Determinants of scholarly productivity among male and female economists.

Maske, Kellie L. ; Durden, Garey C. ; Gaynor, Patricia E. 等

I. INTRODUCTION

The purpose of this article is to test a model of the determinants of articles produced by male and female academics. Of particular interest is how article production is influenced by coauthorship, institutional research orientation, and gender. Several works have identified and demonstrated the trend toward coauthorship in economics and in other disciplines. Using data from the Journal of Economic Literature, Heck and Zaleski (1991) demonstrated that the incidence of coauthorship has increased from about 15% of total articles in 1969 to about 35% in 1989. Durden and Perri (1995) report that as of 1992 the proportion coauthored had grown to more than 38%. Hudson (1996) finds that by 1993 coauthorship rates in the Journal of Political Economy and American Economic Review were 39.6% and 54.9%, respectively, as compared to rates of 6% and 8% in 1950.

Why do economists coauthor? Rational interest theory suggests that if scholars collaborate then there must be a utility-enhancing result, other things equal. This could be in the form of higher salaries and increased probabilities of promotion, greater access to funded research, greater mobility in job markets, and the like, which might result if collaboration significantly increases the overall production of published papers. Studies by McDowell and Melvin (1983), Barnett et al. (1988), and Piette and Ross (1992) suggest that coauthorship among scholars may increase article production through the division of labor made necessary by increased complexity in the subject matter and by the need to saturate markets to increase the probability of getting papers accepted for publication. Laband and Tollison (2000) find evidence that increasing rates of coauthorship result from the greater quantitative content of articles, greater requirement for the use of sophisticated econometric techniques, and the fact that cooperation is cheaper in time and other resources as compared to learning what is necessary to publish in another field or discipline. Hamermesh and Oster (2002) discuss nonpecuniary, purely consumption benefits, such as the pleasure obtained from cooperation, and suggest that higher levels of prestige among colleagues may also be relevant.

Although many reasons for cooperation have been developed, not much has been done to determine whether the rate of coauthorship among scholars actually affects total article production. Using cross section data, Hansen et al. (1978) and Graves et al. (1982) estimate productivity models, but coauthorship is not included in either study. McDowell and Smith (1992) include a coauthorship variable but find that it neither increases nor decreases article production when articles are discounted by the number of authors. Hollis (2001) uses panel data to examine the relationship between coauthorship and research productivity. He finds that coauthorship leads to better, longer, and more frequent publications. When publications are discounted by the number of authors, however, coauthorship appears negatively related to research output. McDowell and Smith's (1992) and Hollis's (2001) results are very counterintuitive, because a viable reason for coauthorship is to increase efficiency, as shown in Durden and Perri (1995). Durden and Perri (1995) estimate a time-series model, finding that coauthorship is highly instrumental in determining article productivity.

As far as we know, the only study by economists that specifically analyzes the effect of institutional research orientation on publishing output is that by Graves et al. (1982). Using data from 240 schools, they estimate total pages published per faculty member in the top 24 economics journals between 1974-78. Using a series of independent variables to proxy influences that affect publication, they find that secretaries per faculty member helps and teaching load hinders production. The effects of Ph.D. status, teaching assistance, and faculty--student ratios are mixed, depending on specification of the model.

The existing literature suggests an area in which this research may provide additional insights on male/female publishing patterns. Previous work suggests that there may be differences between males and females in salaries, in the number of articles published, and some evidence that there are gender-specific differences in propensities to coauthor. (1) Controlling for field, graduate institution, and journal rank, Fish and Gibbons (1989) find that males significantly publish more than women. Significant salary differences between men and women in academics are reported in Broder (1993) and Formby et al. (1993). Differences in coauthorship rates were found by both Ferber and Teiman (1980) and McDowell and Smith (1992), and the latter finds that economists tend to coauthor with members of the same sex.

The emphasis in this article is on analyzing the effect (if any) of coauthorship, experience level, gender, institutional orientation, and other factors on how many published works a scholar in economics will have produced at a given point in his or her academic career. The study proceeds as follows. In section II, we develop a model of the determinants of articles produced. In section III, we present an empirical model and describe the data set and variables used. Section IV provides a discussion of the mean values of the independent variables. Results from pooled, male, and female regressions are presented and analyzed. The summary and conclusions of section V complete the paper.

