Editorial favoritism in economics?

Medoff, Marshall H.

1. Introduction

In 1990, the Journal of Economic Literature listed the contents from over 300 economics journals. The journals ranged from general interest (e.g., American Economic Review, Journal of Political Economy, Quarterly Journal of Economics) to those specializing in specific areas (e.g., business, finance, law, real estate, trade). Economic knowledge is now disseminated primarily through a journal-dominated system. The audience is predominantly professional academic economists who screen the research of each other in order to certify its quality. Publication in these journals is a necessary condition for tenure, promotion, influence, reputation, and mobility. By chance or choice, economics journals hold the keys to success for academic economists.

Not surprisingly, given the rewards from publishing, there has been substantial research on economics journals. A quality hierarchy has been found to exist among economics journals based on such indicators as surveys, citations, and institutional affiliations of authors. The more visible or highly cited economics journals tend to be the more prestigious or higher-quality journals (Moore 1972; Hawkins, Ritter, and Walter 1973; Laband and Piette 1994b).

Considerably more attention has focused on the quality-control appraisal system used in the publication process: peer review. In order to publish in virtually any academic economics journal, an author must receive a favorable evaluation from an anonymous peer saying that the paper is potentially publishable and, if necessary, what further work is needed to make the paper worthy of publishing. The paper is not approved for publication unless and until any suggested revisions are made to the satisfaction of the editor and referee. The peer review process has been examined in terms of (i) what the characteristics and functions of these anonymous referees are and (ii) whether they are fair and objective in their evaluation.

Hamermesh (1994) examined the characteristics of referees at four general and three specialty economics journals and found that referees are overwhelmingly male, with 16 years of experience since the receipt of their Ph.D. degree, and are typically of higher quality than the author of the paper they are reviewing. Higher-quality journals have higher-quality referees who are not systematically assigned to review higher-quality authors. Mackie (1998) surveyed referees at seven economics journals and found that referees use a set of highly subjective and interpretive criteria in assessing the significance or quality of a research paper. The criteria included originality, novelty, creativity, innovativeness, advances in existing economic knowledge, and relevance to real economic problems.

Laband (1990) found that referees have two functions. First, they screen the quality of the research conducted by their professional peers in order to determine ff the paper meets a minimum quality standard. Second, referees, through their comments and suggested revisions, increase the quality of a potential publication (as measured by the subsequent number of citations a paper receives in the six years after publication). However, this relationship is only statistically significant. The numerical impact of a referee's comments on the quality of a paper is virtually negligible. Using Laband's figures, a referee's comments increase the number of citations a published paper receives by less than 0.25 per year over the subsequent six-year period.

The second issue of concern about peer review is whether the evaluation process is fair. Peer review takes two forms: single blind (the author does not know the identity of the referee but the referee knows the author's identity) and double blind (neither the author nor the referee knows the identity of the other). Blank (1991) conducted a unique controlled experiment to analyze the effects of single-blind versus double-blind refereeing on papers submitted to the American Economic Review between 1987 and 1989. She found that under the double-blind system, acceptance rates are lower and referee comments more critical. Acceptance rates of authors at the top five ranked universities were not affected by the type of review system used. Authors at the near-top universities (ranks 6-50) had lower acceptance rates under double-blind reviewing. These results, however, provide no conclusion regarding the fairness of the peer review process. (1)

Very little research has focused on the behavior of the key journal decision makers in the review process: the editor/coeditors. Editors must decide whether research submissions are of sufficient quality to warrant publication in their limited number of journal pages. An editor's objective, presumably, is to produce a journal of the highest possible quality. Journal editors compete with each other to attract papers that will make the greatest scientific contributions. Because of the active competition between journals, editors attempt to persuade authors to submit their high-quality papers to them in exchange for a reduction in the transaction costs involved in the reviewing/publication process.

Critics of the editorial review process contend that the absence of any clearly defined criteria of what constitutes a significant high-quality contribution produces editorial favoritism in the review process (Folster 1995; Mackie 1998). It is argued that publication decisions are swayed by an author's personal or institutional connections to the editor or coeditors. The consequence is that nonscientific considerations influence editorial decisions.

