Journal of Economic Literature 2024, 62(1), 269–293

https://doi.org/10.1257/jel.20221653

Publishing Economics: How Slow?

Why Slow? Is Slow Productive?
How to Fix Slow?†
Aboozar Hadavand, Daniel S. Hamermesh, and Wesley W. Wilson*

Economics publishing proceeds much more slowly than in the natural sciences, and
more slowly than in the other social sciences and finance. It is relatively even slower at
the extremes. Much of the lag, especially at the extremes, arises from authors’ dilatory
behavior in revisions. Additional rounds of resubmissions at top economics journals
are related to additional citations; but conditional on resubmission, the delays are
unrelated to greater scholarly attention. We offer several proposals for speeding pub-
lication, including no-revision policies such as Economic Inquiry’s, the use of “cascad-
ing referee reports,” limits on authors’ time revising, and limits on editors waiting for
dilatory referees. (JEL A11, B29)

1. Introduction that had occurred in the publishing process

in ­economics. Today the difficulties are well

T he slowness of publishing in econom-

ics was pointed out by Ellison (2002),
although scholars who had been active in
known and have been discussed by many
editors (e.g., in Szenberg and Ramrattan
­
2014). In this study we first provide evidence
the profession for at least a quarter of a cen- on how the publishing process—in terms of
tury were by then very aware of the changes the lags involved—compares to that in the

*Hadavand: Minerva University. Hamermesh: Uni-

versity of Texas at Austin and IZA. Wilson: University of APSR, and especially Economic Inquiry, and Lea-Rachel
Oregon. We thank Amy Ando, Belinda Archbong, Garry Kosnik for the data underlying figure 3. Most important,
Barrett, George Borjas, André Burgstaller, Colin Cam- we thank the editors of the three “top five” journals, whose
eron, Steven Deller, Co-Pierre Georg, Christopher Gibbs, data underlie much of the analysis, for approving this proj-
Sarah Hamersma, Campbell Harvey, Van Kolpin, Kevin ect, and their editorial assistants who compiled and pro-
Lang, Andrew Leigh, Derek Neal, Andrew Oswald, Glen vided the data. Under confidentiality agreements with the
Waddell, and participants in seminars at several universi- journals, most of the data are not available publicly, nor
ties for helpful comments. Shachar Berkowitz-Regosin and can we publicize the identities of those journals. Dummy
Yonah Hamermesh gave excellent research assistance, and programs and some of the data are available in Hadavand,
IZA provided funding. We also thank the many authors Hamermesh, and Wilson (2024).
†
who responded to email surveys, the people who provided Go to https://doi.org/10.1257/jel.20221653 to visit the
salary information, the editorial staffs of the AEA, the article page and view author disclosure statement(s).

269
270 Journal of Economic Literature, Vol. LXII (March 2024)

“hard sciences” (very much slower) and in Laband 1990 provided subjective evidence
other social sciences (substantially slower). on one aspect, authors’ views of the referees’
There are many components that might contributions to their articles).
explain the slowness of economics journals. Answers to these questions require posi-
Culprits might be dilatory editors/referees, tive analysis. Making normative suggestions
authors who delay responding to initially about how the publication process in eco-
encouraging editorial responses, or lags nomics might be speeded up with no loss of
between a study’s acceptance and its publi- quality constitutes the second major section
cation. We cannot elucidate the underlying of this study. Although basically suggestive, it
causes of each of these possible contributing too has some positive bases, as we examine
factors. There is no way to infer why edi- data describing publications in journals that
tors or referees might “sit on” a paper, why have experimented with alternatives to stan-
authors might hesitate to revise their work dard practices in economics publishing.
quickly, or why economics differs from other
disciplines. All we can do is document the
2. Characteristics of Slowness in
magnitude of each factor’s relationship to
Economics Publishing
slow publication by providing the first evi-
dence on this issue, one novel contribution Much of the analysis in this section is
of this study. based on a set of data collected from leading
The central part of our empirical work economics journals. We asked the editors of
examines the relationship of slowness to each of the “top five” journals in the field for
subsequent scholarly attention. We measure details on each article, excluding Nobel/pres-
this relationship by a study’s post-publication idential addresses, comments, replies, et cet-
(both online and in-print) citations. If a lon- era, that were published in 2012 and 2013.
ger publication process is related to greater The details for each paper include: its initial
notice by other scholars, perhaps we should submission date; the date of the initial edito-
view these benefits as justifying its cost— rial response; the date of first resubmission,
although the evidence suggests that a slow et cetera, through the date of acceptance.2
process reduces the quantity of publications, We use articles published in those two years
as measured by pages written (Conley et al. to allow time for the importance of each arti-
2013). We cannot discover whether slowness cle to be reflected in subsequent research.
in economics leads to more attention from Three of the five editors provided the data,
other scholars than does research in other showing these outcomes for each of 241 pub-
disciplines, nor can we even analyze whether lished articles in these two years, thus allow-
increasing slowness in economics has made ing charting how each article flowed through
economic research generally more influen- the editorial process.
tial.1 We can, however, analyze whether at a
point in time published research with a lon- 2 We view the top five as the American Economic
ger gestation period is related to greater sub- Review, Econometrica, the Journal of Political Economy,
sequent scholarly attention, providing the the Quarterly Journal of Economics, and the Review
first objective evidence on this crucial out- of Economic Studies. We are very aware of differences
in these journals’ average impacts, of the tremendous
come of the publication process (although heterogeneity of impacts of articles within each journal
(Hamermesh 2018), and of the possibly deleterious effects
of overreliance on publication in these outlets (Heckman
1 There is evidence that the influence of economics and Moktan 2020). Nonetheless, we follow convention in
journal articles on scholarship in other disciplines has bibliometric analyses and restrict this part of the study to
increased (Angrist et al. 2020). these journals.
 Hadavand, Hamermesh, and Wilson: Publishing Economics 271

2.1 Slowness in Economics—and Other 2012–13.4 Time to acceptance is crucial for

Fields young scholars seeking tenured positions and
for more senior ones seeking new positions,
Along with other, publicly available infor- since with an acceptance they can include
mation, we use these data to examine the the publication on their curricula vitae. One
speed of publishing in economics, political might argue that a first-round revise/resub-
science, psychology, and the natural sci- mit (R&R) is sufficient indication of schol-
ences, both in 2012–13 and recently (2020). arly success, but many such responses do
The Review of Economics and Statistics not lead to publication in top journals. No
(arguably the most-cited general journal out- risk-averse institution is likely to make job
side the top five), along with one of the two offers or grant continuing tenure based on
top five journals that did not provide com- this indication, and in our experiences R&Rs
plete information, do publish submission, are heavily discounted in these contexts. As
acceptance (and obviously) publication dates Yogi Berra noted, “It ain’t over till it’s over,”
with each article. Adding this information to which is as valid in economics publishing as
that of the three journals in our main dataset, in baseball seasons.
we compare the process among them to that Acceptances are also important for econo-
in three other social science journals: The mists in obtaining public recognition of their
American Political Science Review (APSR), work, as journalists often ask whether a study
leading in its discipline; and the Journal of has been peer reviewed. They are also cru-
Applied Psychology (JApplPsy) and Journal cial to establishing bona fides in expert testi-
of Personality and Social Psychology mony or in providing policy advice generally.
(JPersSocPsy), two of a probable “top five” in Time to publication used to be an important
psychology, which also publish this informa- indicator of how long it took from the time
tion with each article. The same information when authors viewed their research as com-
is included with each article in Nature, one plete to when others could see and use the
of the two most widely cited scholarly jour- finished product. Today, however, this mea-
nals, and in the Proceedings of the National sure seems less important, since many jour-
Academy of Sciences (PNAS), which has a nals include and publicize widely available
five-year impact factor higher than all but online final versions of articles shortly after
one of the economics top five.3 acceptance.
The upper panel in table 1 presents statis- By any measure, the record in econom-
tics describing the distribution of times from ics is discouraging, perhaps epitomized by
initial submission to acceptance and then figure 1. The mean time from submission
to publication among articles published in to acceptance of articles published in these
journals (the REStat and four of the top five)
in 2012–13 was more than two years.5 This
3 Care is required comparing impact factors across dis-
ciplines, since scholars in different fields differ in their
propensity to cite other studies. The average article in the 4 The means are simple averages of the average times in
top five economics journals referenced 56 items in 2019, each journal in each group. Regrettably, several attempts
almost the same as in the APSR. Articles in the two natural to elicit this information from the leading sociology jour-
science journals averaged 42 references to other studies. nals, the AJS and the ASR, in a discipline arguably most
At the other extreme, articles in the two psychology jour- comparable to economics, failed.
nals averaged 106 references per article; and the two lead- 5 Björk and Solomon (2013) compare a large number of
ing sociology journals, the American Journal of Sociology business and economics journals along these measures to
(AJS) and the American Sociological Review (ASR), with journals in other broadly defined disciplines and demon-
impact factors of 5.9 and 8.2, averaged 110 references in strate, even across broad ranges of journals by quality, that
each article (Clarivate Analytics 2020). business and economics is much slower.
272 Journal of Economic Literature, Vol. LXII (March 2024)