II. A MODEL OF ARTICLES PRODUCED

Several studies over the years have shown that the production of refereed journal articles is positively related to the size of academic salaries and to the probability that a given individual will be promoted to the next available rank. (2) Therefore, we begin with the assumption that a researcher's utility, other things equal, depends on total articles produced so that:

[U.sub.i] = U(A)

where [U.sub.i] is the utility gained by a given individual i, all other things equal, from the total articles (A) produced by that individual at a particular point in time in the individual's career. Following McDowell and Melvin (1983), the total number of articles is given by

A = [delta][S.sub.S] + [delta][gamma][S.sub.C] + [L.sub.S] + [gamma][L.sub.C],

where [gamma] is the value of a co-authored article relative to sole authorship, [delta] is the value of a short article relative to long, [S.sub.S] is time spent on sole authored short articles, [S.sub.C] is time spent on coauthored short articles, [L.sub.S] is time spent on sole authored long articles, and [L.sub.C] is time spent on coauthored long articles. We assume utility is of the form [U.sub.i] = A, or

[U.sub.i] = [delta][S.sub.S] + [delta][S.sub.C] + [L.sub.S] + [gamma][L.sub.C].

Maximization of individual utility is subject to a time constraint. Adapting the McDowell and Melvin (1983) model, we assume an academic economist's efforts are divided among three activities: research, teaching, and administration, other things equal. (3) Total time, T, is given by

T = [T.sup.R] + [T.sup.T] + [T.sup.A],

where [T.sup.R] is the time devoted to research, [T.sup.T] is time devoted to teaching and [T.sup.A] is time spent on service activities. Research time [T.sup.R] is

[T.sup.R] = [S.sub.S.sup.T]([S.sub.S] + [alpha][S.sub.C] + [L.sub.S.sup.T]([L.sub.S] + [alpha][L.sub.C]),

where [alpha] represents reduction in time associated with coauthorship relative to sole authorship, [S.sub.S.sup.T] is total time spent on short articles, and [L.sub.S.sup.T] is total time spent on long articles. The time constraint can be rewritten as

T = [S.sub.S.sup.T] ([S.sub.S] + [alpha][S.sub.C] + [L.sub.S.sup.T]([L.sub.S] + [alpha][L.sub.C]) + [T.sup.T] + [T.sup.A].

The utility maximization results are qualitatively similar to those of McDowell and Melvin. (4) Simplifying, the total number of articles produced is

A = f([T.sup.T], [T.sup.A], [alpha], [gamma], [delta], [S.sub.S.sup.T] / [L.sub.S.sup.T]).

Therefore article production is dependent on time devoted to teaching, research, administrative duties, the efficiency of time used, the value of short versus long articles, and the value of sole authored versus coauthored articles.

III. DATA DESCRIPTION

The data for this study are compiled from a survey of economists, which was conducted over a period of several months in 1992 and 1993. The survey form was sent to male and female academics whose names appeared in the American Economics Association membership volume published in 1989. Every tenth male and all females were surveyed (survey form available on request). If the tenth male encountered in the AEA membership listings was not an academic economist, the next academic male economist encountered was surveyed. Nonrespondents from the first wave were surveyed a second time, and nonrespondents from the second wave were surveyed a third time. From these three mailings (approximately 2,400 individuals surveyed) we received about 1,000 usable forms.

The functional form for the empirical model on which regressions were based is

[A.sub.i] = d([T.sup.T], [T.sup.A], [T.sup.R], [alpha], [gamma], [delta], Control).

