tier programs published 28.1 more pages in the first eight years and 26.4 more pages in the second eight years than did the bottom tier (the classification excluded from the regressions). The graduates of the middle tier published some 15 more pages in each period. Thus, the contribution of quality of graduate school to publication success appears to hold up over time.

Employment After Graduation.

Doctorate economists entering academia have more to gain from publishing, we hypothesize, than do those going to jobs in business or government. We also expect to find that economists employed as faculty in the higher-ranked departments that grant the Ph.D. will publish more than those in lower ranked departments who will, in turn, publish more than faculty in departments that do not offer the Ph.D. Moreover, we expect lower publication rates for those employed outside of academia who may not have a work environment that facilitates (or encourages) scholarly research. Goodwin and Sauer (1995) found that economists who leave academia for a stint of government service have lower output rates during that time. Also, the research of government economists is more likely to appear in publications that are not indexed. For example, the CIA employs many Ph.D. economists, but the bulk of their research is classified and not widely distributed.

Table 1 relates type of employment to publication success. As expected, the tobit coefficients show that faculty in all ranks of departments offering the Ph.D. publish more than the excluded group of economists employed in departments not granting the Ph.D. in the eight years after graduation [4]. However, after tenure the publication advantage of all three tiers drops sharply to the point that faculty in the bottom tier of Ph.D. programs do not publish significantly more than faculty in the nonPh.D. programs. These surprising results hold even when we limit the dependent variable to pages published in 24 leading journals.

Doctorates in the sample employed in the government did not publish more than faculty in nonPh.D. programs in either of the two time periods; however, those in academic posts in countries other than the U.S. and Canada publish more before tenure but not after than faculty in non-Ph.D. programs in the U.S. and Canada. Economists in private non-academic jobs published significantly less than faculty in non-Ph.D. programs in the first-eight years but not in the second-eight years of their careers.

Research Field.

Does the research field of the dissertation impact future research productivity? Fish and Gibbons (1989) found that authors of dissertations in "general economics" were more likely to publish. Therefore, we expected to find that generalists who wrote dissertations classified as "general economics" or "quantitative methods" would publish more, given their flexibility in choosing research topics. Table 1 shows that these generalists were no more likely to publish than those who wrote more specialized dissertations during the first eight years, but those with dissertations in "quantitative methods" did publish significantly more the second eight years. When we confine the analysis to publication in the top-24 journals only, writing a general or quantitative dissertation increases publication rates in the first-eight years also, and the coefficients becomes more strongly significant in the second-eight years. These results support arguments about the greater research efficiency over their careers of generalists compared to economists with more specialized education.

Gender.

Many studies of the labor market find that women face a competitive disadvantage due to discontinuous work experience because of family responsibilities. Previous studies of gender differences in publication rates led us to expect to find that women publish fewer pages. For example, while Broder (1993) did not find statistically significant evidence that female assistant professors fall behind in publication, she did find that women produce fewer articles in highly rated journals, regardless of their individual characteristics. Fish and Gibbons (1989) found a higher percentage of male graduates publish than do female graduates. Barbezat (1992) also found that women are somewhat less likely than men to publish in scholarly journals before finishing graduate school. The literature also indicates that women are less likely to co-author articles, and may have lower productivity as a result (McDowell and Smith, 1992).

Table 1 shows that while women publish fewer pages on average than male economists in the periods before and after tenure, the difference is not statistically significant at the 5-percent level when we account for other variables. Moreover, we did not find support for Broder's (1993) finding that women publish less in the leading journals. Like McDowell and Smith (1992), we find that women are less likely to co-author articles than men, but the difference is so small that the female coefficient in the regressions remains statistically insignificant even when we do not control for co-authorship.

Co-authorship.

There may be productivity gains from joint research and writing, especially when the authors bring specialized skills to the project--such as proficiency in econometric techniques or knowledge of literature related to the current project--which are not shared by other co-authors. McDowell and Melvin (1983) advance such a division-of-labor hypothesis. Barnett, Ault, and Kaserman (1988) find empirical support for this hypothesis and for two others. Since all academic economists face pressures to publish, their opportunity cost of providing pre-submission review of manuscripts has risen to the point that they may often be rewarded by co-authorship instead of just an acknowledgment. Their second additional hypothesis holds that co-authorship represents an attempt by authors to submit more manuscripts in the hope of getting more of them past the referral process, where acceptance may contain a random element.