TABLE 1
Acceptance and Publication Lags (in Months), Economics, Other Social Sciences, and “Hard”
Sciences, 2012–13 and 2020*

Weighted Percentile
Mean SE 10 25 50 75 90
2012–13
Four top 5 + REStat (535)**
Submission to acceptance 24.72 (0.63) 10 14 21 32 42
Submission to publication 33.15 (0.66) 17 25 35 43 52
APSR, JApplPsy, JPersSocPsy (371)
Submission to acceptance 12.84 (0.36) 6 8 12 16 22
Submission to publication 18.05 (0.37) 11 13 17 22 38
Nature, PNAS*** (195)
Submission to acceptance 5.77 (0.22) 3 4 5 7 9
Submission to publication 7.80 (0.24) 5 5 7 9 11

2020
Four top 5 + REStat (308)**
Submission to acceptance 26.38 (1.00) 10 15 22 35 50
Submission to publication 34.31 (0.93) 18 25 32 44 59
APSR, JApplPsy, JPersSocPsy (212)
Submission to acceptance 14.37 (0.52) 6 8 13 17 24
Submission to publication 22.99 (0.56) 14 17 21 26 32
Nature, PNAS*** (183)
Submission to acceptance 7.16 (0.41) 3 4 6 8 13
Submission to publication 9.35 (0.41) 5 6 8 11 15

Notes:
  *Number of articles in parentheses. Means weighted by the inverses of the numbers of articles from the
journal in the samples. Several articles (fewer than five in each case) accepted within a month of submis-
sion were deleted from the samples of psychology journals.
**In addition to the REStat and the three top five journals that provided confidential information, we also
include published information from one of the two top five journals that did not provide such data.
***The sampled issues cover only the last several weeks of the respective calendar years.
Source: Most of these data are available from Hadavand, Hamermesh, and Wilson (2024).

outcome compares to slightly more than one suggests problems, the extreme times can
year in the three other social science jour- only be characterized as awful.7 Even at the
nals, and only six months in the two natu- 75th percentile, the time from ­ submission
ral science journals.6 If the average time
(1953) one-page article was published eight weeks after
the discovery was announced.
6 Upon seeing this table, one distinguished economist 7 The 90th percentile statistics are bad enough. The
remarked, “If Watson and Crick had to deal with econom- maximum durations in the sample were 7 years 5 months
ics publishing, their article would have been 70 pages long from submission to acceptance, and 9 years 5 months from
and taken three years to get into print.” Watson and Crick’s submission to publication.
 Hadavand, Hamermesh, and Wilson: Publishing Economics 273

Manuscript submitted Manuscript accepted

Figure 1. Depiction of Publishing Lags in Economics

Source: Wikimedia Commons. Left: Allan Warren, HRH The Prince of Wales. Right: White House, King
Charles III (July 2023).

to acceptance is twice as long in econom- their publication processes less compara-

ics as in the other social sciences, and four ble to economics than those describing the
times as long as in the natural science jour- other social science journals.8 Differences
nals. Moreover, the uncertainty faced by in acceptance rates may be important; but
economic researchers is greater: The 90/50 why the very low acceptance rates in the top
ratios of time to acceptance are 2.0 in eco- five generate much longer publishing lags is
nomics, 1.8 in the social science journals, unclear.9
and 1.8 in the natural science journals. The One might argue that these figures reflect
90/10 ratios are 4.2, 3.7, and 3.0 in the three ancient history, and that the situation has
areas respectively.
While we do not inquire why economics
differs so greatly from other disciplines, it is
8 Very few economics-related papers are published in
worth noting that the acceptance rate in the
Nature, but the PNAS publishes many: In 2013–14, 49
top five journals currently averages 6 per- articles appearing there had at least one coauthor affiliated
cent, compared to 10 percent in the three with an economics department or similar institution. While
social science journals, and to 8 and 15 per- the average article in the PNAS is better-cited than most in
the top five, articles by economists published in that outlet
cent at Nature and the PNAS respectively. were cited at a slightly lower rate, cumulating an average of
At least in the natural science journals, the 59 citations over the eight years post-publication.
9 One is reminded of George Stigler’s perhaps apocry-
papers are shorter, and the supply of jour-
phal response to the then-editor of the AER, who com-
nal space is greater (with nearly weekly plained of having so many good papers to choose among,
publication), making the data describing “Why not publish one occasionally?”
274 Journal of Economic Literature, Vol. LXII (March 2024)

improved greatly over the past decade.10 The possible harm from slow publishing
That argument is wrong, as shown in the is not greatly mitigated by the ever-growing,
bottom panel of table 1. Despite now-uni- at both the extensive and intensive margins,
versal online submission procedures at series of discussion/working papers. These
these journals, the change over the decade are not peer reviewed, and thus lack the
was very small, with the mean time to accep- bona fides of journal articles in the eyes of
tance rising slightly and, worse still, with an other scholars, university administrators,
increase in the mass in the upper tail of the and the media. Moreover, the plethora of
distribution.11 Of the five economics jour- such papers creates congestion external-
nals, the median time from submission to ities, even in the most visible such series
acceptance increased in two, fell in two, (Lusher, Yang, and Carrell 2021), making
and was unchanged in one. The duration it difficult to keep up with what trusted
at the 90th percentile increased in four and experts view as important.
fell in one. Similar increases in the mass Our analysis has concentrated on the
in the upper tails of the distributions of top five journals, the pinnacle of scholarly
acceptance times in the other social science publishing in economics. While lower-level
journals and in the natural science jour- economics journals do have shorter sub-
nals also occurred; but their average speed mit-to-accept times, large numbers of the
and the speed of the slowest publications articles that they publish have gone through
remained far more rapid than in economics. an eventually failed submission process at
To summarize today’s situation succinctly, one or more of the top five or other eco-
an economics article that is at the 50th per- nomics journals. The process described
centile of time to acceptance would be at for these top journals may be even longer
the 85th percentile of times to acceptance at lower-level journals when one includes
in the other social science journals, and at the time from the initial submission at any
the 97th percentile in the two natural sci- journal.
ence journals.12
2.2 Contributors to Slow Publishing

10 One
There are many plausible explanations
of the editors who kindly supplied the data
underlying most of the work in this section questioned our for the slowness in getting an economics
request for 2012–13 data, stating that the journal’s process paper accepted. These include the number
may have been slow in the past but was no longer slow. We of times a paper is resubmitted, the amount
explained that we needed data from those earlier years to
examine the articles’ impacts. We have not had the heart of time that it spends with editors and ref-
to note that, while the mean submit-to-accept time at that erees (denoted here by time in journal’s
journal has speeded up slightly, the mass in the right tail hands), and the length of time that it spends
has increased.
11 The failure of submit-to-accept times to fall in the in author’s (or authors’) revisions (time in
economic journals could not have been due to COVID- authors’ hands). Here we examine the rela-
19-induced delays. Only 13 percent of the papers tab- tive contributions of each of these to the lags
ulated were accepted after April 1, 2020. Given the
rapid turnaround in the natural science journals, how- in publication and consider the characteris-
ever, COVID might explain their (small) increase in tics of the papers and their authors in rela-
­submit-to-acceptance times.
12 Without one of the top five journals, the statistics
tion to these outcomes.
in table 1 are incomplete. We cannot solve this problem
for 2012–13 for this journal, but we can piece together a
good estimate for articles in 2020 using some in-publica-
tion information and an email survey of authors. The data If these estimates were included in the statistics shown in
suggest that its mean submit-to-publication time was 30 the bottom part of table 1, they would reduce the mean
months, with a mean submit-to-accept time of 23 months. times by one month each.
 Hadavand, Hamermesh, and Wilson: Publishing Economics 275

