# Coauthorship and publication efficiency.

I. IntroductionSeveral papers have investigated certain aspects of the recent trend toward proportionally increasing coauthorship among economists. Heck and Zaleski [1991] used an aggregated data set compiled from the Journal of Economic Literature and examined the 1969-89 time period finding, among other things, that coauthorship has increased from about 15.1 percent of total papers listed in 1969 to 35.7 percent of those listed in the JEL in 1989. Using citation counts to isolate important articles published in the American Economic Review between 1965-85, Sophocleus [1993] reported that coauthorship over the period increased from 29.4 percent to 52.8 percent of the total articles published.

Barnett, Ault, and Kaserman [1988] collected a sample of 2,535 papers from the AER over the same 1960-85 period and successfully estimated the determinants of total authors per article. McDowell and Melvin [1983] also modeled and tested the determinants of coauthorship, and Sauer [1988] has estimated the academic returns to coauthorship. McDowell and Smith [1992] investigated the phenomenon of gender sorting among coauthors (men more often work with men and the converse) and found evidence that this practice may decrease productivity among women as compared with men. The problem arises because there are far fewer coauthorship opportunities for women due to their relatively small numbers, especially in smaller schools.

Although some of the studies cited do present and develop the notion that coauthorship among economists enhances productivity, very little formal work has been done to model and test this hypothesis.(1) The purpose of this paper is to develop and test a simple model of the effect of coauthorship on article production and test the model using data (summarized in Table 1) from Heck and Zaleski [1991].

Section 2 of this paper provides a discussion of the underlying reasons for increasing coauthorship and provides intuitive reasoning as to why coauthorship should increase article production. The model is developed in section 3 and the empirical results discussed in section 4. A summary and some conclusions are presented in section 5.

II. Coauthorship as a Basis for Increasing Article Production

Most authors will want to publish individually in order to demonstrate this capability to those who make recommendations for promotion, tenure, and salary increases. Single-authored publications may also be important, at the margin, for the enhancement of mobility. There is little doubt, however, that the growing "publish or perish" requirement plays a major role in the observed proportional increase in coauthorship. This contention is enhanced by the fact that schools not formerly research-oriented at all (e.g., many of those classified as Liberal Arts or Comprehensive)(2) have recently incorporated a substantial research expectation into all review procedures. Given the observed proportional increase in coauthorship and the increased demand for greater research output, it follows that there may be a significant relationship between the number of articles produced per year and the proportion coauthored.

Economic theory suggests that coauthorship may increase article output in several ways. The most obvious reason is that, when authors work jointly, the advantages of specialization and the division of labor may be incorporated. McDowell and Melvin [1983] and Barnett, Ault, and Kaserman [1988] point out that, as the body of knowledge in economics and the degree of sub-specialization have increased, the necessity for individuals to work in a relatively narrow sector has also increased. The relatively high levels of knowledge and expertise in some areas and relatively low levels in others for a given individual will often make cooperation attractive and result in greater production. For example, some authors may be more proficient in the development of models while others may be better at the use of statistics in generating appropriate empirical tests. Some are better able to envision publishable projects while others are more adept at bringing such projects to closure.

Bennett, Ault, and Kaserman [1988] have provided an additional series of reasons why authors jointly publish, including: (1) The opportunity cost of time hypothesis which suggests that formerly "free" reviews by colleagues are too costly to the reviewers so that the expectation of coauthorship as payment has risen; (2) the quality hypothesis, which is that competition for journal space requires a higher-quality result, and this requirement is more easily fulfilled when tasks are divided according to expertise; and (3) the diversification hypothesis which argues that coauthorship spreads the risk of rejection, offsets the problem of extended length of time between acceptance and publication, and other related factors. Although perhaps not as directly relevant as the specialization and division of labor argument, all of these factors are seen as contributing to increased article output.

III. The Model

Suppose there are two individuals, 1 and 2, each of whom has time available (T) which can be devoted to either single-authored or coauthored publications. Let [q.sub.i] equal the output individuals i's single-authored publications, with [Mathematical Expression Omitted] where [t.sub.i] is the time output of spent in individual research and T- [t.sub.i] is i's time input in joint research for coauthored articles. If 0 [less than] [Delta] [less than] 1, then there is diminishing marginal productivity of time in individual research, as will be the case for joint research. Note that for

[t.sub.i] [greater than] 1, [Delta][q.sub.i]/[Delta][Delta] [greater than] 0,

d[Delta] [greater than] 0 implies a larger output of articles. Let the number of coauthored publications equal z, with z = B[(2T - [t.sub.1] - [t.sub.2]).sup.[Delta]] and B [greater than] 1. Thus, for example, if the two individuals each put in x hours working together on research, the number of articles produced would equal B times the number produced if either had spent 2x hours in individual research. The parameter B reflects the gains from coauthorship and will be referred to as the productivity effect of coauthorship.