The dependent variable, [A.sub.i], is the total number of articles produced by individual i, and is constructed with data from questions that asked how many full-length articles and notes the respondent had published at the time of the survey. The literature provides little or no help in deciding how many notes equal a full-length article. Recognizing that other researchers might have chosen differently, we assume that two notes will be equal to one full-length article ([delta] = 0.5) so that total articles produced is [A.sub.i] = (full-length articles + 0.5 x notes). Based on the results from McDowell and Smith (1992), we assume that authored and coauthored papers generate approximately the same utility, and therefore we do not discount coauthored works relative to sole authored. McDowell and Smith find that single-authored and coauthored articles appear to be weighted equally as determinants of rank and salaries ([gamma] equal to one). Their results suggest that "departments tend to count total articles rather than measure independent contributions of coauthors." (5)

The amount of time devoted to teaching ([T.sup.T]) should be affected by the number of courses taught and the general teaching-research orientation of the respondent's university or college. Each respondent was asked to indicate how many courses were taught per semester or quarter, and these were converted to a number taught per year. We assume that the heavier the teaching load, the less time for research, so that the expected sign on Courses is negative. To control for time devoted to administrative duties ([T.sup.A]), we asked if the respondent had spent time as an administrator and created the variable Administration with value of one if the answer was yes, zero otherwise, with the expected sign again negative.

It has been argued here and elsewhere that scholars coauthor because collaboration provides nonpecuniary and pecuniary benefits, the latter resulting from the increased production of total articles and correspondent higher salaries. To test whether collaboration significantly enhances total article output we create the independent variable PercentCoauthored, where PercentCoauthored = [(coauthored articles)/(total articles)] x 100. If, as expected, collaboration generates production efficiencies, then the sign on PercentCoauthored will be positive.

The amount of time an individual will have available for research and the efficiency and intensity with which research time is used depend on certain personal and institutional factors, such as the number of years one has been in academe, research-related human capital accumulation, desire to become a known scholar, ability to conceive and execute research projects, access to appropriate resources, and institution-specific performance expectations. Because all data used here are from respondents who hold a terminal degree (but recognizing that differences in the quality and quantity of training exist among institutions) we assume that anyone who has a Ph.D. in economics has the basic foundation necessary for conducting research. (6) Research-related human capital differences among scholars at a given point in time, other things equal, will then depend on what one has learned from reading, teaching, interacting with other scholars in various settings, trial and error with respect to the research process, and so on. These factors are directly and intimately related to the years of job experience for each respondent. In addition, the longer one has worked in academe, the more articles one should have produced, all else equal. To approximate both the research-related human capital embodied in each respondent, and differences in years of working life accumulated, we use years of Experience, which should have a positive sign. With respect to most professions and occupations, human capital probably begins to diminish at some point in an individual's working life. If this is true in academics, or if a faculty member simply loses interest or becomes complacent as time goes by, the rate of article production will slow, ceteris paribus. To test for this, we include the variable ExpSq, which is the square of Experience, and expect that the direction of influence will be negative.

It seems likely that personal factors (such as interest in publishing and the type of scholarly aptitudes that lead to greater or lesser success in the publication of articles) will differ among economists and that research performance expectations and resources available for research will vary among institutions. If we can assume that those with greater interest and scholarly aptitude are usually employed in institutions where resources and research performance expectations are greater, then our data allow construction of variables designed to capture the related influences on article production. Each respondent was asked to categorize his or her institutional orientation as (1) primarily teaching, (2) effective teaching and research, or (3) effective teaching and strong research. With primarily teaching as the base, we constructed two dummy variables: Teaching-Research equals one if the respondent selected category two and equals zero otherwise; and HighResearch equals one if the respondent selected category three and equals zero otherwise. The expected sign for both is positive, and the incremental effect should be larger for the latter.

We include two control variables that may affect article production either positively or negatively: Books, which is total books published, and Chapters, the sum of such publications as monographs and chapters in edited collections. (7) Given the number of years in academe, more time spent on Books and/or Chapters could mean less time spent producing articles, and the sign on one or both will be negative. If these variables capture complementary effects, or if they measure differences in ability and drive to publish, the signs will be positive. (8)

On the survey form we listed the top 25 journals (identified as such in Liebowitz and Palmer 1984) and asked the respondent to indicate how many articles and notes he or she had published in each. Using the answers, PercentTop25 was created as PercentTop25 = (Top 25 Articles/ Total articles) x 100. If the time necessary to prepare articles for publication in top 25 journals is greater than that required for other journals, then, as the percentage of total publications in top 25 journals increases, the total number of publications would decrease, and PercentTop25 will have a negative sign. Alternatively, there may be complementary effects, in which case the relationship will be positive.