There exists considerable anecdotal evidence regarding the perception that editorial favoritism exists in the review process. Bhagwati, editor of the Journal of International Economics. noted that he published a paper from a former student (Paul Krugman) despite the fact that there were two adverse referee reports from very distinguished experts and he did not normally publish his own students' work (Shepherd 1995, p. 89). Clower, when he was editor of the American Economic Review, frequently accepted research papers for publication without submitting them for peer review (Shepherd 1995, p. 99). While editor of the Review of Economics and Statistics, Houthakker read every incoming manuscript and summarily rejected papers without sending them out for formal review (Shepherd 1995, p. 107).

These examples of editorial discretion am not determinative since it cannot be inferred whether these editorial decisions am systematic or random. The crucial question is whether favoritism influences the choices editors make in the prepublication appraisal process. The methodological problem is that editorial favoritism is difficult to directly detect from acceptance rates for several reasons. First, most researchers do not have access to journal submissions that are needed in order to compare the characteristics of published and rejected authors. Second, an editor's choice of referees may predetermine the publication decision about a research paper. Editors may assign a paper to referees who are ideologically biased (in either direction) toward an author. Third, authors' decisions about what journal to submit their paper to may be influenced by their concerns, positively or negatively, about editorial favoritism.

It is possible to indirectly detect the presence of editorial favoritism from published articles. If editorial favoritism exists, one would expect to find quality differences in the articles by those authors with and without personal or institutional connections to the publishing journal's editor/coeditors. Laband and Piette (1994a) examined articles published in 28 economics journals in 1984. Using as a measure of an article's quality the number of citations received in the subsequent five-year period following publication, (2) they found that articles with an author/editor connection were of higher quality than those without such connections.

There are, however, several serious methodological problems with Laband and Piette's analysis. First, their measure of an author/editor connection was flawed. Laband and Piette (1994a, p. 197) define an author/editor connection to exist whenever "any of the authors of an article received his or her Ph.D. from the same university that the editor, coeditor, or any associate editor who published the paper was affiliated with in 1984 or received his or her Ph.D. degree from, or if any of the authors of a paper was affiliated in 1984 with the same university that the editor, coeditor, or any associate editor was affiliated with in 1984 or received his or her Ph.D. degree from." The problem with this author/ editor connection variable is that it fails to recognize or understand the power relations and structure within a journal. With few exceptions, only the editor/coeditors of economics journals have the power and discretion to accept research papers for publication. (3) Associate editors, assistant editors, or board of editors do not make the final accept/reject decision. The Laband and Piette author/editor connection variable is not only misspecified but is also so aggregative that it is difficult, if not impossible, to determine exactly what effect the variable is actually measuring. Second, the sample of 28 economics journals used by Laband and Piette is heavily weighted toward specialty journals (agriculture, tax, law, business, finance) that do not publish the full spectrum of economic research and, as a consequence, are cited less often. Finally, Laband and Piette utilize an incorrect estimation procedure to estimate their citation equation. They use ordinary least squares. But their dependent variable, the number of citations an article receives, can have only nonnegative values. The use of ordinary least squares of a censored regression model produces biased and inconsistent parameter estimates (Greene 1993).

The purpose of this paper is to empirically investigate whether editorial favoritism exists in economics. Are articles authored by those with explicit and identifiable personal ties or institutional connections to the publishing journal's key decision makers (editor/coeditors) of lower quality as compared to those articles by authors without such connections?

2. Data

The data for this study come from six top economics journals (American Economic Review, Econometrica, International Economic Review, Journal of Political Economy, Quarterly Journal of Economics, Review of Economic Studies). These six journals all appear on Diamond's (1989) list of "core" economics journals and are consistently top ranked in terms of quality. These journals publish the full spectrum of research in economics (theoretical, applied, modeling, econometrics, microeconomics, macroeconomics). Detailed information was collected on the 359 articles and original notes published in this core journal set in 1990. Excluded from this data set were all presidential addresses, Nobel Prize lectures, comments and replies. The information collected included length of the article, name(s) of the author(s), professional affiliation, article placement position, and the subject area in which the article was classified in the Journal of Economic Literature. Detailed information on each author (Ph.D.-granting institution and years of attendance, professional affiliations, and years of employment) was obtained from the December 1989 American Economic Review Survey of Members.