Using the descriptions of each stage of the Table 2 presents descriptive statistics of
submission/review process for each of the all these variables except times in the jour-
241 articles published in the three top five nals’ and authors’ hands, which we examine
journals in 2012–13, we calculated the num- in detail below. The average or median pub-
ber of rounds of submission/resubmission/ lished paper goes through three rounds: It
re-resubmission/re-re-resubmission that is submitted, resubmitted, and resubmitted
each went through. We denote this number again, when it is then accepted; but nearly
by 2 if the second editorial response—the one-fourth of all articles went through a
response after the first resubmission—was fourth round. (For the empirical articles,
an acceptance, 3 or 4 if the third or fourth we consider acceptance as the date when
was an acceptance.13 We can decompose the an acceptance email was sent, that is, thus
total time from submission to publication ­earlier than the final submission that at many
into three parts: time spent in the journal’s journals today must include a documented
hands, time in the author’s/authors’ hands, dataset.)
and time between acceptance and in-print Empirical articles comprise 53 percent
publication. of the total (11 percent using administra-
In addition to these descriptors of the tive data, 42 percent using other data), with
editorial process itself, we gathered other pure theory accounting for 35 percent of the
information for each article about: its Web of publications, and experiments and econo-
Science citations (Clarivate Analytics 2021) metric theory accounting for the remaining
in each year from the year of publication 12 percent. The average article has slightly
through 2020 (nine years of citations to arti- above two authors, but nearly 10 percent
cles published in 2012, eight years to those have four or more authors, reflecting the
published in 2013); the cumulative number stretching of the right tail of the distribution
of Google Scholar citations that the article of ­authors/article noted by Hudson (1996),
had received as of March 2021; its length in Ellison (2002), Card and DellaVigna (2013),
pages;14 the number of references included; Hamermesh (2013), and Jones (2021). The
the number of authors; and the subfield in average article contains nearly 29 printed
which the article might be classified (theory; pages, not including the ubiquitous and
empirical with administrative data; other often voluminous online appendices. There
empirical, including calibration; experimen- is substantial variation in the number of ref-
tal; econometric theory). Characterizing erences included, and its correlation with
the articles’ authors, we obtained the Web article length is only +0.30. Of the articles’
of Science citations of each author in the authors, 22 percent were women, with the
year the article was submitted, used to con- incidence of female authors rising as the
struct the citations of the most-cited author; total number of authors increases.
the post-PhD experience of the most-cited The average article received 70 Web of
author, and the number of female authors. Science citations in its first eight or nine
years in print with, as is always the case in
citations, substantial skewness in this mea-
13 Two of the articles went through a fifth round,
sure (Hamermesh 2018). The skewness is
although in one case the elapsed time in that round was equally pronounced in the distribution of
less than one month. We treat those two as having endured the articles’ cumulative Google Scholar cita-
four rounds.
14 These are calculated based on the average number of
tions. Even more skewness exists in the dis-
characters per page in each journal, with the number nor- tribution of citations to each article’s most
malized to the journal with the most characters per page. cited author at the time of submission. The
276 Journal of Economic Literature, Vol. LXII (March 2024)

TABLE 2
Descriptive Statistics from the Three-Journal Sample, Articles 2012–13 (N = 241)

Variable means:
Number of rounds Field Number of authors (fraction with any female)
2 0.27 Theory 0.35 1 0.21 (0.12)
3 0.51 Admin. data 0.11 2 0.43 (0.18)
4 0.22 Other data 0.42 3 0.27 (0.28)
Experiment 0.09 4+ 0.09 (0.39)
Econometric theory 0.03

Percentile
Variable: Mean SE 10 25 50 75 90
Page equivalents* 29.18 0.41 21 25 29 33 37
References included 46.08 1.18 26 33 44 55 67
Cumulative Web of Science citations 69.52 5.15 10 20 44 86 168
Cumulative Google Scholar citations 299.24 21.67 42 78 168 402 739
Web of Science citations of most cited author during 271.81 36.53 7 28 97 284 701
year of submission
Post-PhD experience of most cited author 16.41 0.66 4 8 14 24 32

Notes: *Pages standardized to the journal with the densest format.

average most cited (on each article) author weighted as one-third of the total.) Several
received over 270 Web of Science citations aspects of the figure are striking:
in the year of submission, about average
among tenured faculty members in eco- (i) The main proximate determinant of
nomics departments that might be viewed as inter-decile differences in the speed
top 30; but the median most cited author of of publication is the huge rise in the
an article was cited only one-third as often. amount of time spent in authors’ hands
The post-PhD experience of the most cited (the sum of times between receiving a
authors averaged around 15 years—typical response from a journal to resubmis-
of relatively young full professors and con- sion on each round) as the total time
sistent with evidence on the age distribution to acceptance and publication rises.
of authors in leading economics journals Among papers in the middle decile,
(Hamermesh 2013). this is 10 months; among those in the
With our focus on the process by which slowest decile, 26 months are spent in
articles are handled, we examine the con- authors’ hands.
tributions of the three components of time
from submission to publication. Their dis- (ii) While the amount of time spent at
tributions are presented in figure 2, con- journals increases with the slowness of
taining decompositions of the average time publication, moving from the middle to
from submission to publication, measured the slowest decile increases that dura-
in months on the vertical axis and shown tion only half as much, from 10 to 18
within each of five deciles. (Each journal is months.
 Hadavand, Hamermesh, and Wilson: Publishing Economics 277

60

48 Journal
Speed in months

Author(s)
36 Accept to publication

24

12

0
Fastest decile Third decile Middle ten percent Eighth decile Slowest decile

Figure 2. Contributions to Slowness, by Decile of Speed

TABLE 3
Mean, Variance, and Decomposition of Variance of the Submit-to-Acceptance Lag, Total and by
Journal (in Months)

Mean Variance Variance due to author(s) Variance due to journal 2*Covariance

All journals 22.24 16.80 9.25 4.37 3.18
Journal 1 15.51 12.03 10.49 0.27 1.27
Journal 2 24.62 14.17 3.73 6.36 4.08
Journal 3 26.24 17.44 10.38 2.83 4.23

(iii) Lengthier submit-to-publication times its two sources and their covariance. The
are essentially unrelated to differences most interesting findings in this table are:
in the time between acceptance and
publication. (i) The substantial heterogeneity in the
length of time that articles are in pro-
Figure 2 aggregates across the three jour- cess—the variance is quite large even
nals and does not reflect the role of the within a journal.
heterogeneity of journals in the total publi-
cation lag. To examine how these contribut- (ii) The heterogeneity across the journals:
ing factors differ across the three journals, journal 1 handles the papers some-
table 3 shows the means and variances of the what more quickly than the other two,
submit-to-acceptance lags among the 241 but, most important, there is almost no
­articles in total and for each journal sepa- variation in the amount of time a paper
rately, and it decomposes the variance into spends with editors/referees.
278 Journal of Economic Literature, Vol. LXII (March 2024)

TABLE 4
Correlations of Time in Journals’ and Authors’ Hands

All papers Completed in round

Round Correlation N= Correlation N=
2 0.023 241 0.255 65
3 0.260 176 0.256 124
4 0.364 52 0.364 52

(iii) Despite the heterogeneity in the in ­column 1 of table 5 we present least-

sources of the acceptance lags, and con- squares regressions of the number of rounds
sistent with the evidence in figure 2, through which a paper travels at a journal as
over half of the variation in lags in accep- a function of all the article/author character-
tance at each journal arises from authors istics on which we have information (except
spending more time on revisions. the number of references included, which
may be partly affected by the number of
The covariance between the time in a rounds and time spent refereeing/revising).
journal’s and author’s/authors’ hands is posi- Column 2 presents the same regressions
tive in table 3, but in no case does it account with journal indicators added to account for
for even one-third of the total variance in the the heterogeneity demonstrated in table 3.
submit-to-acceptance time. Even this low
­ The estimates treat each article equally—
correlation is due mainly to the fact that arti- the observations are unweighted.15
cles that go through more rounds necessarily Authors’ characteristics are unrelated
take more time of both authors and editors/ to the number of rounds an article goes
referees. The correlations between editor/ through: how well-cited an author is, senior-
refereeing time and authors’ response times ity, and gender are all unrelated to this out-
are shown for each round separately in table come. Characteristics of the article are,
4, both for all papers handled in the round however, related to the number of rounds:
and for those completed in that round. The Theory papers are handled in significantly
correlations at each article’s final round aver- fewer rounds, with differences across the
age +0.30. Thus, those articles that take other subfields being small and statistically
more editor/referee time to handle are asso- insignificant (not shown in the table). Papers
ciated with authors spending more time, but with fewer authors are handled more rap-
the relationship is weak. idly. Articles that are longer when published
Various characteristics of the articles
might cause them to go through more 15 Using weights that are inversely proportional to the
rounds of resubmissions at a journal; and number of articles in each journal in the sample produces
they might lead editors and referees to only minute changes in the estimates. Similarly, while
ordered-probit estimation is more appropriate than least-
spend more time handling the paper. The squares, its implications differ little from those of the
same characteristics might lead authors results in the table. Also, replacing the variable “any female
to take longer resubmitting an article that author” with indicators of the number of female authors
and replacing “two or more authors” with indicators of
has received an encouraging initial edito- their number do not change the qualitative conclusions
rial response. To examine the first issue, about the effects of these measures.
 Hadavand, Hamermesh, and Wilson: Publishing Economics 279

TABLE 5
Determinants of the Editorial Production Process, OLS Estimates, N = 241

No. of rounds Months at journal Months with author(s)