We include three final control variables including Gender, Race, and Public. The variable Gender equals one if the respondent is male and zero if female is entered in pooled regressions to determine whether the independent variables used can explain all male/female differentials. Fish and Gibbons (1989) found that residual differences do exist, so we assume that such differences will be found in our data and the sign on Gender will be positive. To determine whether race matters, we created Race equal to one, zero otherwise, if the respondent was other than a white U.S. citizen (with no expectation with respect to sign). (9) As another control for institutional orientation, we created the variable, Public equals one, if the respondent's school is public, zero if private, with an expectation that the sign will be positive. The reason is that many private schools are small, liberal arts schools with less time and resources available for research, so publication rates at public schools should be higher.

IV. DISCUSSION OF MEAN VALUES AND REGRESSIONS RESULTS

Table 1 presents the mean values of variables used for all regressions. The number of articles produced by the average respondent is 11.6, which is somewhat lower but consistent with Grimes and Register (1997). Using a sample of Ph.D. recipients for 1968, Grimes and Register find that from 1969-93, average article production was 13.8. (10) When sorted by gender, our data suggest that males produce about 13.9 and females 7 total articles. (11) Males in the sample publish 1 book, on average, whereas females publish 0.7. One reason for these disparities is that as the regression results later show, experience matters, and females have 12.2 years experience whereas males have 17.2 years. Another is that 6% of females report working in a high-level research institution, compared to 11% for males.

Males and females in the sample are shown to coauthor at almost exactly the same rate. This is consistent with the data reported by McDowell and Smith (1992), in which the percent of articles coauthored is not significantly different for women than men. However, McDowell and Smith (1992) also collect data on the gender of coauthor teams and find that team selection is not gender-neutral. They argue that women are at a disadvantage in terms of total article production because coauthorship tends to occur within gender and the male-to-female ratio in academic economics is at least 8: 1. Our data set does not allow us to examine gender-specific coauthorship.

In the regressions to follow, we use least-squares regression to estimate three versions of the following equation, one for pooled data and one each for males and females. (12)

A = [[beta].sub.o] + [[beta].sub.1] Gender + [[beta].sub.2] Race + [[beta].sub.3] Experience + [[beta].sub.4] ExpSq + [[beta].sub.5] Books + [[beta].sub.6] Chapters + [[beta].sub.7] Courses + [[beta.sub.8] Teaching Research + [[beta].sub.9] High Research + [[beta].sub.10] Administration + [[beta].sub.11] Public + [[beta.sub.12] PercentCoauthored + [[beta].sub.13] PercentTop25 + [epsilon]

Results are presented in Table 2 and show that the explanatory power of the model is relatively high, given that the data are cross-sectional, with adjusted [R.sup.2] values ranging from 0.42 to 0.49, and F-values always highly significant. (13) In the pooled regressions the significant variables include Gender, Experience, ExpSq, Books, Chapters, Courses, Teaching-Research, HighResearch, Administration, Public, and Percent Coauthored. (14) In every case the expected and actual direction of influence is the same. A respondent working in a school emphasizing effective teaching and research produced almost 4.8 more articles than those working in a school emphasizing primarily teaching. The males equation results show the corresponding differential to be 5.6 papers, whereas that for females is 3.9 (in percentage terms these are roughly similar). The variable Courses is negative in all three regressions and significant in the pooled and males regressions, a result consistent with the expected trade-off between teaching and research. Administration is negative in all three equations and significant in the pooled and males regressions. Graves et al. (1982) show that time available for research is an important factor, and these results provide strong evidence that this is true.

