SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data
  5. 3. Regression analysis of the editorship effect
  6. 4. Testing the influence of editors on citations
  7. 5. Concluding Remarks
  8. Appendix
  9. References

Scholars may become journal editors because editors may generate more citations of their own works. This paper empirically establishes that a scholar's publications are more likely to be cited by papers in a journal that is edited by the scholar. We then test if editors exercise influence on authors to cite editors’ papers by either pressuring authors (“editor-pressure” hypothesis) or accepting articles with references to the editors’ papers (“editor-selection” hypothesis), by using the keyword analysis and the forward citation analysis, respectively. We find no evidence for the two hypotheses, which leaves self-selection as a possible cause for the editor effect. JEL classification: J01

Motivations des directeurs de revues. Les chercheurs peuvent devenir directeurs de revues parce que ce rôle peut engendrer plus de citations de leurs propres travaux. Ce mémoire montre empiriquement que les publications d'un chercheur sont davantage susceptibles d’être citées dans une revue qu'il dirige. On teste deux hypothèses à savoir si les directeurs exercent de l'influence sur les auteurs pour citer leurs travaux (soit en pressant les auteurs – hypothèse de pression du directeur – soit en acceptant les articles qui font référence à leurs travaux – hypothèse de sélection du directeur) à l'aide d'une analyse des mots-clés et des références (forward citation) respectivement. Les résultats ne supportent ni l'une ni l'autre des hypothèses, ce qui laisse l'auto-sélection du chercheur comme une cause possible de l'effet-directeur.

1. Introduction

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data
  5. 3. Regression analysis of the editorship effect
  6. 4. Testing the influence of editors on citations
  7. 5. Concluding Remarks
  8. Appendix
  9. References

Why do scholars often wish to be journal editors, despite the significant time and effort required to perform editorial duties? Substantial costs are associated with editorship, as their own research output drops precipitously when conscientious journal editors operate in that capacity (Shepherd 1995). With regard to the benefits from editorship, editors may be inspired by a commitment to serve the professional community or to produce a journal of the highest possible quality. Editors may be also motivated by a desire to be monetarily compensated for editorial duties, including course relief, although this compensation may not be substantial.

Editorship can generate non-pecuniary rewards, too: Editors can more easily acquire scholarly influence or reputation. Their own work may be cited more frequently in the journals they edit, either because they have solicited citations explicitly and/or implicitly, or because they have simply attracted submissions concerning the same types of subject as in their research. Publications or citations can be transformed into pecuniary rewards, as it has been fairly well documented that the number of publications or citations is related positively to tenure, promotion, and salaries, both in and out of the academic world.1 In this paper, we empirically evaluate the effect of editorship on citations in major economics journals. In particular, we attempt to determine whether the publications of an economist are more likely to be cited in articles published in a journal that the economist edits.

There are two categories of possible explanations for the effect of editorship on citations. First, an author may submit a paper to a journal whose editor is particularly interested in the paper's subject, and that paper is generally more likely to include references to the editor's publications. We refer to this explanation as the “self-selection” hypothesis.

The other explanation pertains to the influence of journal editors. Editors can exert influence on references in articles they edit in two ways. First, the author of an article may be either implicitly or explicitly persuaded by the journal editor to introduce references to the editor's papers. The author may voluntarily bend to tacit pressure and introduce those references, or the editor may explicitly suggest references to the editor's papers. We refer to this explanation as the “editor-pressure” hypothesis. Second, the editor may prefer to accept those papers that include references to the editor's papers. We refer to this latter explanation as the “editor-selection” hypothesis. We propose and carry out two distinct methods to test each of the two hypotheses.2

Table 1. Three hypotheses for the editor effect on citations
 DescriptionTest method
Self-selection hypothesisAn author may submit a paper to a journal whose editor is particularly interested in the paper's subject, and the paper is generally more likely to include references to the editor's publications. 
Editor-pressure hypothesisAn author may voluntarily bend to tacit pressure and introduce references to the editor's publications, or the editor may explicitly suggest references to the editor's papersSuppose an author publishes two papers, one in journal J and the other in journal K, both of which cite the same article written by an editor for journal J. If the paper in journal J was published under the editor's pressure, it is less closely related in content to the editor's article than the paper in journal K is to the editor's article.
Editor-selection hypothesisThe editor may prefer to accept those papers that include references to the editor's papers.Suppose an author publishes two papers in the same journal, one of which cites the journal editor's articles and the other does not. Under the editor-selection hypothesis, the former paper is likely to be of lower quality compared with the latter.

A number of studies in the literature have addressed the influence or power exercised by journal editors in the publishing industry. For example, Stigler, Stigler, and Friedland (1995) argued:

There is considerable discussion of the possible role of editors in steering disciplines, pushing or suppressing various lines of research. It is certainly true that the tastes of an editor can influence the contents of his journal. Davis Dewey, editor of the AER from 1911 to 1940, made that journal unreceptive to the growing technical rigor and formalization of economics, but the effect was a good deal stronger on the AER than on the profession. In effect, Dewey subsidized the rise of Econometrica. Similarly, Keynes's long reign at the Economic Journal probably discouraged its publication of econometric work, of which he was a skeptic, again, a subsidy to Econometrica, and his policies also helped the Review of Economic Studies. (344)

Shepherd (1995) provides other examples of editor discretion: Clower, when he was the editor of the American Economic Review, frequently accepted research papers for publication without submitting them for peer review (89). 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 (107). These stories are generally consistent with the point Kuhn (1970) made regarding the editor's power, namely, that journal editors set the research agenda, choose the issues to be tackled, establish the pertinent theoretical and analytical frameworks, and control the empirical methods utilized. Our paper makes a contribution to this body of literature in that it empirically assesses one dimension of the editor's influence, using citation count data.

Citation count has been utilized extensively in earlier studies as a quality measure of papers or authors in economics, as well as in other disciplines (Laband and Piette 1994a, Scott and Mitias 1996, Cherkashin et al. 2009, and Ellison 2010, for economics; Byrnes 1997, for psychology; Zsindely, Schubert, and Braun 1982, for chemistry; Sievert and Haughawout 1989, for education). Laband (1986), Laband and Piette (1994b), Medoff (2003), and Hudson (2007) attempted to determine the manner in which the number of citations of an economics article over the five- or ten-year period since its publication was related to various factors, most notably paper length and author productivity. In contrast to our paper, these papers have analyzed cross-sectional data on citations, rather than using panel information, and thus have ignored, for example, the profile of citations over the paper's age. Fok and Franses (2007) utilized panel data regarding annual citations to estimate a model similar to the Bass product diffusion model. However, this paper as well as those aforementioneddid did not attempt to determine the manner in which editorship affects citations to the editor's papers. Nor did those researchers test the “editor-pressure” or the “editor-selection” hypothesis.

We found one study in a discipline other than economics that addressed essentially the same question as ours, namely, the effect of editorship on citations of editors’ publications in the field of psychology. Lange and Frensch (1999) attempted to determine whether a researcher's assumption of editorship increased the frequency with which that researcher is cited in psychology journals. They demonstrated that during their editorship editors of American journals enjoyed an increased citation rate in their journal, but this was not true for editors of German journals. Because the average citation rates were compared prior to, during, and after editorship, they did not control for other possible factors that may have influenced the citation rate, in contrast to our paper, in which those factors are taken into account.