3. Model

The specification of the article quality equation is

(1) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

The measure of an article's quality (CITATIONS, 1991-2000) is the total number of citations received during the 10-year period (1991-2000) following the 1990 publication of article i from each of the six journals previously listed (excluding self-citations). (4)

To account for the variation in journal page size, the number of pages of each article was standardized to American Economic Review equivalent size pages (PAGES). Higher-quality research, generally, requires greater exposition than shorter, less substantive research.

The measure of journal quality is the Laband and Piette (1994b) 1990 normalized weights (0-100) of the relative impact articles published in each of the six core journals have on the economics profession (JOURNALQUALITY).

Author quality (AUTHORCITES, 1970-89) is measured by the total number of citations received by author i (or the average for multiauthored papers) during the previous 20 years (1970-1989) prior to the publication of article i, excluding self-citations. Citations to article i may be influenced by the quality/reputation of author i for several reasons. First, the total number of citations an author has received in the past is an indicator of the expected scientific contribution of the current article. (5) Second, the stock of prior citations may also reflect Merton's (1968) "Matthew effect," which argues that articles published by researchers of known past reputation will tend to receive greater increments of recognition than articles by less well-known researchers. Third, the total number of prior citations accumulated by author i may provide a signal to journal editors/referees that the citing researchers are aware/familiar with the reputational ranking of author i (David 1994). Fourth, the stock of prior citations may influence subsequent citations to article i because of parochial citation loyalty. The citation practices of researchers may reflect greater knowledge or familiarity with the professional standing or ideological preferences (biases) of the more prolific faculty of their doctoral institution (Stigler and Friedland 1975).

In economics, as in other disciplines, the number of citations to an article may depend, to some extent, on the subject matter. Articles written in certain areas may have greater interest, visibility, or significance. Each article from the core journal sample was categorized using the 1991 Classification System for Articles and Abstracts in the Journal of Economic Literature (JEL). This classification system replaced the previous 10 broad economics subject categories (e.g., general economics included both microeconomics and macroeconomics) with more detailed subject area categories. The subject area variables are 15 binary variables equal to one if the article appeared in the JEL's subject area classification A, C, E, F, G, H, I, J, K, L, M, N, O, Q, and R. (6) The omitted control category was subject area D (Microeconomics), which had the largest percentage of articles published in the sample.

It has been suggested that the position of an article in a journal issue provides a market signal to readers about the expected quality of the article (Laband and Piette 1994a). The signal provides readers a means for identifying research that is expected to provide the greatest substantive value. Thus, article position in a journal issue may influence subsequent citations. One measure of article position is the variable LEADARTICLE, which equals one ff article i was the lead article from each of the 1990 issues of the six journals previously listed. (7) As a further check on the accuracy of my results, I also estimated Equation 1 using the position number of article i in the particular journal issue instead of the variable LEADARTICLE. Because the number of articles differs between journals as well as issues (being the fifth article out of six articles is different than being the fifth out of 20 articles), the article placement position numbers were normalized (NORMALIZEDARTICLE#). The first article in a journal issue is assigned the value one, the last article is assigned the value zero, and the rest of the articles are between zero and one. Articles in the front half of a journal issue receive values above 0.50, and those in the bottom half receive values below 0.50.

The variable COAUTHOR equals one if article i has more than one author. It has been argued that economists who collaborate produce scientific contributions of higher quality than sole authors because collaboration allows economists to capture the efficiency gains from specialization and division of labor (Sauer 1988; Hamermesh and Oster 2002).

Three variables are used to encompass the full range of possible institutional connections and personal ties between authors and the editor/coeditors of the publishing journal. Institutional connections are measured by the binary variable INSTITUTIONALCONNECTION, which is equal to one if (i) author i presented the paper, prior to its 1990 publication, at a seminar or workshop held at an affiliation of any of the publishing journal's editor or coeditors; (8) or (ii) author i was affiliated, as of 1990, at the same university of any of the publishing journal's editor or coeditors; or (iii) author i is affiliated, as of 1990, with a university that any of the publishing journal's editor or coeditors received his Ph.D. degree from; or (iv) author i was a former graduate student of any of the publishing journal's editor or coeditors; or (v) author i attended graduate school contemporaneously with any of the publishing journal's editor or coeditors. (9)

Two binary variables are used to measure the personal ties or connections between authors and the publishing journal's editor/coeditors. The first variable, EDITORIALBOARD, is equal to one if author i was a member of the 1990 editorial board of the publishing journal. The second variable is THANKEDITOR, which equals one if author i, in the notes section of the article, thanked or acknowledged the publishing journal's editor or coeditors for constructive comments or suggestions on the paper. The means and standard deviations of the variables used in Equation 1 are reported in Table 1.