The variable PercentCoauthored is a significant factor in explaining variation in articles produced. This supports Durden and Perri (1995), who find that coauthorship enhances production efficiency. The pooled results show that when the coauthored index rises by 1 point, overall production rises about 0.05. This means that if the ratio of coauthored to single-authored papers increases by 20 percentage points, then total output increases by about one article. The coefficient estimate for males is 0.06 and for females 0.04. This suggests that a 20% increase in the rate of coauthorship will, on average, increase output for men by about 1.1 papers, or about 8%. For women, the corresponding numbers are 0.82 papers, or about 12%, suggesting that joint production may be more important for males in absolute terms and for females in relative terms.

The time variable Experience is highly significant and positive in all three regressions. For the pooled results, the marginal year of experience is associated with an increase of 0.76 articles, an increase of 0.99 papers for males, and 0.45 for females. ExpSq is negative in all regressions and significant in the pooled and male regressions, suggesting that human capital deteriorates over time, the payoff to scholarly work declines, one simply tends to lose interest, or some combination of these factors.

We earlier suggested that individuals are likely to differ with respect to personal characteristics, such as innate ability and drive, and that persons with greater ability to produce articles are more likely to be employed in institutions with the highest research expectations. Results for the variable HighResearch suggest that this is true, at least for males. The pooled results show that the highest research expectation corresponds to 6.4 more articles produced as compared to those working where teaching is the primary interest. Males in high research expectation institutions produce 7.23 more articles. The HighResearch coefficient for males is highly significant. The corresponding difference for females is 2.7 more papers produced, but the relationship is not highly significantly--about the 15% level of confidence. Publishing in top 25 journals (PercentTop25) does not appear to effect overall productivity of either male or female economists.

Recall our earlier assumption that the production of books and other scholarly output could reduce article production if time and other trade-offs dominate. Alternatively, if such efforts are complementary or if they indicate a higher level of ability or drive, they will be associated with greater article production. We find that Books is highly significant and positive in all three regressions and that the marginal book is associated with an increase of 2.17 articles for males and 0.45 articles for females. (15) Chapters is also significant, and the coefficients for males and females are, respectively, 0.83 and 0.57. These findings suggest that the two variables pick up differences in ability and drive and/or that there are complementary effects.

Of the three remaining controls the Race variable (value equals one if the respondent is other than white U.S. citizen, zero otherwise) is never significant. When males work in a Public institution, they produce, on average, about 3 more papers, and their female counterparts produce about 1.25 more. The coefficient in the male regression is highly significant and that for the female regression is significant at 15%. The Gender variable (value equals one if male) from the overall regressions is positive and significant, suggesting that the typical male produces 4.1 more articles, other things equal. To determine what might cause this disparity, we employ the decomposition method pioneered in economics by Oaxaca (1973) and Blinder (1973). (16) The Oaxaca-Blinder procedure, commonly used in studies of male and female (or race-based) earnings or income differences, separates the productivity differential into two parts. One part is that associated with such variables as experience or institutional research orientation, and the second part is a residual.

The Oaxaca and Blinder adjustment reveals that 41.3% of the difference between male and female article production is explained by experience, number of courses taught, type of university orientation, and other control factors. More than half is left as an unexplained residual, a result that is consistent with the findings of McDowell and Smith (1992). The unexplained difference between typical male and female article publication rates may be related to discriminatory practices (the usual implication in wage and earnings studies) in the publication process (Ferber and Teiman, 1980). However, unmeasured choice factors may also be contributory. Perhaps, for example, women are more involved in service activities at the expense of research. Another possible explanation is that research output may fall during childrearing years, although Kahn (1995) finds that the proportion of tenure-track women dropping out of the labor force for extended periods is very small. The data collected for this project do not allow us to provide a thorough analysis, so explanation of the male/female publication rate differential is left to future researchers.

V. CONCLUSIONS

This article develops and tests a model of the determinants of articles published by individual scholars, using data from a survey of academic economists who were members of the American Economics Association in 1989. The empirical results indicate that for both males and females, article production varies with teaching load, teaching and research orientation of one's institution, years of experience and related human capital levels, the rate of coauthorship among scholars, and a number of other control factors. Neither race nor ethnicity appears to affect total articles produced. Of particular interest is the statistically significant positive relationship, for both men and women, between total article production and the rate of coauthorship. This result suggests that comparative advantage and other factors that promote production efficiency are enhanced by cooperation and may help explain the increasing rate of coauthorship that has been observed in recent years.