This paper is organized as follows. Section 'Data' describes the data on references in papers published in the top six general-interest economics journals. In Section 'Regression analysis of the editorship effect', we explain our empirical specifications and report the empirical findings from the regression analysis of citations. Section 'Keyword analysis for the editor-pressure hypothesis' describes the methodology for and the results from testing the “editor-pressure” hypothesis and the “editor-selection” hypothesis. For testing these two hypotheses, we utilize the keyword similarity analysis and the forward citation analysis, respectively. Our concluding remarks are provided in Section 'Concluding Remarks'.

2. Data

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data
  5. 3. Regression analysis of the editorship effect
  6. 4. Testing the influence of editors on citations
  7. 5. Concluding Remarks
  8. Appendix
  9. References

The following steps outline our database construction procedure. We initially collected the names of the top 100 economists from the list of “Top 5% Authors, as of January 2009 (average rank)” as provided by IDEAS (http://ideas.repec.org).3

We then acquired information on publications by these economists and all the citations to those publications up to March 2009 from Thomson's ISI Web of Science.4 We retrieved the information by searching the author name field for the names on our list. Because different individuals often share the same name, we verified the publications of each economist from the ISI Web of Science by checking each economist's curriculum vitae, which we downloaded from the internet. We were unable to locate the curriculum vitae of seven economists (Bernanke, Cox, Fernandez, Frey, Griliches, Nijkamp, and Smith) and thus we excluded them from our database. Our dataset included all varieties of scientific papers, such as full-length articles, short papers and proceedings, editorial materials, comments and reviews, and book chapters, and we excluded self-citations from our dataset. For each paper that was written by the economists in our database or that cited any papers written by our selected economists, the variables retrieved from ISI Web of Science included the author's name, paper title and type, pages, publication year, published journal and publisher, abstract, language, subject category, ISSN, and references.5 Ultimately, we compiled information on the entire set of publications written by our 93 economists (7,784 papers in total) and 136,476 papers authored by 101,295 people that cited the former set of papers.

We then determined, again using the curriculum vitae, the occasions on which the economists in our dataset served as editors with any of the top six general-interest journals: American Economic Review (AER), Econometrica (ECTA), Journal of Political Economy (JPE), Quarterly Journal of Economics (QJE), Review of Economic Studies (REStud), and Review of Economics and Statistics (REStat). We then assessed the “volume information” of each journal with JSTOR (http://www.jstor.org) to collect historical information on members of the editorial board and to verify the editorship of our economists. Members of the editorial board are, in general, composed of three types of editorship: editors, co-editors, and associate editors. These types are referred to by different titles in each journal, and the editorship structures have changed over time. Table 2 shows the composition of editors of the six journals since 1961. Because the roles of different types of editor differ, as does their influence, we grouped editors into two categories in our analysis: editors and co-editors as core editors, and associate editors as broadly defined editors. The list of economists who served as editors and their years of service are reported in Table 3.

Table 2. Composition of the editorial board in six journals
  Composition of the editorial board
JournalsYearEditorCo-editorAssociate Editor
NOTES
  1. *JPE had a position called associate editor only in 1973, which we regard as equivalent to an editor.

  2. ** In 1963, there was an acting managing editor for REStud.

  3. *** REStat had a managing editor only in 1962 and 1963.

AER1961–1982Managing editor Board of editors
 1983–1985Managing editorAssociate editorsBoard of editors
 1986-presentEditorCo-editorsBoard of editors
ECTA1961–1965Managing editorCo-editorsAssociate editors
 1966 – presentEditorCo-editorsAssociate editors
JPE1961-presentEditor*  
QJE1961–1974Editor  
 1975–1979EditorAssociate editors 
 1980–1984Board of editors  
 1985-presentBoard of editors Associate editors
REStud1961–1965Managing editor**American editoreconomic study society board
 1966–1967Joint managing editorAmerican editoreconomic study society board
 1968–1972Joint managing editorForeign editorEconomic Study Society, board
 1973–1979Joint managing editorForeign editorSociety forEconomic Analysis Ltd., board
 1980 – presentJoint managing editorForeign editorEditorial board
REStat1961–1963EditorAssistant to editor***Associate editor
 1964-presentEditorActing editor (1965–1967, 1974)Associate editor
Table 3. Editors in six journals
JournalEditors/Co-editorsAssociate editors
AERTaylor (1985–88), McCallum (1988–91), Milgrom (1990–93), Campbell (1991–93), Besley (1999–05), Card (2002–05), Gertler (2005–09)Becker (1968–71), Hall (1972–75), Stiglitz (1973–75), Feldstein (1974–79), Fama (1975–77), Gordon (1975–77), Barro (1976–79),Diamond (1979–81), Mishkin (1982–85), Akerlof (1983–91), McCallum (1985–88), Obsfeld (1987–90), Campbell (1990–91, 1994–96), Hamermesh (1990–94),Auerbach (1991–97), Christiano (1992–95), Milgrom (1993–00), Romer (1994–96), Grossman (1995–98), Gali (1996–01), Rebelo (1995–01), Romer (1996–02), Besley (1997–98), Woodford (1997–99), Reinhart (2001–05), Kehoe (2002–09), Rajan (2002–03)
ECTASims (1977–81), Deaton (1980–88), Card (1994–97), Fudenberg (1996–00), Blundell (1997–01), Acemoglu (2007–09)Hall (1972–78), Sims (1984), Engle (1976–81), Barro (1978–81), Deaton (1978–80), Dixit (1978–83), Phillips (1978–84), Taylor (1981–85), Hart (1984–87), Pesaran (1984–85), Tirole (1984–99), Fudenberg (1985–96), Milgrom (1987–90), Andrews (1988–09), Card (1988–93), Woodford (1990–93), Diebold (1994–97), Campbell (1996–98), Stock (2005–07)
JPEGordon (1970–73), Barro (1973–75, 1983–85), Lucas (1978–81, 1988–02), Heckman (1981–87), Becker (1993–94), Cochrane (1998–03) 
QJEBlanchard (1980–98), Freeman (1980), Summers (1984–90), Maskin (1984–90), Shleifer (1989–99), Katz (1991–09), Alesina (1998–04),Glaeser (1998–08), Barro (2004–09), Helpman (2008–09)Fudenberg (1984–89, 2008–09), Grossman (1984–03), Rogoff (1984–97), Akerlof (1985–03), Dornbusch (1985–09), Romer (1990–98), Alesina (1992–98, 2006–09), Borjas (1992–98), Kruger (1992–97), Hart (1995–03), Woodford (1995–00), Rajan (1998–02), Acemoglu (1999–02), Besley (2006–09)
REStudStiglitz (1968–76), Sims (1973–75), Nickell (1975–78), Maskin (1977–82), Hart (1979–83), Laffont (1980–85), Shiller (1982–84), Blanchard (1983–84), Tirole (1986–94), Campbell (91), Fudenberg (1993–96), Cochrane (1995–98)Feldstein (1965–67), Dixit (1970–75, 1979–80), Nickell (1973–74, 1979–87), Deaton (1975), Phillips (1975–80), Stiglitz (1978–80), Heckman (1982–85), Blundell (1983–84, 1988–93), Campbell (1990), Tirole (1994–96), Besley (1995–05), Gali (1998–00), Andrews (2005–09)
REStatStock (1992–02), Campbell (1996–02), Borjas (1998–05), Rodrik (2002–09), Acemoglu (2003–07)Feldstein (1968–02), Freeman (1970–02), Hall (1972–78), Sims (1975–77), Sargent (1977–81), Mankiw (1990–02), Christiano (1993–96), Diebold (1993–02), Poterba (1993–02), Rogoff (1993–08), Taylor (1993–96), Engle (1994–02), Heckman (1994–02), Eichenbaum (1996–02), Auerbach (1997–02), Borjas (1997–98), Rodrik (1997–01)