Presumably each individual will allocate his or her available time to maximize article production. However, it is likely that a coauthored article will yield less in terms of salary increases, prestige, and so on, than a single-authored article.(3) Thus the individual will maximize [y.sub.i], where [y.sub.i] = [q.sub.i] + [Phi]z and 0 [less than] [Phi] [less than] 1. Both individuals are identical in that the same time input will yield the same number of publications. This does not imply that there is no comparative advantage: one may (individually) produce theoretical articles, and the other may do empirical work. Together they may produce articles with both kinds of analysis. In order to focus on the effects of B, [Phi] and [Delta] on time allocation, the authors have assumed a simple production technology which yields the same level of z for any total time input of the coauthors. Clearly this is not realistic, but adding realism in this regard would take us away from our focus on the "pure" effects of B, [Phi], and [Delta] on coauthorship.

When individual i chooses [t.sub.i] to maximize [y.sub.i], the result is with [Theta] [equivalent] [Phi] B:

[t.sub.i] = (2T-[t.sub.1]-[t.sub.2])[[Theta].sup.1/[Delta]-1]. (1)

With [t.sub.1] [equivalent] [t.sub.2]:

[Mathematical Expression Omitted].

Let [Psi][equivalent][[Theta].sup.1/[Delta]-1]. Then [Mathematical Expression Omitted], and [Mathematical Expression Omitted].

Since

[Delta][Psi]/[Delta][Theta] [less than] 0,

if little weight is accorded to coauthored publications, then [Psi] gets large and [t.sub.i] approaches T. A higher weight for coauthored articles (d[Phi] [greater than] 0), or a larger productivity gain from coauthoring (dB [greater than] 0) will lead to an increase in the time devoted to coauthored articles (d[t.sub.i] [less than] 0).

The above results are obvious and do not require a formal model. However, the impact of [Delta] on [t.sub.i] is not so obvious. Using equation (2), [Delta][t.sub.i]/[Delta][Delta] = - ln([Theta])k, where k is a positive term. Evidence presented in the next section suggests that publishing productivity is declining over time. In terms of our theoretical model, this would imply that d[Delta] [less than] 0. As coauthorship increases (decreases) over time, then this implies that [t.sub.i] decreases (increases), which suggests that [Theta] is less than (greater than) one. Thus the theoretical model may lead to indirect estimates of the relative magnitudes of the productivity effect of coauthorship (B), and the way in which evaluators discount coauthored articles ([Phi]).

IV. The Data, Estimating Equation and Results

The data used with the model are presented in Table 1. Here the growth in number of articles is shown to be quite substantial, from 3,924 in 1969 to 11,636 in 1992, a nearly three-fold increase.(4) The number coauthored has grown even faster, from 594 to 4,466, a factor of more than seven. This works out to a percentage increase of from 15.1 percent in 1969 to 38.2 percent in 1992. The number of journals indexed by the JEL has somewhat less than doubled, suggesting that the average journal now produces more articles.

Articles produced have grown considerably faster than AEA membership, which has been generally stable, but has waxed and waned somewhat with economic conditions. Articles per AEA member have grown from .240 to .659 during the 24-year period of this analysis. These data leave little doubt that economists have recently become more active and that an increased emphasis on coauthorship is evident. The statistical tests reported in section 5 suggest that the two phenomena are related as theory suggests. The specific hypotheses to be tested can be summarized in a very simple estimating equation:

[(Articles).sub.i] = a + [a.sub.1] [(Time).sub.i] + [a.sub.2] [(Timesq).sub.i] + [a.sub.3] [(AEA).sub.i] + [a.sub.4] [(Coauthored).sub.i] +[a.sub.5] (Csq) i[e.sub.i] (3)

[TABULAR DATA FOR TABLE 1 OMITTED]

where Time is a trend variable with values 1-24; the relevant population is measured by AEA membership in year i; Coauthored is the number of coauthored articles in that year; Csq is the square of Coauthored; and [e.sub.i] is the stochastic error term. The inclusion of Time in the regressions is for the purpose of capturing any productivity influences which are not specifically spelled out in the model. Timesq is the square of Time and should capture related nonlinearities. The inclusion of AEA membership should capture the relationship between population growth and article authorship, ceteris paribus. The number of Coauthored articles per year should contribute to more articles if coauthorship does increase total output because of the efficiency characteristics discussed earlier. Csq is entered to determine whether the influence of coauthorship decays as more coauthored papers appear.

The empirical tests of the model are summarized in Table 2. As is generally true for time-series estimations, R-Squared values are in the upper nineties. For ease of reference, both the t-statistics and the related p-values (marginal significance levels of the relevant t-values) are reported beneath the appropriate parameter estimate.

[TABULAR DATA FOR TABLE 2 OMITTED]

Two versions of the model were run, one for the determination of total articles published, and one for articles per-capita (Articles/AEA membership). In both cases, the Durbin-Watson tests for serial correlation were inconclusive. Because of this, a second equation was run for both versions which included an adjustment for serial correlation. The statistical package, MicroTSP, performs this procedure when so directed, and the result is shown in regressions 2 and 4 for the variable AR(1). The adjustment suggests no serial correlation since the null hypothesis is not rejected. Importantly, the adjustment had virtually no effect on any of the relevant coefficients or their p- or t-values, with the exception that Timesq in equation 2 exhibits an increase in significance level. All variables are shown to be statistically significant at the 5 percent level or better (two-tail test), except for AEA in regressions 1 and 2 and Timesq in regressions 3 and 4.