Ind. Var.:
Citations to most cited 0.009 0.005 –0.256 –0.129 –0.161 –0.121
author/100 (0.008) (0.008) (0.086) (0.070) (0.127) (0.126)
Experience of –0.006 –0.006 0.072 0.071 0.074 0.079
most cited author (0.005) (0.005) (0.049) (0.039) (0.073) (0.071)
Any female author 0.044 0.028 –0.902 –0.413 2.467 2.570
(0.112) (0.109) (1.145) (0.920) (1.696) (1.659)
Two or more authors –0.171 –0.162 –1.577 –1.700 –0.013 0.078
(0.115) (0.118) (1.190) (0.955) (1.762) (1.722)
Equivalent pages –0.004 –0.005 0.121 0.039 0.043 –0.074
(0.007) (0.007) (0.075) (0.064) (0.112) (0.115)
Theory –0.238 –0.205 –1.070 –2.235 –1.808 –2.142
(0.098) (0.096) (1.000) (0.810) (1.481) (1.461)
Adj. R2 0.018 0.081 0.041 0.383 0.006 0.052

Notes: Standard errors in parentheses. Columns 2, 4, and 6 contain journal indicators. The correlation of the resid-
uals in columns 3 and 5 is 0.24, between the residuals in columns 4 and 6 it is 0.31.

are handled in no more rounds than shorter impact on the profession matters much
articles. more than one’s seniority in relation to
Columns 3 and 4 of table 5 present esti- how an author is treated. Other than these
mates of the correlates of the length of time effects, none of an article’s characteristics is
that the journals take to handle a submission. correlated with the time that it spends at a
The clearest result is that theory papers are journal.
dealt with significantly and substantially more The final two columns of table 5 describe
quickly than other articles, 2.5 months on a the determinants of the time that authors
mean of 10 months, at each journal (again spend revising their papers in response to a
with only small differences across the other requested resubmission. The only correlates
subfields). Weaker evidence shows that hav- whose relation to this outcome are even
ing multiple authors is associated with more marginally significant are the presence of a
rapid treatment by the journals, perhaps female author (2.5 months extra on an aver-
because coauthors help iron out problems age of 12 months) and the negative estimate
that might otherwise lead editors and ref- for articles classified as theoretical.16 There
erees to spend more time handling the arti- is no evidence that authors whose prior work
cle. There is weak evidence that better-cited has had a greater scholarly impact or those
authors receive somewhat faster treatment
and that, conditional on an author’s prior 16 Less time is spent revising theory articles at each
scholarly recognition, more senior authors’ round of the publication process. There are no significant
submissions are handled more slowly, other differences across the other subfields. Thus, articles using
administrative data take no longer than other non-the-
things equal. These last two results are con- ory articles to revise, and similarly for articles based on
sistent with the ­observation that one’s prior experiments.
280 Journal of Economic Literature, Vol. LXII (March 2024)

who are more senior are differentially slow each article, since additional references in
in handling requests for revision. Here and an article might, for scholarly or invidious
throughout this table, variations in the length reasons, generate more subsequent citations
of articles have essentially no effect on the to the article. Column 1 shows least-squares
outcomes. The results demonstrate that estimates of the relationship of citations to
most of the variation in authors’ behavior is the number of rounds the article has gone
idiosyncratic.17 through and to various control variables.
With average annual citations of about eight,
2.3 Slowness and Citations the estimated relationship of subsequent
citations to the presence of an additional
The most important question in judg- coauthor is low (within this set of studies in
ing whether the uniquely lengthy publica- these leading journals), although not much
tion process in economics is worthwhile is different from that found in other studies
its relation to the scholarly attention to the (Hollis 2001, Medoff 2003, Bosquet and
research that survives this very lengthy treat- Combes 2013, Hamermesh 2018). Having a
ment at these major outlets. We recognize female author on a study has a substantial,
that research published in these top journals but not quite statistically significant, positive
often has important influences beyond those relation to the scholarly impact of the article,
on economists or other scholars, for exam- larger than found in other studies (Laband
ple, on debates about policy or on inchoate 1987, Ferber and Brün 2011, Hamermesh
popular feelings about economic issues. 2018), maybe because of within-subfield dif-
Nonetheless, economic research, indeed ferences by gender in the topics on which
any scholarly research, is judged at least economists work, or perhaps because these
in part by the extent to which subsequent are better articles.19 Lengthier articles have
work acknowledges its influence. We, there- no greater association with subsequent cita-
fore, answer this question by measuring the tions than do shorter ones, perhaps due to the
relationship of the outcomes examined in relatively narrow range of page lengths in the
tables 3 and 5 to annual patterns of (Web sample. A one standard deviation increase in
of Science) citations up through 2020 to the the number of references is related to a sta-
articles published in 2012 and 2013.18 tistically significant 0.06 additional standard
Table 6 lists the results. Each observation deviations in citations.
is an article/year, necessitating clustering Theory papers on average receive roughly
standard errors on the individual articles. half as many citations per post-publica-
In addition to the regressors in table 5, we tion year as do otherwise identical articles
add the number of references included in in other subfields, a result consistent with
evidence comparing leading specialized
journals in different subfields.20 Articles by
17 Here and in the next subsection we also experiment
with a measure of heterogeneity—the standard deviation
of citations across coauthors. This measure is uncorrelated
with the time coauthors spend revising, and its inclusion 19 The articles cited on this issue, which is quite second-
has minute effects on the estimated impacts of the other ary to the crucial points of this study, are part of a burgeon-
regressors. ing and now voluminous literature.
18 Checchi et al. (2021) show that there is a remarkably 20 The average five-year impact factor among the Journal
high correlation between this objective bibliometric mea- of Development Economics, Journal of Econometrics,
sure and subjective peer-based evaluations of individual Journal of International Economics, Journal of Labor
research products, suggesting that a subjective approach to Economics, Journal of Monetary Economics, and Journal
measuring impact would yield results that would arguably of Public Economics was 3.67 in 2019. The average five-
be similar. year impact factor of Games and Economic Behavior and
 Hadavand, Hamermesh, and Wilson: Publishing Economics 281

TABLE 6
Determinants of Annual Post-Publication Citations (N = 241 Articles, 2,049 Citation-Years)

OLS LAD
Ind. Var:
3 rounds 2.541 3.261 1.314 1.827 0.815 1.020
(0.956) (1.038) (0.924) (1.033) (0.323) (0.357)
4 rounds 3.066 5.909 2.349 3.987 0.977 1.295
(1.671) (2.145) (1.645) (2.209) (0.382) (0.552)
Journal hands — –0.145 — 0.010 — 0.259
(0.071) (0.086) (0.306)
Author(s) hands — –0.147 — –0.140 — –0.662
(0.057) (0.062) (0.170)
Citations to most-cited 0.508 0.447 0.463 0.446 0.407 0.409
author/100 (0.091) (0.085) (0.088) (0.086) (0.057) (0.079)
Experience of –0.114 –0.091 –0.118 –0.106 –0.103 –0.095
most-cited author (0.041) (0.042) (0.040) (0.041) (0.013) (0.016)
Any female author 2.953 3.064 2.696 3.011 0.755 0.918
(1.695) (1.638) (1.610) (1.600) (0.372) (0.376)
Two or more authors 1.206 1.301 1.139 1.360 0.494 0.429
(1.108) (1.111) (1.097) (1.104) (0.322) (0.344)
Equivalent pages –0.099 –0.068 –0.027 –0.037 0.005 0.011
(0.078) (0.076) (0.082) (0.083) (0.022) (0.023)
Number of references 0.072 0.065 0.058 0.061 0.029 0.029
(0.026) (0.025) (0.025) (0.025) (0.007) (0.007)
Theory –4.087 –4.224 –3.734 3.859 –1.969 –2.036
(0.898) (0.913) (0.937) (0.967) (0.257) (0.266)

R2 0.253 0.275 0.273 0.284 0.257 0.266

Notes: Standard errors in parentheses, clustered on articles. Columns 3–6 contain journal indicators. Each equation
also includes a vector of indicators of year post-publication. OLS stands for ordinary least squares and LAD stands
for least absolute deviation.