3. Regression analysis of the editorship effect

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data
  5. 3. Regression analysis of the editorship effect
  6. 4. Testing the influence of editors on citations
  7. 5. Concluding Remarks
  8. Appendix
  9. References

3.1. Model pecification

We test the effect of editorship on citation against publication-level panel data. The basic specification for our analysis of the editorship effect is a panel-data negative binomial model with cited-paper-specific random effects, which was developed in Hausman, Hall, and Griliches (1984):6

  • display math(1)

in which subscripts i, j, k, and t denote economist i (i = 1,…, 93),7 jth publication of the economist, journal k where citing papers appear (k = AER, ECTA, JPE, QJE, REStud, REStat), and year t, respectively. Note that we estimate this model individually for each of the six journals, since our regressors may have distinct influences on citations in each journal because each has a unique review process and varies in terms of the roles of editors. There are 7,784 papers written by 93 economists in our sample and subscript t for each paper runs from the year of its publication until 2009, which gives us a total of 139,972 paper-year observations in the regression for each of the six journals (i.e., the average age of the papers as of 2009 is about 17.98 years). E is the expectation operator and the dependent variable CITijkt is the total number of papers appearing in journal k in year t that cite the jth paper of economist i. We employ the negative binomial model because (i) our dependent variable is a count variable with non-negative values, frequently taking a value of zero, and (ii) the negative binomial model is preferred to the Poisson model, as we encounter the overdispersion problem in our data with the latter model (Wooldridge 2001). The cited-paper-specific random effects term included in our specification picks up variations in paper quality as well as writer quality.

Our key regressor is an indicator variable (EDITORikt) that equals 1 if economist i was an editor or co-editor in journal k in year t − 2. Note that the editorship variable is constructed to have a two-year lag. This is due to the submit-publish time lag. According to Ellison (2002), the average submit-accept time for papers in our six journals during the period 1970–99 is approximately 18 months, and we take two years for the submit-publish time lag in our analysis, with the extra six months for the accept-publish lag. For sensitivity we also conducted trials with other lags (e.g., lags of three years and of one year) and found the results to be qualitatively similar.8 As editors at different levels may have differing degrees of influence, we create a second measure of editorship (ASC_EDITORikt), which equals 1 if economist i was an associate editor in journal k in year t − 2. If a paper receives more citations because it is written by an editor, this influence should disappear after the editor steps down. To test this conjecture, we introduce into our analysis an additional indicator variable, AFTERikt, which equals 1 if economist i was an editor in year t − 5, but not in the next five years.

In order to measure the effects of age or labour market experience on citations, we include as a regressor in our specification EXPit, which indicates the years elapsed at year t since the publication of the first paper by economist i. As studies into the age-productivity profiles of economists (e.g.,, Oster and Hamermesh 1998) evidence an inverted-U shaped profile and a flattening of productivity, variable EXPit is entered in cubic forms in our model.

Fok and Franses (2007) demonstrated that the number of citations received by a paper assumes an inverted-U shape with a flat tail in relation to the age of the paper for several possible reasons: the paper becomes outdated, is replaced by better research, or becomes so well known that citations are no longer needed. To account for the effects of paper age, we include DURijt as a regressor, which measures the years elapsed at year t since the publication of economist i's jth paper.

As measures of the quality of a cited paper, we include three variables as regressors in our analysis: the logarithm of the number of pages (ln PAGEij), the number of papers appearing in all economics journals besides journal k that cite economist i's jth paper in the first five years after its publication (FIRST5ijk), and the number of papers appearing in all economics journals besides journal k in year t that cite economist i's jth paper (CURRijk't). We utilize the first regressor because high-quality research may require more exposition than does shorter, less substantive research, or may contain more sources to be cited (Laband and Piette 1994b; Hudson 2007). We attempt to capture the paper's quality in general with the second variable, as the lifetime citation count is related positively with the number of citations in the earlier years after publication.9 Readers’ interest in a paper may fluctuate over time, owing to paradigm shifts in economics, which generate fluctuations in citation counts over time. The third variable is included to account for these fluctuations.

To control for other sources of bias in citation, we include three indicator variables: NOBELit, which is 1 if economist i has received the Nobel Prize in economics within or prior to year t; CLARKit, which is 1 if economist i has received the John Bates Clark Medal within or prior to year t; and SAMEijk, which equals 1 if both economist i's jth paper and the papers that cite it are published in the same journal k. The first two variables are included because social reputation may perform a function in the choice of references by fellow researchers in their publications (see Stephan 1996). It has been previously argued in the literature that articles tend to cite more frequently papers that are published in the same journal, possibly because (i) the citing articles address the same subjects as those papers, or (ii) the editors may nudge authors into citing papers published in the journal they edit, in efforts to elevate the popularity or reputation of the journal. We include the third indicator variable to evaluate this notion.

The final regressor included in our benchmark specification is the logarithm of the total number of articles published in journal k in year t, ln ARTIkt, which reflects the supplyside of citations. A paper is expected to receive more citations in a journal that publishes more articles within a given year. Table 4 reports the description and summary statistics of the variables utilized in our analysis.

Table 4. Description and summary statistics
VariableDescriptionMeanSt. Dev.MinMax
CITijktTotal number of papers citing the jth paper of economist i in journal k in year t0.0350.227010
EDITORikt1 if economist i was an editor or co-editor in journal k in year t − 20.0150.12301
EXPitYears elapsed at year t since the publication of the first paper by economist i25.129.93153
DURijtYears elapsed at year t since the publication of economist i's jth paper12.038.72153
lnPAGEijLogarithm of the number of pages in the jth paper of economist i2.560.91805.27
FIRST5ijk’number of citations the jth paper of economist i receives in all economics journals besides journal k in the first 5 years after publication8.4916.580413
CURRijk'tnumber of citations the jth paper of economist i receives in all economics journals besides journal k in year t2.708.500283
NOBELit1 if economist i has received the Nobel prize in economics in year t or before0.0590.23601
CLARKit1 if economist i has received the John Bates Clark Medal in year t or before0.1580.36501
SAMEijk1 if economist i's jth paper is published in journal k0.0500.21901
AFTERikt1 if economist i was an editor or co-editor in year t − 5, but not in the next 5 years0.0030.0501
ASC_EDITORikt1 if economist i was an associate editor in journal k in year t − 20.0380.1901
lnARTIktlogarithm of total number of articles published in journal k in year t4.1980.5432.086.37
lnEDSIZEktlogarithm of the number of editors and co-editors in journal k in the first issue of year t1.500.56002.77

3.2. Empirical findings from the citation regressions

Table 5 shows our estimation results. We report the results for each of our six general-interest journals in each column of this table: AER, ECTA, JPE, QJE, REStud, and REStat in the first to sixth columns, respectively.