The coefficients on AEA and Time are negative in all regressions, with that for Time consistently significant and those on AEA significant in regressions 3 and 4. The related p-values range from less than 1 percent to about 1.17 percent for the Time coefficients and less than 1 percent for the AEA coefficients in the per-capita equations. This suggests that factors other than coauthorship may have worked to reduce article output. Coauthored is always positive and is significant at far greater than the I percent level (p [less than] 1 percent). This result, in combination with the results related to the passage of time and to AEA membership, suggests that coauthorship plays a very major role in the observed increases in total and per-capita article production. The significant (p [less than] 1 percent) and negatively signed coefficient on Csq indicates that the ability of coauthorship to increase article output decreases as coauthorship grows. This is logical since journal space will always be limited and increasing coauthorship cannot indefinitely increase output.

If [Delta] has decreased over time, the increase in the percentage of articles that is coauthored (Table 1) suggests that [Delta][t.sub.i]/[Delta][Delta] [greater than] 0. As demonstrated above, this requires [Theta] [less than] 1. Thus, our theoretical model suggests that: (1) coauthored articles are weighted lower than single-authored articles ([Phi] [less than] 1), and (2) the lower weight for coauthored articles dominates the productivity effect of coauthorship (B [greater than] 1) so that [Theta] = [Phi]B [less than] 1.

V. Summary and Conclusions

This study has reported tests of a simple model of the determinants of economics journal articles published per year for 1969-92. Results indicate that not only is coauthorship increasing at a very fast rate, as noted by Hock and Zaleski [1991], but also that coauthorship appears directly and very strongly related to increased output. This appears to be true for both total articles published and articles published per-AEA member. Since coefficients on both Time and AEA are negative in all regressions, it appears that coauthorship may be the major reason for recent observed increases in article production, more than compensating for an apparent decline in average publishing ability. Finally, indirect evidence suggests that coauthored articles are weighted (by those who determine rewards for publication) lower than single-authored articles, and that this effect dominates the productivity effect of coauthorship.(5)

The pressures on economists and other academics to publish refereed journal articles will almost certainly continue at the large research schools. Such pressures may well increase for those working at comprehensive and liberal arts institutions. Results from this study suggest that an increasing tendency to coauthor will be one means of achieving the required higher rates of scholarly output.

1 One exception is found in McDowell and Smith [1992]. They used a cohort sample of 89 men and 89 women receiving Ph.D.s between 1968-75. In a sub-sample of 106 persons, they found no evidence that coauthorship increased the number of articles per scholar when total articles were discounted by the number of coauthors.

2 The classification of schools as research, doctoral, comprehensive, and liberal arts is found in a report from the Caarnegie Foundation for the Advancement of Teaching [1987].

3 When departments attempt to quantify the relationship between raises and scholarly production, coauthored papers are generally discounted. However, McDowell and Smith [1992] find that coauthored articles have the same weight in promotion decisions as single-authored articles.

4 The original table from Heck and Zaleski [1991] has been extended by three years and a new column, "Articles Per Capita," has been added.

5 Leibowitz and Palmer [1986] report a survey of department chairs which suggests that, on average.

REFERENCES

Barnett, Andy H.; Ault, Richard W.; Kaserman, David L. "The Rising Incidence of Coauthorship in Economics: Further Evidence." Review of Economics and Statistics, August 1988, pp. 539-43.

Carnegie Foundation for the Advancement of Teaching. A Classification of Institutions of Higher Education. Princeton: The Carnegie Foundation, 1987.

Heck, Jean Louis; Zaleski, Peter A. "Trends in Economic Journal Literature." Atlantic Economic Journal, December 1991, 19, pp. 27-32.

Leibowitz, S. J.; Palmer, J. P. "Assessing Assessments of Economics Departments." Mimeo, University of Western Ontario, 1986.

McDowell, John M.; Smith, Janet Kilholm. "The Effect of Gender-Sorting on Propensity to CoAuthor: Implications for Academic Promotion." Economic Inquiry, January 1992, 30, pp. 68-82.

McDowell, John M.; Melvin, Michael. "The Determinants of Coauthorship: An Analysis of the Economics Literature." Review of Economics and Statistics, February 1983, 65, pp. 155-60.

Sauer, R. "Estimates of the Returns to Quality and Coauthorship in Economics Academia." Journal of Political Economy, 96, 1988, pp. 855-66.

Sophocleus, John P. "Further Exploration into the Iron Law of Important Articles." Mimeo, Auburn University. 1993, 9 pages.

Printer friendly Cite/link Email Feedback | |

Author: | Durden, Gary C.; Perri, Timothy J. |
---|---|

Publication: | Atlantic Economic Journal |

Date: | Mar 1, 1995 |

Words: | 2793 |

Previous Article: | Golden parachutes, information and shareholder value. |

Next Article: | Neutrality of money and the Fisher hypothesis: further empirical test. |

Topics: |