authors whose prior work has been more careers will not “make it” even with work
heavily cited receive more attention; but published in a leading outlet.
conditional on that measure, more senior The central variables of interest indicate
authors’ work is cited less. As with the the number of rounds at the journals. The
impact of these measures on the time that results suggest, other things equal, that the
journals spend handling the paper, this jux- 51 percent of articles that require a third
taposition suggests an autocorrelation of round (two resubmissions) are related to
scholarly attention to one’s work, and that greater subsequent citations than the 27 per-
those who have not “made it” earlier in their cent of papers that go through only two
rounds (are accepted after the first resubmis-
sion), the excluded category in table 6. On the
the Journal of Economic Theory was 1.49 (from Clarivate other hand, the marginal additional citations
Analytics 2020). related to the fourth round (the 22 percent
282 Journal of Economic Literature, Vol. LXII (March 2024)

of articles that are resubmitted, resubmitted citations to the time that an article
again, and then accepted after yet another spends with the journal disappears.
resubmission) is smaller, although positive.
One might think that greater editorial (ii) The estimated relationship of a fourth
attention or more time that authors spend round at a journal to subsequent cita-
revising before resubmission(s) would tions is reduced, but becomes about
improve the quality of the article in terms of equal to that of a third round, with both
its subsequent impact. The specification in effects not quite reaching standard lev-
column 2 thus adds measures of time spent els of statistical significance.
at the journal and with author(s). Given the
number of rounds an article goes through, (iii) Most importantly, the negative rela-
greater lags in the process are negatively tionship of citations to additional time
related to its subsequent attention by other spent in authors’ hands is essentially
scholars.21 These estimates are statistically unchanged and statistically significant.
significant, not huge, but not small either: A
one standard deviation increase in the time (iv) None of the estimated coefficients on
at a journal is associated with 0.11 ­standard the control variables is altered in any
deviations fewer subsequent citations. important way.
Similarly, a one standard deviation increase
in the time that authors spend revising is As the statistics in table 2 demonstrate,
associated with 0.13 standard deviations citations to the articles in this sample are
fewer citations. highly skewed, as are prior citations received
These estimates ignore the tremendous by their authors. The regressions in columns
heterogeneity across journals in the kinds 1–4 of table 6 describe the average experi-
of articles published and, as table 3 showed, ence of these published articles; but given
in how they are treated. This difficulty is the skewness in these variables, they do not
accounted for by our preferred estimates, describe what the median author faces. To
shown in columns 3 and 4 of table 6, which infer that, columns 5 and 6 present least
include journal indicators. The major com- absolute deviation (LAD) estimates, with
parisons to the results presented in the first the same specifications as in columns 3 and
two columns are: 4, including the journal indicators. While the
parameter estimates of the control variables
(i) Not surprisingly, given the heterogene- are smaller than in the ordinary least squares
ity shown in table 3 and the relatively low (OLS) estimates, they are qualitatively quite
within-journal variation in this measure, similar. The estimated coefficients on of the
the negative relationship of ­subsequent number of rounds through which an arti-
cle passes and the amount of time spent in
editor/referees’ or author’s/authors’ hands
21 To account for citations to articles prepublication, are also smaller; but the basic inference
we reestimated the equations here and in table 7 using remains the same. The marginal association
cumulative Google Scholar citations (through March of another round at a journal to subsequent
2021) instead of annual Web of Science citations. This
­respecification does not qualitatively alter any of the infer- citations is positive and statistically significant
ences. With cumulative Google Scholar citations equaling and, conditional on the number of rounds,
roughly 3.5 times annual Web of Science citations in this authors’ slowness in revising their work has
sample, the coefficient estimates differ in proportion. The
measures that are significantly related to citations in table a significant negative relation to its later
6 remain significant. citations.
 Hadavand, Hamermesh, and Wilson: Publishing Economics 283

TABLE 7
Determinants of Post-Publication Citations, Theory Articles vs. Others

Ind. Var.: Non-theory Theory

OLS LAD OLS LAD
3 rounds 0.916 –0.206 1.578 0.727
(1.670) (1.048) (1.283) (0.637)
4 rounds 3.628 0.559 1.082 0.314
(3.255) (1.368) (1.406) (0.922)
Journal’s hands 0.099 0.069 0.025 0.033
(0.124) (0.063) (0.075) (0.042)
Author(s)’ hands –0.193 –0.083 –0.004 –0.004
(0.076) (0.038) (0.068) (0.030)

R2 0.288 0.273 0.315 0.270

N (articles, observations) (158, 1,345) (83, 704)

Notes: Standard errors in parentheses, clustered on articles. Each equation also includes
a vector of indicators of year post-publication. Also included in each equation are journal
indicators and all the other independent variables included in the estimates of table 6.

Table 6 shows that there are differences in author(s)—to subsequent citations have dif-
the subsequent scholarly attention received ferent implications. The conclusions from
by theoretical and other articles, while table 6 apply mainly to articles in subfields
table 5 showed that journals spend sharply other than theory: among those sub-fields,
different amounts of time dealing with them the relationship of citations to a third resub-
and that authors of theory articles spend mission (a fourth round) is positive at the
less time revising in response to resubmis- margin; and, most importantly, as in the
sion requests. Perhaps this is because upon entire sample, additional time that authors
submission a theory paper is clearly correct spend revising remains associated with a sta-
or incorrect, with fewer inherent possibil- tistically significant lesser scholarly impact.
ities for revision, and the main issue being Among theory articles, the time spent either
whether the result is sufficiently important. at the journal or by authors is unrelated
Regardless, to examine the theory–other to subsequent citations, although there is
subfield distinction further, table 7 presents ­evidence that the marginal relationship of a
estimates of equations specified like those second resubmission (a third round) to cita-
in columns 4 and 6 of table 6, but with the tions is positive, while that of a fourth round
articles separated into subsamples of theory is not.
and other papers. For each type of article, The analysis of this sample leads to the
the first column shows OLS estimates, the conclusion that multiple rounds of editing/
second LAD estimates. handling at these journals may be useful
Depending upon the type of article, esti- (in terms of articles’ relations to the atten-
mates of the relations of the crucial vari- tion in subsequent scholarly work), although
ables—an extra round of resubmission and statistical significance is low. Publishing lon-
the times spent at the journal and with ger papers (within the range of full-length
284 Journal of Economic Literature, Vol. LXII (March 2024)

articles included in the sample) is, however, positively related to later scholarly attention,
unrelated to scholarly attention. The stron- however, that gain must be traded off against
gest conclusion is that authors spending more the cost to (younger) scholars’ careers, in
time responding to requested resubmissions that additional back-and-forth with journals
is associated with less scholarly attention. postpones their ability to demonstrate their
scholarly prowess. Thus, considering alter-
native editorial arrangements and rules is
3. Solutions to Slowness
important. A second margin is in the time
The results in sections 2.2. and 2.3. do authors “sit on” their papers after hearing
not reflect ex ante random assignment of back from journals, time that our results sug-
papers to quicker or longer processes; nor gest is related to lesser scholarly attention.
were articles randomly assigned to differing The final margin is the time that editors and
amounts of time spent by e­ditors/­referees referees spend handling papers, time that at
or by authors. (We do not see how such the margin is unrelated to attention by later
randomness could be ethical, although ran- research.
domly nudging some submitting authors to
3.1 Fast-Tracking—The Economic Inquiry
choose faster decision routes might work.)
Innovation22
Without a true experiment, we cannot be
sure that articles that went through more We first consider the oldest major effort
rounds were not inferior to others ab initio to accelerate the refereeing/publication pro-
and required extra attention to bring them cess, Economic Inquiry’s (EI’s) introduction
up to par. Similarly, articles on which authors in 2007 of a two-track process. Submitting
spent more time, conditional on the number authors could choose between a fast track, in
of rounds, might have needed that time to which the article receives a simple accept or
rise to the minimum acceptable quality level reject; or a regular track, which might lead to
of the journal. The former caveat may be an acceptance with minor revisions, or one
important, although we saw that additional or more revise/resubmit responses with sub-
rounds had a positive relation to subsequent sequent additional refereeing, or to rejec-
citations. The latter does not seem credible, tion (McAfee 2010) (https://weai.org/view/
especially given the low correlation of the EI-No-Revisions). Several journals have now
time a journal takes to generate a first R&R instituted a similar quick turnaround pol-
request and the author’s/authors’ time spent icy, although none appears to offer authors
responding to it. a choice of tracks or a definite no-revision
We assume throughout that, given the track. This policy change is obviously not a
returns to publishing in these top journals, randomized experiment: authors may non-
the disincentives posed by the lengthy ref- randomly self-select into the fast-track group
ereeing process deter almost no submissions and, as shown below, they did so along one
(despite concerns expressed by Azar 2007). interesting dimension.
Assuming, therefore, that shortening the To examine how this experiment worked
process would be desirable, the findings in out, we obtained confidential data from the
the previous section point to three margins Western Economic Association International
along which the publishing process might be on the 935 articles published in EI between
improved with no loss of quality. Additional 2009 and 2018 inclusive, yielding a usable
back-and-forth between authors and edi-
tors—more rounds with a journal—has some 22 This subsection is a very much shortened version of
scholarly value. Even if third revisions are Hadavand, Hamermesh, and Wilson (2020).
 Hadavand, Hamermesh, and Wilson: Publishing Economics 285

TABLE 8
Selection Equation, and Relation to Citations, Economic Inquiry
Dependent variables
Independent variables Fast-track selection Citations
Any author [5–10] years post-PhD 0.085
(0.029)
Years post-PhD of most-cited author 0.0023 0.0087
(0.0014) (0.0057)
Five prior years citations of most-cited author (/1,000) 0.0345 0.380
(0.0159) (0.125)
Fast track 0.510 0.419
(0.188) (0.203)
Two-round regular 0.147 0.103
(0.138) (0.155)
Years from submission to acceptance 0.051
(0.116)