Following this literature on co-authorship, we hypothesize that co-authors will publish more pages per author than authors who write alone. Although this hypothesis underlies previous attempts to explain co-authorship, as far as we are aware, it has not been tested to date. In past studies, co-authorship has been treated as the dependent variable, whereas in our study it is an independent variable explaining differences in publication rates among individuals. In considering the effects of co-authorship, we use the mean number of authors per paper for each doctorate. The variable ranges from 1 (no co-authors) to 3 (2 co-authors). For instance, if Jane Economist has published two articles, one solo and the other with a co-author, the value of the co-authorship observation for Jane is 1.5. To explore the possibility of diminishing returns to co-authors, we also include the square of the co-authorship variable in the regressions. If we take the derivatives of our fitted equations with respect to co-authorship an d set them equal to zero, the page maximizing number of authors per article is 1.9 in 1982-89 and 2.1 in 1990-97. [5] This increase in the optimal number of authors per paper may represent authors finding more productive co-author matches with experience.

III. Summary and Conclusions

We have used multivariate analysis to investigate the correlates of variations in publication rates among doctorate economists. The following results stand out: Graduates of top- and middle-ranked departments publish more than the graduates of the lowest-ranked departments in the eight years before tenure and in the eight years after. Economists employed in Ph.D. programs are the leading publishers; however, the publication rates of faculty in the bottom tier are not significantly different from those of faculty in non-Ph.D. departments after tenure. In the post-tenure period, dissertation field gains power in explaining publishing success. The increasing relative publishing success of economists who wrote in general economics and quantitative methods later in their careers indicates an advantage of general over specialized training. Contrary to other studies, we find that the publishing records of women, following control for other publication correlates, do not differ significantly from those of men in eit her time period. Finally, our study finds that co-authorship increases publication rates, but that there are diminishing returns to coauthorship.

(*.) We wish to thank Barry Hirsch, John Huston and an anonymous referee for their helpful comments. Department of Economics, Trinity University, 715 Stadium Drive, San Antonio, TX 78212-7200, 210-999-7221, jdavis@trinity.edu, dpatters@trinity.edu

Notes

(1.) We use the top-24 journals identified by Graves, et al. (1982)

(2.) The sample is separated into three approximately equal tiers according to the graduate program rankings of Hirsch, et al., (1984), which are based on faculty publication records in the early 1980s when our sample graduated. The top-tier departments are Chicago, Harvard, Stanford, Pennsylvania, Yale, Northwestern, MIT, Wisconsin, Berkeley, UCLA. The middle-tier departments are Cornell, Columbia, Princeton, Minnesota, Michigan, Rochester, Illinois, University of British Columbia, NYU, Carnegie Mellon, Washington, Ohio State, North Carolina, W. Ontario, UCSD, Purdue, Virginia, Toronto, Texas A&M, Rutgers, VPI, UCSB, Penn State, Michigan State, Florida, USC, UC-Davis, and Maryland. The third tier includes graduates of 57 other programs.

(3.) We also estimated the equations with OLS. As would be expected, the coefficients on the explanatory variables are smaller with OLS, which does not correct for the limited dependent variable. The statistical significance of the explanatory variables is the same as with tobit.

(4.) Faculty employed in Ph.D. granting departments are separated into three tiers. Those in the top tier work in the 22 highest-ranked departments, using the Hirsch, et al. (1984) rankings. Faculty in the middle tier work in departments ranked 23 to 60, while those in the bottom tier are employed in 32 other departments.

(5.) In order to keep non-publishers in the sample, we coded them with a 1, the same as solo authors. We find a similar but weaker relationship between co-authorship and pages published when we exclude non-publishers from the regressions.

References

Barbezat, Debra A. "The Market for New Ph.D. Economists," Journal of Economic Education, Summer 1992, 23(3), pp. 262-76.

Barnett, Andy H., Ault, Richard W., and Kaserman, David L. "The Rising Incidence of CoAuthorship in Economics: Further Evidence," Review of Economics and Statistics, Aug. 1988, 70(3), pp. 539-43.

Baumann, Michael G., Werden, Gregory J., and Williams, Michael A. "Rankings of Economics Departments by Field," American Economist, Spring 1987, 31(1), pp. 56-61.

Broder, Ivy E. "Professional Achievements and Gender Differences among Academic Economists," Economic Inquiry, Jan. 1993, 31 (1), pp. 116-27.

Conroy, Michael, Dusansky, Richard, Drukker, David, and Kildegaard, Arne. "The Productivity of Economics Departments in the U.S.: Publications in the Core Journals," Journal of Economic Literature, Dec. 1995, 33(4), pp. 1966-71.

"Eightieth List of Doctoral Dissertations in Political Economy in American Colleges and Universities." American Economic Review, Dec. 1983, 73(5), pp. 1179-1202.

Fish, Mary, and Gibbons, Jean D. "A Comparison of the Publications of Female and Male Economists," Journal of Economic Education, Winter 1989, 20(1), pp. 93-105.

Goodwin, Thomas H., and Sauer, Raymond D. "Life Cycle Productivity in Academic Research: Evidence from Cumulative Publication Histories of Academic Economists," Southern Economic Journal, Jan. 1995, 61(3), PP. 728-743.