Table 5. Citation regressions (dependent variable: total number of citing papers)
 AERECTAJPEQJEREStudREStat
NOTE
  1. We use a negative binomial model with cited−paper-specific random effects. Value of t-statistics in parentheses.

EDITOR0.3710.0700.5320.8200.3830.728
 (4.24)(0.57)(6.89)(10.78)(3.73)(7.47)
EXP−0.0150.0340.0260.0500.0010.029
 (−0.89)(1.33)(1.28)(2.14)(0.04)(1.19)
EXP2−0.001−0.003−0.003−0.004−0.002−0.003
 (−1.28)(−2.55)(−3.12)(−3.36)(−1.50)(−2.72)
EXP36.6E−60.2E−40.4E−40.4E−40.1E−40.4E−4
 (0.71)(1.37)(3.38)(3.29)(0.94)(2.59)
DUR0.1250.2350.1710.1260.2590.299
 (11.79)(14.36)(12.25)(8.49)(15.20)(19.28)
DUR2−0.009−0.015−0.011−0.009−0.015−0.018
 (−13.55)(−14.06)(−12.33)(−9.77)(−14.29)(−18.31)
DUR30.2E−30.2E−30.2E−30.2E−30.2E−30.3E−3
 (12.53)(12.79)(11.19)(9.44)(12.51)(15.55)
lnPAGE0.9540.6540.7510.6270.7290.789
 (26.21)(14.24)(18.31)(16.11)(16.70)(19.78)
FIRST50.0160.0120.0150.0190.0170.014
 (16.27)(9.63)(12.35)(16.25)(12.96)(13.50)
CURR0.0220.0140.0200.0230.0230.020
 (25.58)(12.68)(18.19)(18.91)(17.33)(22.20)
NOBEL0.4420.8200.3080.5330.5150.382
 (6.32)(8.32)(3.25)(5.05)(4.96)(4.04)
CLARK0.2470.0060.2370.3460.0630.293
 (4.85)(0.08)(3.72)(5.53)(0.85)(4.49)
SAME1.4521.9801.2281.1501.4431.474
 (20.30)(16.53)(13.43)(12.11)(11.34)(8.82)
lnARTI0.0580.6801.6511.1500.9340.985
 (0.37)(9.78)(19.51)(10.20)(9.86)(10.98)
Obs.139972139972139972139972139972139972

In all journals with the exception of Econometrica, the estimated effects of EDITOR are both significant and positive, thereby indicating that the publications of a given editor (or co-editor) are more likely to be cited by articles published within that editor's journal. The magnitude of the EDITOR effects, however, appears to vary to a great extent across these journals. The estimated coefficients for EDITOR in QJE and REStat are more than twice as large as that measured in REStud. Interestingly, it is those in-house journals such as JPE, QJE, and REStat where it has become increasingly common for editors to reject a paper outright without soliciting comments from referees (Wu 2007) that evidence a profound editorship effect.

When we replace EDITOR with ASC_EDITOR in specification (1), the effect of the latter variable is either quantitatively smaller than that of the former (AER, QJE and REStat) or it becomes insignificant (REStud), as shown in panel A of Table 6.10 Econometrica is an exception, wherein the effect of ASC_EDITOR is significantly positive, as opposed to the non-existent effect of EDITOR. Our findings indicate that associate editors in general receive fewer additional citations than editors or co-editors, possibly because they have less influence in the publication process than do editors or co-editors.

Table 6. Additional sensitivity tests
 AERECTAJPEQJEREStudREStat
NOTE
  1. Value of t−statistics in parentheses. *JPE has no associate editor. The number of observations for each journal in panels (A) and (B) is the same as in Table 5.

(A) Effect of associate editors
 ASC_EDITOR0.1640.175*0.331−0.0870.199
 (3.19)(2.38)*(4.32)(−0.71)(3.14)
(B) Effect of editors after they are gone
 EDITOR0.3780.0810.5510.8220.3950.709
 (4.31)(0.65)(7.08)(10.77)(3.83)(7.27)
 AFTER0.2700.2360.3450.0910.221−41.532
 (1.25)(0.89)(2.07)(0.40)(1.34)(−0.00)
(C) Sample of journal editors only
 EDITOR0.3180.0040.5310.6770.3080.696
 (3.65)(0.03)(6.85)(8.70)(2.92)(6.95)
 Obs.584665846658466584665846658466
(D) Effect of the number of editors with a subsample of EDITOR = 1
 lnEDSIZE−1.048−0.388−0.268−0.3340.3810.855
 (−2.13)(−0.38)(−0.63)(−0.48)(0.77)(0.81)
 Obs.151711881574444217412497

The editorship effect is likely to last only during the time that the editors are on active duty. In order to assess this, we estimate a model with another regressor, AFTER, added to specification (1) in section 3.1. The results are reported in panel B of Table 6. We find evidence to support our hypothesis: the effects of AFTER are insignificant in all journals except for JPE.

Although our sample of top economists was not selected solely on the basis of citation counts, the total number of citations is one of the principal criteria for ranking them. This may cause a sample selection problem in our data. Determining whether this issue is actually a pertinent one, we utilize a subset of economists in our dataset who had at one time been editors or co-editors in any of the major six journals, and then we re-estimate our model. The effects of EDITOR in this case, as shown in panel C of Table 6, are qualitatively and quantitatively similar to those from the original estimation in Table 5.

One might argue that an individual editor may become less influential when more editors share editorial duties. In order to evaluate this hypothesis, the number of editors in a journal (EDSIZE) is added to our regression model, using only those observations of cited papers written by an editor (i.e., EDITOR = 1). Panel D of Table 6 shows no supporting evidence for this hypothesis: the effect of EDSIZE is insignificant in all journals except AER.