According to the data used here (and consistent with conclusions from earlier studies) males produce more articles than females in part because males have more experience and more often teach at high-research environment schools. Even after all controls are accounted for, however, a significant residual, gender-specific difference in article production remains. This result suggests an area of analysis that is ripe for future study.

APPENDIX: OAXACA/BLINDER DECOMPOSITION METHODOLOGY

The decomposition methodology developed by Oaxaca and by Blinder requires that number of articles (A) be estimated, using the regression method of ordinary least squares, for each of the two groups (i = male, female) to be compared, according to:

(1) [A.sub.i] = [[summation of].sub.k.sub.j=o] [B.sub.j][X.sub.j] + [e.sub.j]

where [A.sub.i] the number of articles produced for each group, [B.sub.o] is the constant term, X is a vector of the j independent variables that are theoretically linked to articles produced, [B.sub.j] is a vector of j corresponding parameters, and [e.sub.j] is the error term.

Taking males (m) to be the favored group, a total articles difference is computed between this group and females (f)

(2) E[D.sub.f] = [A.sub.m] - [A.sub.f] = Total difference in number of articles produced between males and females

This total difference is then separated into two components, one of which is attributable to differences in the attributes of the groups (for example, experience, number of courses taught, type of university, etc.) and the other, which is generally assumed to be predominately gender-based. The former measure has been called the "endowment effect" or "skill" (END) and the latter is known as the "residual effect" (RE). To separate that portion of total articles produced that is related to gender differences, the following estimations and computations are performed.

(3) R[E.sub.f] = E[D.sub.f] - EN[D.sub.f]

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

where R[E.sub.f] is the residual effect for females, E[D.sub.j] is the total articles differential, EN EN[D.sub.f] is the females endowment effect, [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] is the sum of parameter estimates times the mean values of the independent variables for the males equation, and [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] is the sum of parameter estimates for the males equation (1) times the mean values from the females version of equation (1). See Table 3 for results.

TABLE 1
Means and Standard Deviations of Variables

Variable Males and Females Males Females

 Mean SD Mean SD Mean SD

Articles 11.57 15.50 13.92 17.33 6.95 9.54
Gender 0.66 0.47 -- -- -- --
Race 0.89 0.31 0.87 0.33 0.93 0.26
Experience 15.48 9.09 17.15 9.25 12.21 7.80
ExpSq 322.26 351.19 379.60 377.55 209.78 258.48
Books 0.90 2.29 1.03 2.46 0.65 1.91
Chapters 3.32 6.67 3.54 7.03 2.87 5.88
Courses 5.10 2.06 5.09 2.15 5.13 1.88
Teaching-Research 0.37 0.48 0.38 0.49 0.37 0.48
High Research 0.09 0.29 0.11 0.31 0.06 0.24
Administration 0.24 0.43 0.26 0.44 0.19 0.40
Public 0.57 0.50 0.61 0.49 0.51 0.51
Percent Coauthored 39.50 35.40 39.20 34.10 40.10 37.80
Percent Top25 13.10 22.60 13.10 22.30 12.88 23.10

TABLE 2
Determinants of Articles Produced

 Males and Females (a)

 Regression
Variable Coefficient p-Value

Gender 4.05 0.00
Race 0.51 0.66
Experience 0.76 0.00
ExpSq -0.01 0.00
Books 1.80 0.00
Chapters 0.78 0.00
Courses -0.35 0.09
Teaching-Research 4.82 0.00
High Research 6.44 0.00
Administration -1.52 0.09
Public 2.51 0.00
PercentCoauthored 0.05 0.00
Percent Top 25 0.00 0.91
Constant -6.60 0.00

 Males (b) Females (c)

 Regression Regression
Variable Coefficient p-Value Coefficient p-Value

Gender
Race 0.04 0.98 1.16 0.46
Experience 0.99 0.00 0.45 0.01
ExpSq -0.02 0.00 0.003 0.59
Books 2.17 0.00 0.45 0.06
Chapters 0.83 0.00 0.57 0.00
Courses -0.47 0.08 -0.14 0.58
Teaching-Research 5.59 0.00 3.86 0.00
High Research 7.23 0.00 2.70 0.14
Administration -1.95 0.10 -0.96 0.39
Public 3.12 0.00 1.25 0.13
PercentCoauthored 0.06 0.00 0.04 0.00
Percent Top25 0.002 0.99 0.01 0.56
Constant -4.56 0.12 -4.02 0.11

(a) F-statistic 71.99; adjusted [R.sup.2]
0.48; number of observations 1000.