Year post-publication (9) X X

Issue number X X
N pages X
JEL category (10) X
N authors (4) X
Any female author X
Pseudo-R2 or R2 0.020 0.087 0.117
N= 835 3,889 3,889

Note: Standard errors in parentheses, clustered on articles.

sample of 835 articles that were not invited initial response and final acceptance on reg-
and were at least ten pages long. In addition ular-track papers. There is a very long tail
to all published articles, we have information among these papers, with a 90th percentile of
on the track used for 5,178 rejected arti- 17 months (compared to 7.5 months among
cles. We obtained information that allowed fast-track papers). Aside from the obvious
the construction of variables that are similar risk of rejection, submission along the regu-
mutatis mutandis to those used in section 2. lar track carries a small risk of involvement in
Fast-track papers were only slightly, albeit a dragged-out process, less arduous than at
statistically significantly, more likely to be top five journals, but still quite long.
accepted for publication than those submit- Certain characteristics of authors gen-
ted through the regular track (an acceptance erated predictable differences in the track
probability of 0.159, s.e. = 0.004, versus chosen. As the probit derivatives in column 1
0.149, s.e. = 0.002). There is little difference of table 8 suggest, more successful (in terms
in the time between submission and first of prior scholarly impact) and more senior
decision among accepted papers along the authors were more likely to choose the fast
two tracks. Rejection times are also similarly track. Most interesting, we searched over
distributed across tracks. The difference various ranges of the seniority of authors
between them arises from the lag between to find where the likelihood function was
286 Journal of Economic Literature, Vol. LXII (March 2024)

maximized. This occurred using the closed regular-track papers receive more citations
interval [5, 10] years post-PhD If any author than those published papers that were ref-
was in this experience range, the probability ereed multiple times, demonstrating that
of a fast-track submission was significantly the relationship of an extra revision to subse-
higher. With only 19 percent of published quent citations was zero.24
articles submitted along this track, the We stress that these estimates do not
parameter estimate implies that this choice account for the possible endogeneity of the
is over one-third more likely if an author is choice of track; this sample provides no
in this range of experience—presumably evidence that fast-track handling reduced
facing an impending job-security and/or subsequent attention to an article.25 The
promotion decision. The evidence suggests apparent absence of any relation of third or
that this type of fast-tracking might aid junior higher-order rounds of refereeing to citations
scholars who face a rapidly approaching ten- contrasts with the results for the third and
ure decision. fourth rounds in section 2. The difference
Of accepted submissions, 51 percent along may, of course, simply result from sampling
the regular track went through more than two differences or from underlying unmeasur-
rounds of submissions, far below the 73 per- able quality differences between these arti-
cent among top five journals. Paralleling cles and those published in top five journals.
the analysis in section 2.3., we estimate the A substantive explanation is that refereeing
determinants of annual citations to each of at the top journals may be of higher qual-
the 835 usable articles. As was done there, ity than at EI, with the difference in quality
each article is included as an observation being more pronounced on later-round ref-
in each post-publication year. Column 2 of eree reports.
table 8 presents a simple model, including
3.2 Fast-Tracking: AER: Insights and the
only the track chosen; if on the regular track,
AEJs
whether an article went through “only” two
rounds of submission, and two variables that While Economics Letters has promised
mechanically alter the number of citations. and delivered rapid turnaround with little or
Column 3 adds the length of time (in years) no revision for many years, its policy was only
from submission to acceptance, the post- adopted by top general economics journals
PhD experience and prior citations of the with the creation of the American Economic
most cited author, and controls for: number Review: Insights (AERI), which published its
of pages; JEL category, aggregated into 10 first articles in 2020. The question is whether
groups; number of authors; and whether at the AERI has lived up to its promise to
least one author was female. short-circuit the publication lags that charac-
The least-squares estimates in table 8 terize the top five journals. To examine this,
demonstrate that, whether we include we obtained information from the American
covariates or not, fast-track papers are cited
significantly more than articles submitted 24 In a survey reported in Hadavand, Hamermesh, and
through the regular process.23 Two-round Wilson (2020), we find no evidence that fast-track submis-
sions had previously been rejected at more journals than
regular-track submissions.
23 Reestimating the equations in columns 2 and 3 using 25 Instrumenting for fast-track using an indicator of
Poisson estimation to account for the count nature of the whether any author of the article was in the five- to ten-
dependent variable, which contains many zeros, yields year range of experience reduced the impact of fast-track-
essentially the same conclusions as the table. The same was ing on citations somewhat; but, with the instrument quite
true for tables 6 and 7, although the near absence of zeros weak, as implied by the estimates in column 1 in table 8, it
in those data make this technique less appropriate. is not clear what this search for exogeneity tells us.
 Hadavand, Hamermesh, and Wilson: Publishing Economics 287

Economic Association (AEA) on the submis- publications in the AEJ: Applied in 2020 and
sion and acceptance dates of all articles pub- 2021.27 Of the 80 articles, 26 were submitted
lished in the AERI in 2020 and 2021. with prior referee reports, with 25 enclosing
The top panel of table 9 presents infor- reports from the AER, and one with reports
mation like that in table 1, submit-to-accept from the AERI. The bottom two panels
and submit-to-publish times. The record so of table 9 show the submit-to-accept and
far is extremely encouraging. The mean sub- ­submit-to-publication times for articles with-
mit-to-accept time is four months, and there out and with a refereeing cascade. Papers
is remarkably little variance around this aver- including a cascade of reports are handled
age: the 90th percentile is only five months. on average five months more quickly than
Of course, submit-to-publish times are lon- other papers; they are handled with slightly
ger, about 14 months on average, but they less variability than others; and the upper
too suggest that this innovation has been end of the distribution of submit-to-accept
highly successful so far.26 Whether it will be times is substantially shorter. This compari-
successful in generating the same scholarly son suggests the value, in terms of speeding
attention as the top journals cannot yet be up acceptance/publication, of allowing cas-
known; but at least scholars whose work is cades. As noted above, we cannot construct
accepted by the AERI will know that their the appropriate experimental counterfactual
work can be circulated quickly with the that would allow us to infer whether cascad-
imprimatur of a widely recognized outlet. ing causes more rapid treatment; nor can we
The second novelty is the creation of the tell whether it raises the citations that “cas-
opportunity for what we call “cascading ref- caded” articles will eventually receive.
eree reports”—allowing authors to submit
3.3 Fast-Tracking: Desk Rejections
referee reports on their work from journals
that had previously rejected the paper. This Desk rejections speed the review process
policy was adopted at their inception by the and have become de rigueur among top five
American Economic Journals (AEJs), and it journals, with these journals now reporting
has now spread to many other journals. It is desk rejection rates of around 50 percent.28
not possible to tell whether it has speeded Until the 1990s desk rejection was essentially
up the editorial process, since authors’ unknown in these journals, with the AER
choices about whether to include prior ref- explicitly refusing to desk reject submissions,
eree reports depend on the nature of those since its board felt that the journal repre-
reports. But with many papers rejected not senting the profession should be open to all
because of errors, but because editors view submissions.
them as being of insufficient interest for their The growth of desk rejections is a rational
top five journal, creating a cascade of reports response to the growth in the number of sub-
can be a good strategy for authors and can missions and the burden that providing ref-
save editors’ and referees’ time. eree reports on all of them would entail. So
To examine whether published articles that long as the desk rejections are rapid, there is
are submitted with a cascade move through
the publication process more rapidly, we 27 These data and those on the AERI are available in
obtained information from the AEA on all Hadavand, Hamermesh, and Wilson (2024).
28 Web-based and directly communicated editors’ state-
ments report desk rejection rates at the AER, Econometrica,
26 These “short” articles averaged 17 pages, far below the Journal of Political Economy, the Quarterly Journal of
those in the “top five;” but they are longer than the average Economics, and the Review of Economic Studies of 46, 43,
article in the AER was before the late 1990s. 55, 63, and 55 percent respectively.
288 Journal of Economic Literature, Vol. LXII (March 2024)

TABLE 9
Acceptance and Publication Lags (in Months), Economics, “Novel” Economics Journals, 2020–21

Percentile
Mean SE 10 25 50 75 90
AER: Insights (N = 64)
Submission to acceptance 4.05 (0.15) 3 3 4 5 5
Submission to publication 14.72 (0.37) 11 13 14 16 18

AEJ: Applied, No cascade (N = 54)

Submission to acceptance 15.13 (1.21) 8 10 14 19 24
Submission to publication 29.13 (1.26) 20 23 28 34 39

AEJ: Applied, Cascade (N = 26)

Submission to acceptance 10.04 (0.95) 5 6 9 14 17
Submission to publication 23.77 (1.12) 17 20 22 29 32

Note: These data are available from Hadavand, Hamermesh, and Wilson (2024).