To understand more precisely the relationship between the labour market experience of a scholar (EXP) and the number of citations of the scholar's publication in each journal (CIT), in Figure 1 we show the estimated regression lines based on the results provided in Table 5.11 Two types of relationship patterns emerge: a monotonically decreasing or an inverted-U-shaped association. The former type is evident in AER, JPE, QJE, and REStud, whereas ECTA and REStat evidence the latter type, in which the number of citations peaks in the early career of economists (with about five to seven years of experience), and then diminishes sharply afterward. This result is consistent with the findings in the literature. Lehman (1953) reported an early peak in productivity over career in a variety of scientific and artistic fields, and Levin and Stephan (1992) provided clear evidence for a decline in productivity in science, even after careful attempts to account for individual and cohort heterogeneity. Oster and Hamermesh (1998) have determined that economists’ productivity (as measured by publications) over their careers declined quite sharply with age.

image

Figure 1. Labour market experience profile

Download figure to PowerPoint

The regression results in Table 5 support our conjecture that the number of citations received by a paper assumes an inverted-U shape with regard to the age of the paper (DUR). According to the results in Table 5, the number of citations reaches a peak in the eight to twelve years after a paper was published (see Figure 2).

image

Figure 2. Duration profile

Download figure to PowerPoint

The logarithmic value of the page count of a cited paper (ln PAGE), which may reflect paper quality, because more substantive research can have greater exposition or can contain more sources to be cited, is shown in Table 5 to be significantly positively associated with citations. Editorial restrictions on the number of pages are not as tight in some journals, and this variable may just reflect the difference in the propensity to be cited across journals where a cited paper appeared. To entertain this possibility, we introduced as additional regressors in the basic specification the interaction terms of ln PAGE with seven dummy variables for journals where cited papers appeared (AER, ECTA, JPE, QJE, REStud, REStat, and all the rest of journals). We found that the number of pages is positively associated with citations among papers published in each of the seven journal categories (results are not shown to save space).

Our results in Table 5 also show papers that received more citations in the first five years of publication (FIRST5), another measure of paper quality, are more frequently cited over the entire lifespan of the papers. In an effort to account for yearly variations in citation counts due to fluctuations in research interest in a paper, we relate the number of citations in journal k with the number of citations the paper receives in all economics journals besides journal k in the same year (CURR). Our results demonstrate that citation count in other journals is associated positively with that in journal k.

Table 5 demonstrates that Nobel Prize winners receive more citations after winning the prize and John Bates Clark Medal winners also receive more citations in most journals except ECTA and REStud. This implies that the choice of references to include appears to be affected by professional reputation. Interestingly, the coefficient estimates on CLARK are quantitatively smaller than those on NOBEL, which is consistent with the conventional view that the Nobel Prize is better reputed.

Our findings, as shown in this table, confirm the common notion that articles tend to cite papers published in the same journal more frequently: variable SAME significantly positively affects CIT. As anticipated, the total number of articles published in a journal within a given year (ARTI) is shown in this table to have a positive effect on citations in all journals except for AER.

It is plausible that a group of papers will never be cited in a particular journal, perhaps because its research topic or its type (e.g., whether a paper is a full-length article, a book review, or a proceedings paper) is not well suited for the journal. This group of papers, hence, always has zero citation counts. On the other hand, a paper that could be cited but has not been, owing to its poor quality, would not be included in this group. The zero-inflated negative binomial (ZINB) model (Greene 1994) assumes that the probability of a paper belonging to the first group (with zero counts at all times) is ϕ and this probability is determined by characteristics of the paper. Citation counts are thus generated by two processes in this model. Zero citation counts arise with probability ϕ from a binary (logit or probit) process, and both zero and positive counts can be generated by a negative binomial process (see Long 1997 for the discussion of the ZINB model).

As regressors for the binary process that characterize the attributes of a paper, we include the logarithm of the number of pages (lnPAGE) and dummy variables for various types of paper (see the note in Table 7 for the types considered). We assume that ϕ is determined by a logit model.

Table 7. Zero-inflated negative binomial model
 AERECTAJPEQJEREStudREStat
NOTE
  1. Value of t-statistics in parentheses. To save space, we do not report coefficients associated with EXP and DUR in the negative binomial model of this table. Dummy variables included in the logit model: T_ARTI ( = 1 if a paper is a full-length article), T_BOOK ( = 1 if it is a book review), T_EDIT ( = 1 if it is editorial material), T_LETT( = 1 if it is a letter), T_NOTE ( = 1 if it is a note), T_PROC ( = 1 if it is a proceedings paper).

Negative binomial      
 EDITOR0.5710.7511.0111.0400.7490.900
 (5.68)(4.40)(9.73)(13.67)(6.31)(9.24)
 lnPAGE0.3520.5870.161−0.0880.0090.301
 (4.17)(3.25)(1.49)(−1.27)(0.07)(3.47)
 FIRST50.0120.0070.0090.0140.0100.015
 (6.37)(3.98)(5.50)(8.11)(5.29)(7.35)
 CURR0.0600.0430.0480.0470.0510.051
 (11.86)(7.17)(11.21)(9.95)(9.76)(10.24)
 NOBEL0.5691.3080.7260.6720.9380.882
 (5.72)(8.98)(6.51)(5.10)(6.87)(6.92)
 CLARK0.251−0.0510.1930.364−0.0710.254
 (3.87)(−0.51)(2.56)(4.95)(−0.71)(3.32)
 SAME1.0581.7071.0921.0371.1481.357
 (16.57)(21.74)(15.30)(14.09)(12.27)(12.65)
 lnARTI0.2040.9051.8961.4101.0071.106
 (1.12)(10.90)(18.88)(11.33)(8.78)(9.71)
Logit      
 lnPAGE−1.1130.708−0.927−1.285−1.132−2.431
 (−3.53)(0.73)(−3.66)(−7.36)(−3.68)(−2.65)
 T_ARTI−2.360−2.452−1.605−1.723−2.102−5.173
 (−4.98)(−2.61)(−3.29)(−3.71)(−6.74)(−2.91)
 T_BOOK1.6054.20421.4661.5802.078−1.345
 (2.08)(1.61)(39.65)(1.80)(1.94)(−0.78)
 T_EDIT0.2421.4720.5130.6551.421−2.785
 (0.41)(0.80)(0.69)(0.95)(1.71)(−1.25)
 T_LETT−1.6512.605−1.014−1.048−0.751−5.495
 (−2.32)(1.14)(−1.53)(−1.81)(−1.17)(−1.64)
 T_NOTE−1.201−1.785−0.733−0.836−1.511−4.946
 (−1.83)(−0.72)(−1.27)(−1.32)(−3.11)(−1.66)
 T_PROC−0.8590.019−0.526−1.231−0.386−5.076
 (−1.71)(0.02)(−1.02)(−2.19)(−1.03)(−2.04)
 Obs.139972139972139972139972139972139972

The upper panel of Table 7 reports the estimation results from the negative binomial model, and the lower panel presents the results from the logit model. The effects of EDITOR on citations in the negative binomial model are positive and significant in all journals, including even Econometrica. Its effects are more pronounced in this model than in those in Table 5.

The estimation results from the logit model indicate that longer papers have lower odds of being included in the group with zero citation counts at all times in all journals. Full-length articles are less likely to be never cited in all journals of our sample compared with other types of papers.

4. Testing the influence of editors on citations

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data
  5. 3. Regression analysis of the editorship effect
  6. 4. Testing the influence of editors on citations
  7. 5. Concluding Remarks
  8. Appendix
  9. References

4.1. Keyword analysis for the editor-pressure hypothesis

A paper whose author has been persuaded, either implicitly or explicitly, by a journal editor to include references to the editor's papers would clearly not have included references to the editor's papers if the paper had been published in another journal. Therefore, it appears likely that papers that cite the editor's papers as the result of applied editorial pressure would not be closely related to the editor's papers in terms of content. For example, Campbell had two papers published in 1993, one in QJE and the other in REStud, both of which cite the same paper written by Shleifer who was an editor of QJE in 1989–99. If Campbell's QJE paper was published under the editor's pressure, this paper is less closely related in contents to the Shleifer's paper than is his REStud paper. We thus test the “editor-pressure” hypothesis by comparing the closeness of Campbell's QJE paper and Shleifer's, relative to the pair of Campbell's REStud paper and Shleifer's, in terms of common keywords in the abstracts of each pair.