4. Empirical Results

Since the dependent variable in Equation 1, the total number of citations received by article i from 1991 to 2000, is left censored at zero, the Tobit maximum likelihood estimation technique is used. The empirical results appear in Table 2, column 1, when LEADARTICLE is used as the article position variable in Equation 1, and in column 2, when the normalized article placement position number variable is used. (10)

As expected, article length, journal quality, and author quality all have a significantly positive impact on the number of subsequent citations an article receives. Of the 15 JEL subject area binary variables (which are not shown in Table 2 because of space limitations), only category O (Economic Development, Technological Change, and Growth) has a statistically greater number of citations relative to category D (Microeconomics). The coauthorship variable is positive but not statistically significantly different from zero. Neither LEADARTICLE in column 1 nor the normalized article placement variable in column 2 has a statistically significant impact on the subsequent number of citations an article receives.

In both specifications of Equation 1 reported in Table 2, the institutional connection variable and the two personal connection variables have a statistically significantly positive impact on the number of citations to an article. The three connection variables to the publishing journal's editor/ coeditors are also numerically significant. Their coefficients indicate that, ceteris paribus, authors with these connections receive between 12 and 41 more citations than papers authored by those without such connections. The larger value of 41 is from articles authored by editorial board members of the publishing journal. To put these figures in context, Laband (1986) reported that 85% of all economics articles published are cited fewer than 10 times. Only 3% of all articles are cited more than 30 times.

In order to test the robustness of the previous empirical results, the EDITORIALBOARD variable was redefined as the number of authors of article i on the editorial board of the publishing journal divided by the total number of authors of article i (e.g., if a paper was coauthored and one of the coauthors was an editorial board member, the variable value is 0.5). Using this variable, Equation 1 was reestimated with LEADARTICLE and also with NORMALIZEDARTICLE#. The empirical results were virtually identical to those reported in Table 2, columns 1 and 2, with the institutional and personal connection variables within 0.01% of the previously reported values. I also used for author quality (AUTHORC1TES, 1970-89) of multiauthored papers the highest number of prior citations received by the most heavily cited author. (11) Once again, regardless of which article placement variable is used in Equation 1, the empirical results were within three decimal places of the results reported in Table 2, columns 1 and 2. (12)

The empirical results are consistent with the contention that journal editors/coeditors use their connections to reduce the search costs involved in identifying and attracting high-quality manuscripts to their journals. Journal editors/coeditors use their personal connections or ties to actively recruit high-quality authors to serve on their editorial boards not only for their expertise in refereeing manuscripts but also to attract submissions from these high-quality authors in order to have the right of first refusal on their high-impact articles.

It is possible that the previous results that articles by authors with personal or institutional connections to the publishing journal's editor/coeditors receive more citations occur not because their articles are of higher quality but because publication in an influential journal transmits a false signal to readers about the quality of the article. Slow (1991) argues that if refereeing is perceived to be more accurate in judging substantive scientific contributions in higher-quality journals, readers can then assume that a published article is a high-quality or high-impact paper. If, however, there exists a large random component in the editorial review process, as many have contended (Gans and Shepherd 1994), then Siow's model implies that publication in a journal is an imprimatur that may, initially, serve as an inaccurate proxy for the quality of the article. Over time, the marketplace of economic ideas determines the true quality of an article (provided that there is a perfectly elastic supply of new readers and a relatively small cost of determining the quality of an article). (13) This suggests that the timing of citations matters if editorial favoritism is present. If editorial favoritism exists, then as readers accurately assess an article's true quality, the initial positive impact of editorial connections should decrease over time. On the other hand, if editorial connections are actively used to recruit high-quality manuscripts, then the impact of editorial connections on an article's citations should remain positive and be constant over time.