(b) F-statistic 52.96; adjusted [R.sup.2]
0.49; number of observations 662.

(c) F-statistic 21.33; adjusted [R.sup.2]
0.42; number of observations 338.

Table 3
Oaxaca/Blinder Decomposition Results

 Percent of
 Oaxaca/Blinder Total

E[D.sub.f]=Difference 6.97
in the number of articles

EN[D.sub.f]=Difference due 2.88 41.3
to variables included
in regression

R[E.sub.f]=Gender differences 4.09 58.7

(1.) Kahn (1995) provides an overview of the status of women in economics. Hansen et al. (1978) finds that women economics produce fewer publications, and Johnson and Stafford (1974) and Farber (1977) find that women economists have a lower chance of promotion.

(2) For examples, see Katz (1973), Siegfried and White (1973), Johnson and Stafford (1974), and McDowell and Smith (1992).

(3.) McDowell and Melvin (1983) only account for time available for research, whereas our model includes three uses of time: research, teaching, and service.

(4.) The reader is referred to McDowell and Melvin (1983) for a more detailed discussion of the utility maximization problem.

(5.) Sauer (1988) finds that the salary return from a coauthored paper with n authors is 1/n times that of a single-authored paper. However, his data set includes only associate or full professors at seven "top 40" academic departments. Our data do not allow for discounting articles by (1/n).

(6.) To better preserve confidentiality, we did not collect information on respondents' graduate institutions. The variables Teaching-Research, HighResearch, and Books should pick up quality differences.

(7.) The questionnaire was worded so that textbooks, edited volumes, and so on, were excluded from Books.

(8.) Simple-order correlations between Articles and Books and Chapters are in the range of 0.5.

(9.) Ideally, separate regressions would be run, but there were insufficient observations.

(10.) The observed difference between Grimes and Register's (1997) data and our data could be due to the fact that there are more females in our sample, more respondents from low-research expectation institutions, or other factors.

(11.) Using a data base of 960 make and female economists graduating between 1969 and 1984, Fish and Gibbons (1989) found that by 1986, of the 30% who had published, 56% were men and 44% were women. Over the much shorter time period than for this paper or for Grimes and Register (1997), Fish and Gibbons found that average production for males (females) was 1.76 (1.01) articles.

(12.) Ordinary least squares regression is used for estimating the model, with survey responses weighted to adjust for the fact that females are oversampled. The weights reflect the relative proportions of males and females in research, doctoral, comprehensive, and liberal arts schools.

(13.) We also estimated four other versions of the model. The model presented in this paper produced the highest F-statistics and adjusted [R.sup.2]. For the pooled regressions, the alternative models produced no changes in sign or significance for the variables used here. For the male regressions, Administration becomes statistically insignificant in one of the alternative models. For females, Courses becomes significant in one model, and Public. is significant in another model. Results are available on request.

(14.) The reader is urged to interpret the coefficients presented as indicators rather than point estimates of the respective parameters.

(15.) The questionnaire was worded so that textbooks edited volumes, and so on were excluded from Books.

(16.) More details on the Oaxaca/Blinder decomposition method are provided in the Appendix.

REFERENCES

Barnett, A., R. Ault, and D. Kaserman. "The Rising Incidence of Co-authorship in Economics: Further Evidence." Review of Economics and Statistics, 70(3), 1988, 539-43.

Blinder, A. "Wage Discrimination: Reduced Form and Structural Estimates." Journal of Human Resources, 8(4), 1973, 436-55.