The precise steps for our keyword analysis can be illustrated via the following example. In the first step, we select a paper published in QJE during the period 1991–2001, which cites a paper by Shleifer, who was an editor for QJE in 1989–99. Note that we have a two-year lag, owing to the submit- publish time lag (see Section 'Article quality analysis for the editor-selection hypothesis' for details). We refer to the citing paper as paper A and the cited paper (written by the editor) as paper B. In the second step, we find another paper (called paper C) written by the same author of paper A,12 which (i) appeared in a journal other than QJE in the same year as A, or a year prior to or after that year, and (ii) cites B.13 We now have two pairs of papers: A and B (pair I), and B and C (pair II). In the third step, we utilize a computer algorithm for part-of-speech tagging to collect nouns and adjectives from the abstracts of papers A, B, and C.14 In the fourth and final step, we calculate the Jaccard Similarity Coefficients for pairs I and II.

The Jaccard Coefficient is a statistic utilized to compare the similarity of two objects—in our case, a pair of papers. Suppose that papers A and B have N number of keywords that appear in any of the two papers and M number of keywords that appear in both. The Jaccard Coefficient is expressed as 100M/N, which is 100 if two papers have the exact same keywords, and 0 if they have no common keyword.15 Under the “editor-pressure” hypothesis, the Jaccard Coefficient for pair I is expected to be smaller than that for pair II.

In this analysis, possible candidates for the aforementioned “paper A” are those papers that cite any article by an editor of the six journals (only among those in our sample of 93 economists) and that are published during the time that the editors are on active duty. For each candidate for “paper A,” we may have more than one paper as a candidate for “paper C”. In that case, we randomly selected one paper to designate as “paper C.”

Panel A of Table 8 reports the mean Jaccard Coefficients of pair I and II for each journal, wherein the coefficients are calculated with keywords expressed in nouns and adjectives. This shows that the mean Jaccard Coefficient of pair I is greater than that of pair II for AER, ECTA, JPE, QJE, and REStud. The only case in which the opposite is true is REStat, and the difference between the two coefficients is 0.76445. With the assumption that the difference between the two Jaccard Coefficients is a normal random variable with mean zero, the t-statistic for testing the null hypothesis that the two Coefficients are the same in REStat is 0.7975 (with d.f. = 50), and thus we are unable to reject the null hypothesis with a significance level of 10%. In other words, we find no evidence to support the “editor-pressure” hypothesis—that authors are implicitly or explicitly persuaded by an editor of our sample journals to add references to the editor's papers. As part of our sensitivity analysis, we calculate the Jaccard Coefficients using keywords in nouns only, and report the results in panel B of Table 8, which warrants the same conclusion as from panel A.

Table 8. Testing the “Editor-Pressure” hypothesis: keyword analysis
 Average Jaccard Coefficient Observations
JournalPair IPair IIDifference (I – II)I ≥ III < II
NOTE
  1. We refer to a paper that cites an editor's paper as paper A, while the editor's paper is denoted as paper B. We refer to a paper by the same author of A in another journal that also cites B as paper C. Papers A and B comprise pair I and papers B and C comprise pair II.

(A) Keywords in nouns and adjectives   
 AER10.550289.178841.371441412
 ECTA11.4997710.851180.64860812
 JPE10.861679.808821.052863425
 QJE9.925258.302051.623194429
 REStud12.1465911.967260.179331112
 REStat8.933779.69822−0.764452625
(B) Keywords in nouns only   
 AER11.9804810.665491.314991610
 Econometrica13.8445912.92150.923101010
 JPE12.5708111.145491.425313326
 QJE11.594759.0952262.499534429
 REStud14.4082413.724840.68341149
 REStat10.7728911.10606−0.333172427

Table 8 also reports the numbers of cases for each journal wherein the Jaccard Coefficient of pair I is greater than that of pair II, or vice versa. In our sample journals, we find that the cases in which the coefficient of pair I is greater than that of pair II are more numerous or the difference of cases is marginal when the coefficient of pair II is greater, which corroborates our conclusion.

Since the actual editor who handled “paper A” in our sample cannot be identified, our sample for “paper A” may include papers edited not by the editor in question but by another editor in the same journal, which may bias the estimated Jaccard Coefficient of pair I. However, the bias is likely to be a downward bias because papers not assigned to the editor in question should be less related to the editor's research area and thus the editor's paper that is cited. This bias thus cannot invalidate the conclusion from our keyword analysis.

4.2. Article quality analysis for the editor-selection hypothesis

If an editor prefers to accept articles that include references to the editor's papers, the articles with such references are likely to be of lower quality in comparison to other articles published in the same journal without such references. Suppose we have two papers by the same author published in the same journal, one of which cites the journal editor's paper (called paper P) and the other does not (paper Q). We test the “editor-selection” hypothesis by comparing the quality of the two papers in terms of the total number of citations that each paper receives in the first five years after its publication. In order to control for the age effect of authors and for fluctuations in citation counts over time due to paradigm shifts, we select papers P and Q from our sample, one of which was published within a two-year span of the other's publication.16

The first two columns of Table 9 report the average number of citations of papers P and Q for each journal, respectively. The table shows that the average citation number of paper P is greater than that of paper Q for AER, JPE, and QJE, which does not lend support to the “editor-selection” hypothesis. On the other hand, the former average is less than the latter for ECTA, REStud, and REStat. With the assumption that the difference between the two averages is a normal random variable with mean zero, the t-statistics for the null hypothesis that the two averages are the same for each of these three journals are 1.2297 (with d.f. = 12), 0.4588 (with d.f. = 2), and 0.5780 (with d.f. = 16) for ECTA, REStud, and REStat, respectively. This result indicates that we cannot reject the null hypothesis with a significance level of 10%. We therefore find no evidence to support the “editor-selection” hypothesis.

Table 9. Testing the “Editor-Selection” hypothesis: article quality analysis
 Average number of citations Observations
JournalPaper PPaper QDifference (PQ)PQP < Q
NOTE
  1. We refer to a paper that cites an editor's paper as paper P and a paper by the same author of P that is published in the same journal but does not cite any editor's papers as paper Q.

AER38.6730820.5769218.096152626
ECTA14.3076921.23077−6.92307767
JPE26.9615422.846154.1153853319
QJE43.0000029.0833313.916672721
REStud1.000001.666667−0.66666712
REStat9.94117612.23529−2.29411898

Table 9 also reports the numbers of cases for each journal wherein the average citation number of paper P is greater than that of paper Q, or vice versa. In our sample journals, we find that the cases in which the average citation number of paper P is greater than that of paper Q are more numerous or the difference of cases is marginal when the average citation number of paper Q is greater, which is consistent with our conclusion of no editor-selection.