In order to determine whether the impact of editorial connections on an article's subsequent citations decreases with an article's age, the dependent variable used in Equation 1, the total number of citations received from 1991 to 2000, is disaggregated into two separate dependent variables: (i) the total number of citations received from 1991 to 1995 and (ii) the total number of citations received from 1996 to 2000. Equation 1 was then reestimated with each dependent variable. The empirical results appear in Table 3, columns 1 and 2, when LEADARTICLE is used in Equation 1 and columns 3 and 4 when normalized article placement number is used.

Article length, journal quality, and author quality have a significantly positive and stable impact on subsequent citations over the two time subperiods. The effect of the institutional connection variable and the two personal ties variables (service on the publishing journal's editorial board and acknowledging the assistance of the publishing journal's editor/coeditors) is significantly positive in both time periods. The null hypothesis of equality of coefficients for each of the three editorial connection variables (INSTITUTIONALCONNECTION, EDITORIALBOARD, THANKEDITOR), between the two time intervals, cannot be rejected. The quality of articles with editorial connections is significantly higher than those without such connections, and this quality differential does not diminish over time.

5. Conclusion

There exists a belief among many academic economists that the publication process favors those authors with connections or personal ties to the publishing journal's editorial board. Articles from six core economics journals are examined to determine whether articles by authors with connections to the publishing journal's key decision makers, editor(s)/coeditors, are of lower quality than articles by those without such connections.

The empirical results show that articles authored by those with editorial connections, particularly serving on the publishing journal's editorial board, are both statistically and numerically of higher quality. Furthermore, this quality differential does not decrease over time.

The empirical results support the proposition that journal editors, in order to reduce the search costs involved in identifying high-quality manuscripts, use personal ties and institutional connections to persuade high-quality authors to submit their papers to them. Journal editors/coeditors attract these submissions by inducing high-quality authors to serve on their editorial boards as well as by offering constructive comments and suggestions on a high-quality author's paper, reducing the author's transaction cost of publishing.

Table 1. Variable Means and Standard Deviations

Variable Mean Standard Deviation

CITATIONS, 1991-2000 24.21 48.17
PAGES 15.73 6.14
JOURNALQUALITY 68.56 28.76
AUTHORCITES, 1970-89 176.97 493.41
NORMALIZEDARTICLE# .54 .30
COAUTHOR .40 .49
INSTITUTIONALCONNECTION .31 .46
EDITORIALBOARD .06 .23
THANKEDITOR .14 .34

Table 2. Tobit Estimation Results of Equation 1 (a)

 Dependent Variable: Dependent Variable:
 Total Citations, Total Citations,
Independent Variables 1991-2000 1991-2000

PAGES .9585 1.2701
 (1.99) * (2.36) *
JOURNALQUALITY .2803 .2738
 (2.79) ** (2.72) **
AUTHORCITES, 1970-89 .0188 .0195
 (2.97) ** (3.08) **
LEADARTICLE 14.3086
 (1.44) --
NORMALIZEDARTICLE# -- 9.5693
 (.90)
COAUTHOR 1.0661 -.1439
 (.18) (.02)
INSTITUTIONALCONNECTION 12.2120 13.2192
 (1.96) * (2.10) *
EDITORIALBOARD 39.5811 41.8607
 (3.21) ** (3.40) **
THANKEDITOR 18.4865 18.3265
 (2.33) * (2.31) *
CONSTANT -28.7927 -27.1448
 (2.71) ** (2.51) *
Log likelihood -1294.17 -1294.80

(a) Because of space limitations, the 15 JEL SUBJECT coefficients are
not shown. Absolute value of t-statistics in parentheses: ** =
significant at 0.01 level; * = significant at 0.05 level.