Graves, Phillip E., Marchand, James R., and Thompson, Randall. "Economics Departmental Rankings: Research Incentives, Constraints, and Efficiency," American Economic Review, Dec. 1982, 72(5), pp. 1131-41,

Hirsch, Barry T., Austin, Randall, Brooks, John, and Moore, J. Bradley. "Economics Departmental Rankings: Comment," American Economic Review, Sept. 1984, 74(4), pp. 822-26.

Laband, David N. "A Ranking of the Top U.S. Economics Departments by Research Productivity of Graduates," Journal of Economic Education, Winter 1986, 17(1), pp. 70-76.

McDowell, John M., and Melvin, Michael "The Determinants of Co-Authorship: An Analysis of the Economics Literature," Review of Economics and Statistics, Feb. 1983, 65(1), pp. 155-60.

McDowell, John M., and Smith, Janet Kiholm. "The Effect of Gender Sorting on Propensity to Coauthor: Implications for Academic Promotion," Economic Inquiry, Jan. 1992, 30(1), pp. 68-82.

Medoff, Marshall H., and Skov, I. Lee. "Ratings of Women Economists by Citations," Journal of Economics and Business, Feb. 1990, 42(1), pp. 81-88.

 Pages Published in Scholarly Journals
 By the Ph.D. Class of 1982-83
 1982-89
 Pages
Category N (mean)
Source of Ph.D.
 Higher-ranked Ph.D. program 221 47.7
 Middle-ranked Ph.D. program 217 32.2
 Lower-ranked Ph.D. program 204 18.0
Employer
Higher-ranked Ph.D. program 74 74.8
Middle-ranked Ph.D. program 71 53.2
Lower-ranked Ph.D. program 85 43.5
Foreign academic 46 32.0
Government 114 24.4
Private non-academic 82 14.5
Non-Ph.D. program 170 16.1
Dissertation field
 General economics 97 43.6
 Quantitative economic methods and data 42 44.6
 Other fields 503 30.0
Gender
 Female 98 27.4
 Male 544 34.0
Co-authors
 Average number of authors per article
 Average number of authors per article squared
Constant
Squared correlation between observed and
 expected values
All 642 33.0
 1990-97
 Tobit
 Regression Pages
Category Coefficient (mean)
Source of Ph.D.
 Higher-ranked Ph.D. program 28.1 50.2
 (5.1) [**]
 Middle-ranked Ph.D. program 15.2 33.3
 (2.8) [**]
 Lower-ranked Ph.D. program -- 19.2
Employer
Higher-ranked Ph.D. program 53.6 76.5
 (7.0) [**]
Middle-ranked Ph.D. program 33.3 53.5
 (4.4) [**]
Lower-ranked Ph.D. program 28.5 43.2
 (4.1) [**]
Foreign academic 17.9 27.7
 (2.1) [*]
Government 12.1 27.7
 (1.8)
Private non-academic -15.8 12.1
 (2.0) [*]
Non-Ph.D. program -- 21.6
Dissertation field
 General economics 10.8 48.6
 (1.8)
 Quantitative economic methods and data 5.5 63.5
 (.7)
 Other fields -- 29.5
Gender
 Female -4.5 26.6
 (.8)
 Male -- 36.1
Co-authors
 Average number of authors per article 206.0
 (9.3) [**]
 Average number of authors per article squared -53.5
 (8.5) [**]
Constant -182.7
 (10.1) [**]
Squared correlation between observed and .269
 expected values
All 34.6
 Tobit
 Regression
Category Coefficient
Source of Ph.D.
 Higher-ranked Ph.D. program 26.4
 (3.7) [**]
 Middle-ranked Ph.D. program 15.1
 (2.2) [*]
 Lower-ranked Ph.D. program --
Employer
Higher-ranked Ph.D. program 33.4
 (3.4) [**]
Middle-ranked Ph.D. program 21.3
 (2.2) [**]
Lower-ranked Ph.D. program 9.5
 (1.1)
Foreign academic 10.3
 (.9)
Government 4.2
 (.5)
Private non-academic -16.6
 (1.6)
Non-Ph.D. program --
Dissertation field
 General economics 12.5
 (1.7)
 Quantitative economic methods and data 29.2
 (2.7) [**]
 Other fields --
Gender
 Female -8.5
 (1.1)
 Male --
Co-authors
 Average number of authors per article 310.8
 (12.4) [**]
 Average number of authors per article squared -72.8
 (10.9) [**]
Constant -292.6
 (13.4) [**]
Squared correlation between observed and .281
 expected values
All
Note: Absolute t-statistics shown in parentheses.
(*.) Denotes significance at 5 percent, (**.) at 1 percent.