5. Concluding Remarks

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data
  5. 3. Regression analysis of the editorship effect
  6. 4. Testing the influence of editors on citations
  7. 5. Concluding Remarks
  8. Appendix
  9. References

Our regression analysis provides evidence to suggest that economists tend to receive more citations when they serve as journal editors. Our results demonstrate that the numbers of citations are significantly higher for editors in general-interest economics journals, even after we control for the heterogeneity of cited papers, cited economists, and journals. However, the effects of editorship on citations are not uniform across journals. Our findings were robust to various sensitivity analyses we conducted, including models that distinguish the editor effects during and after editorship and those that take into consideration the potential sample selection issue or account for the prevalence of zero counts in the data. Our finding of a non-negative effect of editorship on citations corroborates our thesis that one of the motivations for editors may be to attain greater scholarly influence.

In light of our observation that editors’ papers are cited with greater than normal frequency, we attempt to determine whether this phenomenon is attributable to the influence of editors, which may be exerted either by pressuring authors to include references to editors’ papers or by accepting articles that include references to their papers. According to our findings from the two tests, there is no evidence to support either the “editor-pressure” hypothesis or “editor-selection” hypothesis in all journals of our dataset; this finding leads us to conclude that “self-selection” may be a more significant cause of the effect of editorship on citations as long as these three hypotheses exhaust possible explanations for the editor effect on citations.

Our estimation results are not only statistically significant, but also quantitatively significant. If the editor of JPE in our sample served for one more year at the end of his tenure, the number of papers in JPE citing that editor's publications would increase by 70.2%.17 Because the average annual number of citations in JPE to a paper by an economist who had stepped down from the JPE editorial board in the previous year is 0.1504, and the average accumulated number of publications for the economist in our sample is 59.83, this increase translates to an increase of 6.3 citations ( = 0.702*0.1504*59.83), from 9.0 citations per year to 15.3, in the number of citations in JPE for one year. In the same manner, we can demonstrate that the corresponding increases in citations with one more year of editorship are 2.2 citations for AER, 0.2 for ECTA, 5.1 for QJE, 1.2 for REStud, and 6.4 for REStat.

Appendix

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data
  5. 3. Regression analysis of the editorship effect
  6. 4. Testing the influence of editors on citations
  7. 5. Concluding Remarks
  8. Appendix
  9. References

A. List of Top 100 economists in alphabetical order

Acemoglu, D., Aizenman, J., Akerlof, G.A., Alesina, A., Andrews D.W.K., Auerbach, A., Barro, R.J., Baum, C.F., Becker, G.S., Berger, A.N., Bernanke, B.S., Besley, T.J., Blanchard, O., Blundell, R., Bollerslev, T., Borjas, G., Caballero, R.J., Campbell, J.Y., Card, D.E., Christiano, L., Cochrane, J.H., Cox, N., Deaton, A.S., Diamond, P.A., Diebold, F.X., Dixit, A., Dornbusch, R., Edwards, S., Eichenbaum, M., Eichengreen, B.J., Engle, R.F., Fama, E.F., Feldstein, M.S., Fernandez, P, FrankelJ.A., French, K.R., Freeman, R.B., Frey, B.S., Fudenberg, D., Gali, J., Gertler, M.L., Glaeser, E.L., Gordon, R.J., Granger, C.W.J., Griliches, Z., Grossman, G., Hall, R.E., Hamermesh, D., Hansen, L.P., Hart, O.D., Heckman, J.J., Helpman, E., Jovanovic, B., Katz, L.F., Kehoe, P., King, R.G., Kruger, A.B., Krugman, P.R., La Porta, R., Laffont, J.J., Levine, R., Lopez-de-silance, F., Lucas, R.E., Mankiw, N.G., Maskin, E.S., McCallum, B., Merton, R.C., Milgrom, P., Mishkin, F., Murphy, K.M., Nickell, S.J., Nijkamp, P., Obsfeld, M., Pesaran, M.H., Phillips, P.C.B, Poterba, J., Prescott, E.C., Rajan, R.G., Ravallion, M., Rebelo, S.T., Reinhart, C.M., Rodrik, D., Rogoff, K.S., Romer, D., Romer, P.M., Rose, S.J., Sargent, T.J., Shiller, R.J., Shleifer, A., Sims, C., Smith, B.D., Stiglitz, J.E., Stock, J.H., Summers, L.H., Svensson, L.E.O., Tabellini, G., Taylor, J.B., Tirole, J., Turnovsky, S.J., Woodford, M.

B. Methodology in the economist ranking

A public-access database, RePEc (Research Papers in Economics, http://repec.org) computes 32 different rankings of economists according to 32 different criteria. Those criteria include: seven criteria for number of works, twelve criteria for number of citations, H index (author has written h papers that have each been cited at least h times), two criteria for number of registered citing authors, six criteria for number of journal pages, two criteria for number of abstract views, and two criteria for number of downloads through RePEc service over the previous twelve months. The average rank is computed by taking a harmonic mean of all the above rankings except three (one ranking on number of works, the best rank, and the worst rank).

  1. 1

    The effect on salary, which is an issue of consequence for editors, who are mostly senior researchers, has been extensively studied. Hamermesh, Johnson, and Weisbrod (1982) showed that citation is a more important determinant of dispersion in salaries than either numbers of publications or the status of one's research outlets. Diamond (1986) reported that citation count is a significantly positive determinant of earnings. Sauer (1988) estimates the earnings function to find that salary is related positively with citations, as well as publications. Hamermesh and Pfann (2009) postulate that scholars maximize reputation, which is determined by the number of publications and their quality (citations) and show that quality ranking is the most important factor. For the role of publications in determining salary, also refer to Katz (1973), Tuckman and Leahey (1975), and Hansen, Weisbrod, and Strauss (1978).

  2. 2

    The description of and the test methods for the three hypotheses are summarized in Table 1.

  3. 3

    The list of names in alphabetical order and the description of the methodology utilized for ranking are reported in Appendices A and B, respectively.

  4. 4

    Thomson's ISI Web of Science is a web database that helps scholars quickly find, analyze, and share information in the natural sciences, social sciences, arts, and humanities. It offers four types of citation databases and two types of chemical databases: Social Science Citation Index, Science Citation Index Expanded, Arts & Humanities Citation Index, Conference Proceedings Citation Index, Index Chemicus, and Current Chemical Reactions.

  5. 5

    ISI Web of Science provides keywords for each paper, but we find this information less than useful, because it was missing for a significant number of papers.

  6. 6

    Hausman, Hall, and Griliches (1984) assume that the Poisson parameter λit (i for cross-section units and t for time periods) follows a gamma distribution with parameters (γit, δi) where γit = exp (Xit β), Xit is a vector of regressors, δi = 1/exp (εi), and εi is the random-effects term. In order to easily integrate δi out of the probability density, they further assume that the ratio δi /(1 + δi) is distributed as a beta random variable with parameters (a, b), where the mean is E [δi /(1 + δi)] = a /(a + b) with variance V [δi /(1 + δi)] = ab /(a + b + 1)(a + b)2.

  7. 7

    Hereafter we refer to 93 economists in our sample who “wrote” cited papers as “economists” while those who “authored” citing papers as “authors” for clarification.