Table 3. Tobit Estimation Results of Equation 1, Citations 1991-1995
and 1996-2000 (a)

 (1) (2)
 Dependent Dependent
 Variable: Variable:
 Total Total
 Citations, Citations,
Independent Variables 1991-1995 1991-2000

PAGES .5941 .5443
 (2.88) ** (1.96) *
JOURNALQUALITY .1324 .1628
 (3.08) ** (2.47) *
AUTHORCITES, 1970-89 .0086 .0131
 (3.19) ** (2.83) **
LEADARTICLE 9.1067 7.0606
 (2.16) * (1.90)
NORMALIZEDARTICLE# -- --

COAUTHOR 4.3930 -3.3195
 (1.65) (.86)
INSTITUTIONALCONNECTION 5.2124 7.6877
 (1.96) * (1.98) *
EDITORIALBOARD 16.3703 23.6504
 (3.14) ** (2.99) **
THANKEDITOR 7.3221 12.4229
 (2.17) * (2.44) *
CONSTANT -16.0985 -20.6784
 (3.52) ** (2.96) **
Log likelihood -1040.50 -1101.39

 (3) (4)
 Dependent Dependent
 Variable: Variable:
 Total Total
 Citations, Citations,
Independent Variables 1991-1995 1996-2000

PAGES .6601 .7591
 (2.83) ** (1.96) *
JOURNALQUALITY .1269 .1625
 (2.93) ** (2.47) *
AUTHORCITES, 1970-89 .0090 .0131
 (3.35) ** (2.87) **
LEADARTICLE -- --

NORMALIZEDARTICLE# -.4253 -.7363
 (.90) (1.06)
COAUTHOR 3.6495 -3.9174
 (1.46) (1.03)
INSTITUTIONALCONNECTION 5.7737 8.2343
 (2.13) * (2.02) *
EDITORIALBOARD 17.4229 24.9913
 (3.32) ** (3.17) **
THANKEDITOR 7.0416 12.3878
 (2.06) * (2.43) *
CONSTANT -15.9770 -19.4802
 (3.41) ** (2.75) **
Log likelihood -1040.50 -1101.42

(a) Because of space limitations, the 15 JEL SUBJECT coefficients are
not shown. Absolute value of t-statistics in parentheses: ** =
significant at the 0.01 level; * = significant at the 0.05 level.

(1) One interesting result Blank found is that nearly half the referees of the double-blind papers could correctly identify the identity of the author.

(2) The limitations in the use of citations are discussed and dismissed in Leibowitz and Palmer (1988). They ask rhetorically, If an article is considered to be a high-quality scientific contribution, then why does it generate only a few citations?

(3) Richard Posner (Shepherd 1995, p. 5) notes, "Most of my economics papers have been published by journals edited by close friends (such as Ronald Coase and Bill Landes, when they edited the Journal of Law and Economics, or George Stigler and Sam Peltzman when they edited the JPE, or the Bell Journal when it was edited by Paul MacAvoy), and in many of these cases there weren't even formal submissions." Similarly, Ronald Coase (Shepherd 1995, p. 16) notes, "I have never found any difficulty in getting my articles published. I have either published in house journals (e.g., Economica) or the article was written as a result of a request and publication was assured."

(4) All citation figures were obtained from the Social Sciences Citation Index.

(5) The total number of citations to article i may, to some extent, depend on the vintage of an author's cumulative citation stock. The variable AUTHORCITES, 1970-89 may overstate the scientific worth of some authors who made their most important contributions in the distant past. Equation 1 was also estimated using the cumulative number of citations received by an author in the 10 years (1980-1989) prior to the publication of article i. The empirical results were identical to those reported in the paper.

(6) The JEL subject areas are A = General Economics and Teaching; C = Mathematical and Quantitative Methods; E = Macroeconomics and Monetary Economics; F = International Economics; G = Financial Economics; H = Public Economics; I = Health, Education, and Welfare; J = Labor and Demographic Economics; K = Law and Economics; L = Industrial Economics; M = Business Administration and Business Economics, Marketing, and Accounting; N = Economic History; O = Economic Development, Technological Change, and Growth; Q = Agricultural and Natural Resource Economics; R = Urban, Rural, and Regional Economics.

(7) If the first article in a journal issue was a presidential address or Nobel laureate lecture, the second article was considered to be the lead article in that issue.

(8) The data on paper presentations at seminars or workshops were obtained from the notes or acknowledgments section usually at the bottom of the first page, of each article. The date(s) the paper was presented was never provided by the author of the article.