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

Durden, G. C., and T. J. Perri. "Coauthorship and Publication Efficiency." Atlantic Economic Journal, 23(1), 1995, 69-76.

Farber, S. "The Earnings and Promotion of Women Faculty: Comment." American Economic Review, 67(2), 1977, 199-206.

Ferber, M. A., and M. Teiman. "Are Women Economists at a Disadvantage in Publishing Journal Articles?" Eastern Economics Journal, August--October 1980.

Fish, M., and J. D. Gibbons. "A Comparison of the Publications of Male and Female Economists." Journal of Economic Education, 20(1), 1989, 93 105.

Formby, J., W. Gunther, and R. Sakano. "Entry Level Salaries of Academic Economists: Does Gender or Age Matter?" Economic Inquiry, 31(1), 1993, 123-38.

Graves, P. E., J. R. Marchand, and R. Thompson. "Economics Department Rankings: Research Incentives, Constraints, and Efficiency." American Economic Review, 72(5), 1982, 1131-41.

Grimes, P. W., and C. A. Register. "Career Publications and Academic Job Rank: Evidence from the Class of 1968." Journal of Economic Education, 28(1), 1997, 82-92.

Hamermesh, D. S., and S. M. Oster. "Tools or Toys? The Impact of High Technology on Scholarly Productivity." Economic Inquiry, 40(4), 2002, 539-555.

Hansen, W. L., B. A. Weisbrod, and R. P. Strauss. "Modeling the Earnings and Research Productivity of Academic Economists." Journal of Political Economy, 86(4), 1978, 729-41.

Heck, J. L., and P. A. Zaleski. "Trends in Economic Journal Literature." Atlantic Economic Journal, 19(4), 1991, 27-32.

Hollis, A. "Co-authorship and the Output of Academic Economists." Labour Economics, 8(4), 2001, 503-30.

Hudson, J. "Trends in Multi-Authored Papers in Economics." Journal of Economic Perspectives, 10(3), 1996, 153-58.

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

Kahn, S. "Women in the Economics Professsion." Journal of Economic Perspectives, 9(4), 1995, 193-05.

Katz, D. "Faculty Salaries, Promotions, and Productivity at a Large University." American Economic Review, 63(3), 1973, 469-77.

Laband, D. N., and R. D. Tollison. "Intellectual Collaboration." Journal of Political Economy. 108(3), 2000, 632-62.

Liebowitz, S. J., and J. P. Palmer. "Assessing the Relative Impacts of Economics Journals." Journal of Economic Literature, 22(1), 1984, 77-88.

McDowell, J. M., and M. Melvin. "The Determinants of Coauthorship: Analysis of the Economics Literature." Review of Economies and Statistics, 65(1), 1983, 155-60.

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

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

Piette, M. J., and K. L. Ross. "An Analysis of the Determinants of Co-authorship in Economics." Journal of Economic Education, 23(3), 1992, 277-83.

Sauer, R. D. "Estimates of the Returns to Quality and Coauthorship in Economic Academia." Journal of Political Economy, 96(4), 1988, 855-66.

Siegfried, J. J., and K. J. White. "Financial Rewards to Research and Teaching: A Case Study of Academic Economists." American Economic Review, 63(2), 1973, 309-15.

KELLIE L. MASKE, GAREY C. DURDEN, and PATRICIA E. GAYNOR *

* We wish to thank Alison Aughinbaugh, Todd Cherry, John Dawson, Tim Perri, and Janet K. Smith for helpful comments. We also thank Rachel Jordan for assistance with the data.

Maske: Assistant Professor, Appalachian State University. Boone, NC 28608. Phone 1-828-262-7652, Fax 1-828-262-6105, E-mail maskekl@appstate.edu

Durden: Professor Emeritus, Appalachian State University, Boone, NC 28608. Phone 1-828-262-2148, Fax 1-828-262-6105, E-mail durden_g@bellsouth.net

Goynor: Professor Emeritus, Appalachian State University, Boone, NC 28608. Phone 1-828-262-2148, Fax 1-828-262-6105, E-mail gaynorpe@appstate.edu