  8. 8

    Economists more cited in the past may be more likely to become editors, which implies the possibility of reverse causality. In our analysis, however, we correlate the current status of editorship not with the accumulated number of citations in the past, but with the number of citations two years ahead, which alleviates the reverse causality and the endogeneity problem.

  9. 9

    For example, the correlation between the lifetime citation count and the first five-year citation count for those papers published in 1980 from our sample – there are 133 of them – is 0.86.

  10. 10

    Note that JPE has no associate editor.

  11. 11

    Three coefficients associated with EXP, EXP2, and EXP3 are jointly significant and different from zero in all journals, even though some of them are individually insignificant.

  12. 12

    By “same author” we mean that papers A and C share at least one co-author.

  13. 13

    We consider the three-year span to acquire more matched pairs of papers for our analysis. We also tried the five-year span, which generated the same result.

  14. 14

    The program we use for part-of-speech tagging is the “Stanford Log-Linear POS tagger.” See http://nlp.stanford.edu/software/tagger.shtml for a description of this program.

  15. 15

    If a keyword appears multiple times, we treat each appearance as a separate word.

  16. 16

    If our assumption that forward citations reflect the quality of a paper does not hold, this approach may not be able to exclusively test the editor-selection hypothesis. Suppose, for example, a low-quality paper that would not have been published without the editor”s favouritism happens to be in a particular domain wherein the editor does research so that it cites publications by the editor. If papers that appear in that domain receive more citations due to its popularity, paper P can receive more citations in spite of the fact that the paper is of low quality. We are grateful to an anonymous reviewer for bringing this issue to our attention.

  17. 17

    This calculation is based on the estimated coefficient for EDITOR in column 3 of Table 5, assuming that all the regressors except EDITOR remain constant: exps (0.532*EDITOR)|EDITOR = 1 / exp (0.532*EDITOR)|EDITOR = 0 − 1 = 0.702.

References

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data
  5. 3. Regression analysis of the editorship effect
  6. 4. Testing the influence of editors on citations
  7. 5. Concluding Remarks
  8. Appendix
  9. References
  • Byrnes, James (1997) “Explaining Citation Counts of Senior Developmental Psychologists.” Developmental Review 17, 6277
  • Cherkashin, Ivan, Svetlana Demidova, Susumu Imai, and Kala Krishna (2009) “The Inside Scoop: Acceptance and Rejection at the Journal of International Economics.” Journal of International Economics 77, 12032
  • Diamond, Arthur Jr (1986) “What Is a Citation Worth?Journal of Human Resources 21, 20015
  • Eillison, Glenn (2002) “Slowdown of Economics Publishing Process.” Journal of Political Econom y 110, 94793
  • Eillison, Glenn (2010) “How Does the Market Use Citation Data? The Hirsch Index in Economics.” NBER Working Paper No. 16419
  • Fok, Dennis, and Philip H. Franses (2007) “Modeling the Diffusion of Scientific Publications.” Journal of Econometrics 139, 37690
  • Greene, WilliamH. (1994) “Accounting for Excess Zeros and Sample Selection in Poisson and Negative Binomial Regression Models.” Working Paper No. EC-94–10, New York University
  • Hamermesh, Daniel S., and Gerard A. Pfann (2009) “Markets for Reputation: Evidence on Quality and Quantity in Academe.” NBER Working Paper No. 15527
  • Hamermesh, Daniel S., George E. Johnson, and Burton A. Weisbrod (1982) “Scholarship, Citations and Salaries: Economic Rewards in Economics.” Southern Economic Journal 49, 47281
  • Hansen, W. Lee, Burton A. Weisbrod, and Robert P. Strauss (1978) “Modeling Earnings and Research Productivity of Academic Economists.” Journal of Political Economy 86, 72941
  • Hausman, Jerry A., Bronwyn H. Hall, and Ziv Griliches (1984) “Econometric Models for Count Data with an Application to the Patents R&D Relationship.” Econometrica 52, 90938
  • Hudson, John (2007) “Be Known by the Company You Keep: Citations – Quality or Chance?Scientometrics 71, 238
  • Katz, David A (1973) “Faculty Salaries, Promotions, and Productivity at a Large University.” American Economic Review 63, 46977
  • Kuhn, Thomas (1970) The Structure of Scientific Revolutions. Chicago: University of Chicago Press
  • Laband, David N. (1986) “Article Popularity.” Economic Inquiry 24, 17380
  • Laband, David N., and Michael J. Piette (1994a) “The Relative Impacts of Economics Journals: 1970–1990.” Journal of Economic Literature 32, 64066
  • Laband, David N., and Michael J. Piette (1994b) “Favoritism versus Search for Good Papers: Empirical Evidence Regarding the Behaviour of Journal Editors.” Journal of Political Econom y 102, 194203
  • Lange, L., and P.A. Frensch (1999) “Gaining Scientific Recognition by Position: Does Editorship Increase Citation Rate?Scientometrics 44, 45986
  • Lehman, Harvey C. (1953) Age and Achievement. Princeton: Princeton University Press
  • Levin, Sharon G., and Paula E. Stephan (1992) Striking the Mother Lode in Science: The Importance of Age, Place and Time. New York: Oxford University Press
  • Long, J. Scott (1997) Regression Models for Categorical and Limited Dependent Variables. Thousand Oaks: Sage Publications
  • Medoff, Marshall H. (2003) “Editorial Favoritism in Economics?Southern Economic Journal 70, 42534
  • Oster, Sharon M., and Daniel S. Hamermesh (1998) “Aging and productivity among economists,” Review of Economics and Statistics 80, 1546
  • Sauer, Raymond D. (1988) “Estimates of the Returns to Quality and Coauthorship in Economic Academia.” Journal of Political Economy 96, 85566
  • Sievert, MaryEllen, and Mary Haughawout (1989) “An Editor's Influence on Citation Patterns: A Case Study of Elementary School Journal.” Journal of the American Society for Information Science 40, 33441
  • Scott, Loren C., and Peter M. Mitias (1996) “Trends in Ranking of Economics Departments in the U.S.: An Update,” Economic Inquiry 34, 378400
  • Shepherd, George B. (1995) Rejected: Leading Economists Ponder the Publication Process. Sun Lakes, AZ: Thomas Horton
  • Thomson Rfeuters, Web of Knowledge (1957–2009) Social Science Citation Index. Philadelphia: Institute for Scientific Index
  • Stephan, Paula E. (1996) “The Economics of Science.” Journal of Economic Literature 34, 1199235
  • Stigler, George J., Stephen M. Stigler, and Claire Friedland (1995) “The Journals of Economics.” Journal of Political Economy 103, 33159
  • Tuckman, Howard P., and Jack Leahey (1975) “What Is an Article Worth?Journal of Political Economy 83, 95167.
  • Wooldridge, Jeffrey M. (2001) Econometric Analysis of Cross-Sectional and Panel Data. Cambridge, MA: MIT Press
  • Wu, Stephen (2007) “Recent Publishing Trends at the AER, JPE and QJE.” Applied Economics Letters 14, 5963
  • Zsindely, S., A. Schubert, and T. Braun (1982) “Citation Patterns of Editorial Gatekeepers in International Chemistry Journals.” Scientometrics 4, 6976