(9) For each author, years of attendance at their Ph.D.-degree granting institution was matched, using the 1989 American Economic Review Survey of Members, with the graduate school attendance record as well as the years of university employment of the publishing journal's editor/coeditors.

(10)In order to determine if the empirical results are sensitive to the discrete nature of the dependent variable (number of article i citations = 0, 1, 2, ...), Equation 1 was also estimated using the negative binomial regression model. The negative binomial regression parameter estimates were strikingly similar to the empirical results reported in Tables 2 and 3.

(11) To determine if the results are sensitive to a few authors whose AUTHORCITES, 1970-89 are outliers, so large that they skew the empirical results, Equation 1 was reestimated using the log of AUTHORCITES, 1970-89. The empirical results reported in the paper were unaffected by this change.

(12) The experience of the author(s), measured by the number of years (as of 1990) since the receipt of the Ph.D. degree and a dummy variable if the article had a woman as an author, was also introduced into Equation 1. The quality of an article, as measured by the number of citations received, is not dependent on the experience or the gender of the author. The coefficients of the other variables are virtually identical to those reported in Table 2.

(13) Siow's (1991) theoretical model also suggests that if a reader's time is limited, there is an elastic supply of new scholars, and the cost of switching publications is small, then first impressions about an article are crucial and the optimal strategy of readers is not to go back and reevaluate prior research. However, his empirical results find that first impressions may not be that important.

References

Blank, Rebecca M. 1991. The effects of double-blind versus single-blind reviewing: Evidence from the American Economic Review. American Economic Review 81:1041-67.

David, Paul A. 1994. Positive feedbacks and research productivity in science: Reopening another black box. In Economics of technology, edited by Ove Granstrand. New York: North-Holland, pp. 65-89.

Diamond, Arthur M. 1989. The core journals of economics. Current Contents 21:4-11.

Folster, S. 1995. The perils of peer review in economics and other sciences. Journal of Evolutionary Economics 5:43-57.

Gans, Joshua S., and George B. Shepherd. 1994. How are the mighty fallen: Rejected classic articles by leading economists. Journal of Economic Perspectives 8:165-79.

Greene, William H. 1993. Econometric analysis 2. Englewood Cliffs, NJ: Prentice Hall.

Hamermesh, Daniel S. 1994. Facts and myths about refereeing. Journal of Economic Perspectives 8:153-63.

Hamermesh, Daniel S., and Sharon M. Oster. 2002. Tools or toys? The impact of high technology on scholarly productivity. Economic Inquiry 40:539-55.

Hawkins, Robert G., Lawrence S. Ritter, and Ingo Walter. 1973. What economists think of their journals. Journal of Political Economy 81:1017-32.

Laband, David N. 1986. Article popularity. Economic Inquiry 24:173-80.

Laband, David N. 1990. Is there value-added from the review process in economics?: Preliminary evidence from authors. Quarterly Journal of Economics 105:341-52.

Laband, David N., and Michael J. Piette. 1994a. Favoritism versus search for good papers: Empirical evidence regarding the behavior of journal editors. Journal of Political Economy 102:194-203.

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

Leibowitz, Stanley J., and John P. Palmer. 1988. Assessing assessments of the relative quality of economics departments. Quarterly Review of Economics and Business 28:77-88.

Mackie, Christopher D. 1998. Canonizing economic theory. Armonk, NY: M. E. Sharpe.

Merton, Robert K. 1968. The Matthew effect in science. Science 159:56-63.

Moore, William J. 1972. The relative quality of economics journals: A suggested rating system. Western Economic Journal 10:156-69.

Saner, Raymond D. 1988. Estimates of the returns to quality and coauthorship in economic academics. Journal of Political Economy 96:855-66.

Shepherd, George B. 1995. Rejected." Leading economists ponder the publication process. Sun Lakes, AZ: Thomas Horton.

Siow, Aloysius. 1991. Are first impressions important in academia? Journal of Human Resources 26:236-55.

Stigler, George J., and Claire Friedland. 1975. The citation practices of doctorates in economics. Journal of Political Economy 83:477-507.

Marshall H. Medoff Department of Economics, California Slate University, Long Beach, Long Beach, CA 90840, USA.

Received June 2002; accepted January 2003.