little time (but often substantial psychologi- year among 25 percent. Of the 176 articles
cal) cost to authors.29 But are there enough that went through three or four rounds,
desk rejections? Even with half of submis- 15 percent spent more than six months in
sions being summarily turned down, accep- the author(s’) hands between the second
tance rates among those articles that do response and the second resubmission. Most
receive reports remain at 10 percent at these surprising is that nine of the 52 papers that
journals. It is difficult to believe that editors went through four rounds were worked on
feel that half of submissions have a serious for more than three months between the
chance of being worthy of appearing in their third editorial response and the final resub-
journals. If editors exercised still more dis- mission. While the times to resubmission
cretion and desk rejected more submissions, decrease with the number of resubmissions,
publication would be accelerated, the bur- they remain very long.
den on referees would lessen, and perhaps With the demonstration that the addi-
the quality of referee reports would increase. tional time that authors spend resubmitting
is at best useless, the question arises as to
3.4 Limiting Revision Time
why so much time is spent. One reason may
In the dataset describing top five articles be that procrastinating authors produce low-
published in 2012–13, the time between er-quality research, so that more r­evisions
receipt of the first decision and the first are ­ necessary to produce an acceptable
resubmission exceeded six months on 56 per- paper. Yet another reason may be that they
cent of the 241 articles; and it exceeded one are too busy to devote the real time neces-
sary for producing high-quality revisions
(although the estimates in table 4 showed
29 Our experience, and that of colleagues, is that desk that these lags are unrelated to authors’ char-
rejections occur within one week of submission. The quick- acteristics that might indicate that their time
est we know of was on a submission that was sent from
Europe at noon on a Sunday, with the rejection occurring is more valuable than others’). An alternative
in the United States at 9am that same day. explanation is that some authors may use the
 Hadavand, Hamermesh, and Wilson: Publishing Economics 289

submission process to obtain comments on because for some reason the ethos that gen-
a paper that was not well-polished and was erates publishing lags in the top five has
submitted prematurely, with revision time not infected it as much as other special-
needed to bring the paper up to a minimally ized journals, or because of something else
acceptable level. Yet a fourth possibility con- unique to its subspecialty. Regardless, this
sistent with slowness on the second and sub- admonition might be included in all revise/
sequent resubmissions is that the feedback resubmit letters. By providing at least a soft
received in response to the first resubmission deadline, journals might take advantage of
is of reduced quality because the authors and, incentives that induce collaborators to move
especially, the editors and referees failed to together more quickly (Bonatti and Hörner
remember all the nuances of a subject that 2011). Going further, the evidence in this
they handled many months or even years study of the negative relation between sub-
before. Regardless, it is clear that allowing sequent citations and lags in authors’ revi-
authors free rein to delay resubmission is sions suggests that imposing and enforcing
unrelated to their articles’ scholarly value. a six-month limit on time spent revising, as
Requiring rapid resubmission is standard appears to be done in psychology, would not
in the natural sciences, common in other be harmful to their eventual scholarly atten-
social sciences, but uncommon in econom- tion. If nothing else, it would help pull the
ics. As an illustrative exception, the American right tail in the distribution of submit-to-
Journal of Agricultural Economics (AJAE), acceptance times to the left.
the oldest and most distinguished in its sub- An objection to this proposal is that authors
field, does not impose a deadline but does are busy. Of course they are. But for most
include in its R&R letter, “please submit the authors publishing an article in these journals
revised manuscript and separate responses is a jackpot prize, one that merits putting an
to the reviewers … within six months of invited resubmission on the “front burner”
receiving this letter.”30 It had a median of activities. Very few, if any, requested revi-
submit-to-acceptance time of 10 months sions take more than six months of actual
in 2020, with the 90th percentile being 22 work; it is more likely that the delays simply
months.31 Both statistics are unsurprisingly result from authors’ procrastination. Given
far below the comparable statistics shown the rewards, procrastination is difficult to
in table 1; but they are also below those of explain, and it can be costly.32
the six ­leading specialized applied journals While some top journals specify length
listed in footnote 20, which had median sub- limits on submissions, published versions of
mit-to-accept times of 12, 15, 19, 18, 12, and accepted articles suggest that those limits are
12 months respectively. Of those six, in only often violated. Figure 3 shows the average
one was the 90th percentile of the distribu- pages in each top five journal and the average
tion as short as at the AJAE. in all five journals over the past half century,
We cannot tell whether turnaround times based on data provided by Kosnik (2022).33
in this journal are relatively rapid because It documents the tremendous “page creep”
of the moral suasion in its R&R letters,
32 In at least one case, an author delayed 18 months in
responding to an R&R request from a top five journal. The
30 Email communication, Amy Ando, Coeditor, AJAE, eventual resubmission was quickly rejected by the new edi-
March 19, 2021. tor who was uninterested in the topic.
31 The AJAE has an impact factor of 3.44, almost identi- 33 Adjusting figure 3 for differences in character counts
cal to the average of the six specialized non-theory journals per page across journals and for the changes in page size
listed in footnote 20. that occurred in the AER barely changes this figure.
290 Journal of Economic Literature, Vol. LXII (March 2024)

that has occurred in the profession—a tri- a ­deadline. While there are deadlines in
pling over the half century. Aside from the requests to referees, they are not enforced:
flouting of these limits, incentivizing journals and referees can backload indefinitely.
and authors to drag out the decision process, Monetary incentives merely shift a few
it also sacrifices journal space that might be delayed reports across the margin to qual-
devoted to other authors’ work. Raising the ify for the payment (Hamermesh 1994).
remarkably low acceptance rates (compared Nonmonetary incentives, for example, the
to other economics journals and other social American Economic Review’s or Journal of
science fields) at top ­ economics ­ journals Political Economy’s lists of referees, or free
could be a beneficial result of enforcing lim- journal subscriptions to reward rapid refer-
its on page counts. The page creep in eco- eeing, are unlikely to provide much moti-
nomics journal publishing may be related to vation to overcome procrastination. Public
the increasing lags in editorial decisions, lags shaming of delinquent referees is a possibil-
that might be reduced if page limits were ity, but journals may be unwilling to engage
enforced both ab initio and throughout the in it; and, in any case, it is unclear whether
editorial process. such shaming would reduce delinquency.
Some referees are simply unreliable; since
3.5 Limiting Refereeing/Editing Time
refereeing deadlines are not e­nforceable,
The evidence in section 2 made it clear journal editors may feel stuck with
­
that editor/refereeing lags are not the more delinquents. There is a solution: “fire” the
important contributor to the excessive times delinquent after some short period of nonre-
from submission to acceptance and publi- sponse. If an article is so narrowly focused that
cation at the top five economics journals. It only two or three scholars can provide useful
also demonstrated, however, that conditional comments/recommendations to the editor,
on the number of rounds of back-and-forth, it probably does not belong in a top gener-
additional time spent by referees and editors al-interest journal. A reasonable requirement
has no relationship to the eventual attention is thus that no referee be allowed more than
that an article receives. This suggests that three months to handle an article (a policy
there is room for marginal improvements that is currently implicit and tightly enforced
along this dimension too. by our journal 1, and which explains the low
While the data used here cannot distin- variance in the time articles spend in that
guish between the contributions to publica- journal’s hands). If a referee fails to respond
tion lags of dilatory editors and the referees within that time limit, the editor should
whom they assign, we do know (Hamermesh immediately request a report from another
1994) that most referees who complete referee, if reports are still needed. Given
their assigned task do so quickly, with only the evidence that a large majority of referees
20 percent of them taking more than three complete the task within this time limit, it is
months. The difficulty is that a small fraction unlikely that this constraint would reduce the
are never heard from (5 percent) or quickly supply of useful refereeing talent (as opposed
decline the refereeing request (17 percent). to those who agree to ­referee but fail to do
But almost two-thirds of referees solicited so). The ­general equilibrium effects of such a
agree to do the job and complete it within limit seem minor.
three months. As demonstrated in section 2, 22 per-
The theory of procrastination (see, e.g., cent of articles at top five journals went
Akerlof 1991) suggests that people back- through four submissions/resubmissions,
load completion of tasks until just before back-and-forth with the journal. While
 Hadavand, Hamermesh, and Wilson: Publishing Economics 291

60

Top five
50 AER
ETRCA
JPE

40 QJE
REStud
Pages

30

20

10

0
1969 1974 1979 1984 1989 1994 1999 2004 2009 2014 2018

Figure 3. Average Page Length in Top Five Journals, 1969–2018

Note: Excludes comments/replies/rejoinders and Nobel/presidential addresses. Based on data from Kosnik (2022),
which are available in Hadavand et al. (2024).

these additional rounds were related to clear that, as one former top five editor sug-
some additional subsequent citations, is gested, an initial resubmission will only be
that worth the delay in making research sought if the additional work is “doable” and
more visible and in authors improving their can be ­handled by the author(s) in a reason-
CVs? Journal editors might solicit no more able length of time, as recommended in the
than two resubmissions, with the second previous subsection.
requesting only “cleaning up” and “pol- The editors of most top five journals are
ishing.” If implemented, this innovation paid for their work, with substantial time
would also reduce the incidence of multiple released from teaching and/or monetarily and
rounds of resubmissions that end in rejec- often quite lucratively. (One top five journal
tion. The broader adoption of refereeing pays its editors $51,500 per annum; another
cascades would facilitate this reduction. pays $32,000—with $64,000 to the editor-in-
Admittedly, this recommendation requires chief.) They should be well-paid—their work
that editors exercise judgment when solic- is important and time-consuming. Asking that
iting the first resubmission. They should be they abide by the dictum that they only solicit
292 Journal of Economic Literature, Vol. LXII (March 2024)

resubmissions on papers on which there is a how some innovations instituted in the past
clear path to publication, and therefore exer- 15 years have accelerated the publication
cise more judgment early on, is not unrea- process. While the evidence supporting
sonable. Moreover, given their remuneration, these recommendations comes from data
“sitting on” a paper longer than a month upon describing top five journals, the new AEA
submission/resubmission or upon receipt journals, and one lower-level general jour-
of a sufficient number of referee reports is nal, Economic Inquiry, they are equally
inexcusable. valid at other lower-level journals (whose
decision processes are also distinctly slow).
Some observers might be concerned that
4. Escaping the Low-Level Equilibrium these ­recommendations would increase the
advantages of “insiders,” which are already
The economics profession is in a low-level substantial at some journals (see Bethmann
equilibrium trap, with much longer decision et al. 2023); but it is unclear why speeding
times than any other discipline that validates things up would aid those whose value of
ideas through their acceptance by peer-re- time is ­highest more than those with fewer
viewed journals. Today the lags between professional demands on them.
an article’s acceptance and publication are In all these journals, the burden of improv-
unimportant. Online publication often ing the situation—of putting the econom-
occurs within a few weeks of an article’s ics profession on the same footing as other
acceptance. Even ignoring the now techno- disciplines—rests on editors. They need to
logically irrelevant lag between acceptance change their behavior, to insist that referees
and publication, however, economics pub- behave as gatekeepers rather than coauthors,
lishing remains woefully slower than that in and to be sure that authors respond reason-
other disciplines. ably rapidly to editors’ requests for resub-
The long lags hurt the profession and, as we missions. The low-level equilibrium trap
have shown, are at least partly unrelated to developed because editors let it develop. We
the amount of attention that articles receive will not escape it until editors change how
from other scholars. They have especially they deal with referees and authors.
severe negative impacts on younger schol-
REFERENCES
ars facing tenure/job-security decisions, with
cases where PhD essays are hanging in the Akerlof, George A. 1991. “Procrastination and Obedi-
balance at a journal even when the person’s ence.” American Economic Review 81 (2): 1–19.
tenure case is being considered. In many Angrist, Joshua, Pierre Azoulay, Glenn Ellison, Ryan
Hill, and Susan Feng Lu. 2020. “Inside Job or Deep
institutions junior economists are compared Impact? Extramural Citations and the Influence of
to peers in other disciplines, even in other Economic Scholarship.” Journal of Economic Liter-
social sciences, whose research oeuvres do ature 58 (1): 3–52.
not suffer the same lags in acceptance/pub- Azar, Ofer H. 2007. “The Slowdown in First-Response
lication. Economists making decisions about Times of Economics Journals: Can It Be Beneficial?”
Economic Inquiry 45 (1): 179–87.
their colleagues’ futures understand this Bethmann, Dirk, Felix Bransch, Michael Kvasnicka,
problem, but “higher-level” administrators and Abdolkarim Sadrieh. 2023. “Home Bias in Top
often do not, creating needless stumbling Economics Journals.” IZA Discussion Paper 15965.
blocks to tenure for active junior economists. EQ: Added author and changed date from 2021 to 2023
We have outlined several steps that might Björk, Bo-Christer, and David Solomon. 2013. “The
Publishing Delay in Scholarly Peer-Reviewed Jour-
reduce the time between an article’s submis- nals.” Journal of Informetrics 7 (4): 914–23.
sion and its acceptance, and have indicated
 Hadavand, Hamermesh, and Wilson: Publishing Economics 293

Bonatti, Alessandro, and Johannes Hörner. 2011. “Col- Hamermesh, Daniel S. 2013. “Six Decades of
laborating.” American Economic Review 101 (2): ­Economics Publishing: Who and How?” Journal of
632–63. Economic Literature 51 (1): 162–72.
Bosquet, Clément, and Pierre-Philippe Combes. 2013. Hamermesh, Daniel S. 2018 “Citations in Economics:
“Are Academics Who Publish More Also More Measurement, Uses, and Impacts.” Journal of Eco-
Cited? Individual Determinants of Publication and nomic Literature 56 (1): 115–56.
Citation Records.” Scientometrics 97 (3): 831–57. Heckman, James J., and Sidharth Moktan. 2020. “Publish-
Card, David, and Stefano DellaVigna. 2013. “Nine ing and Promotion in Economics: The Tyranny of the
Facts about Top Journals in Economics.” Journal of Top 5.” Journal of Economic Literature 58 (2): 419–70.
Economic Literature 51 (1): 144–61. Hollis, Aidan. 2001. “Co-authorship and the Output
Checchi, Daniele, Alberto Ciolfi, Gianni De Fraja, of Academic Economists.” Labour Economics 8 (4):
Irene Mazzotta, and Stefano Verzillo. 2021 “Have 503–30.
You Read This? An Empirical Comparison of the Hudson, John. 1996. “Trends in Multi-authored Papers
British REF Peer Review and the Italian VQR Bib- in Economics.” Journal of Economic Perspectives 10
liometric Algorithm.” Economica 88 (352): 1107–29. (3): 153–58.
Clarivate Analytics. 2020. “Journal Citation Reports.” Jones, Benjamin F. 2021. “The Rise of Research Teams:
https://jcr.clarivate.com/ (accessed February 2021). Benefits and Costs in Economics.” Journal of Eco-
Clarivate Analytics. 2021. “Web of Science.” https:// nomic Perspectives 35 (2): 191–216.
clarivate.com/products/scientific-and-academic-re- Kosnik, Lea-Rachel. 2022. “Who Are the More Dismal
search/research-discovery-and-workflow-solutions/ Economists? Gender and Language in Academic
webofscience-platform/ (accessed February 2021). Economics Research.” AEA Papers and Proceedings
Conley, John P., Mario J. Crucini, Robert A. Driskill, 112: 592–96.
and Ali Sina Önder. 2013. “The Effects of Publica- Laband, David N. 1987. “A Qualitative Test of Journal
tion Lags on Life-Cycle Research Productivity in Discrimination against Women.” Eastern Economic
Economics.” Economic Inquiry 51 (2): 1251–76. Journal 13 (2): 149–53.
Ellison, Glenn. 2002. “The Slowdown of the Econom- Laband, David. 1990. “Is There Value-Added from
ics Publishing Process.” Journal of Political Economy the Review Process in Economics? Preliminary Evi-
110 (5): 947–93. dence from Authors.” Quarterly Journal of Econom-
Ferber, Marianne A., and Michael Brün. 2011. “The ics 105 (2): 341–52.
Gender Gap in Citations: Does It Persist?” Feminist Lusher, Lester R., Winnie Yang, and Scott Carrell. 2021
Economics 17 (1): 151–58. “Does Economics Have a Working Paper Problem?”
Hadavand, Aboozar, Daniel S. Hamermesh, and Wesley NBER Working Paper 29153.
W. Wilson. 2020. “Is Scholarly Refereeing Productive McAfee, R. Preston. 2010. “Edifying Editing.” Ameri-
(at the Margin)?” NBER Working Paper 26614. can Economist 55 (1): 1–8.
Hadavand, Aboozar, Daniel Hamermesh, and Wesley Medoff, Marshall H. 2003. “Collaboration and the
Wilson. 2024. “Replication data for: Publishing Eco- Quality of Economics Research.” Labour Economics
nomics: How Slow? Why Slow? Is Slow Productive? 10 (5): 597–608.
How to Fix Slow?” American Economic Association Szenberg, Michael, and Lall Ramrattan, eds. 2014.
[publisher], Inter-university Consortium for Polit- Secrets of Economics Editors. Cambridge, MA: MIT
ical and Social Research [distributor]. https://doi. Press.
org/10.3886/E175941V1. Watson, James D., and Francis H. C. Crick. 1953.
Hamermesh, Daniel S. 1994. “Facts and Myths about “Molecular Structure of Nucleic Acids: A Struc-
Refereeing.” Journal of Economic Perspectives 8 (1): ture for Deoxyribose Nucleic Acid.” Nature 171 (3):
153–63. 737–38.