Empirical analysis of factors influencing delay in article acceptance in accounting journals

Using a sample of articles published in top accounting journals, we estimate a multivariate model and find empirical evidence of factors influencing delay in article acceptance. Acceptance delay is measured as the time from article submission to acceptance. We find that article length and journal workload are significantly and positively associated with acceptance delay, whereas the distance to historical aspirations in terms of journal impact factor, top corresponding authors and affiliations, as well as corresponding author affiliation with institutions in the USA or the Netherlands, are significantly and negatively associated with acceptance delay. We also find significant differences in acceptance delays related to specific journal procedures and editorial policies. Finally, we observe that acceptance delay increases over time.


INTRODUCTION
Publication delay is a major concern in the academic community (Nguyen et al., 2015).This delay includes the time between submission and reviewer assignment, time spent reviewing the manuscript, time spent responding to reviewers' comments, and the additional period used by editors to make the acceptance decision.This causes loss of scientific knowledge and hinders knowledge dissemination (Ding & Du, 2023;Yu & Li, 2006).The speed of the publication process is crucial in science, especially in fields and subjects where timeliness and updates are important (Cooke et al., 2016).Moreover, publication delays may have dramatic consequences for the advancement of researchers' academic careers (Allen et al., 2022;Coronel, 2020;Street & Ward, 2019).
Luwel and van Wijk (cited in Shen et al. (2015)) found that journals have significantly reduced the time between processing manuscripts and making articles immediately available through quick editing procedures and early access availability, or preview of accepted articles, which promotes article dissemination after acceptance.Electronic submissions may have also contributed to speeding up the editorial process.Many journals have also shortened acceptance delays by accelerating the review procedure.
However, despite differences across disciplines and journals, evidence exists that overall publication delay has increased over time (Björk & Solomon, 2013), with obvious increases in economics and related fields (Ellison, 2002).According to Bilalli et al. (2021) and as reported by publishers, publication delay is much longer than that required to ensure article quality.The time spent reviewing a manuscript is a major component of publication delay (Lotriet, 2012), and its relative importance varies across journals and knowledge fields (Amat, 2008).The delay in social science, specifically economics and business, is significantly longer than in natural sciences and technology (Björk & Solomon, 2013).The specific acceptance delay is also longer in economics and social sciences (Huisman & Smits, 2017).acceptance, which is the most critical delay in the overall publication process.Accounting is the primary field of analysis, as it experiences long publication delays, even longer than those in economics (Argilés-Bosch & Garcia-Blandon, 2011).Extensive research has been conducted on the review process in the accounting field, which is typically criticized for being excessively long (Adler & Liyanarachchi, 2011;Argilés-Bosch & Garcia-Blandon, 2011;Moizer, 2009;Wood, 2016).However, there is currently no empirical analysis that rigorously quantifies the acceptance delay of accounting journals.
Using a sample of articles published in top accounting journals over 6 years, we find that some manuscript characteristics, such as article length and journal workload, are positively associated with acceptance delay.Conversely, increasing the JIF relative to historical aspiration and some reputational or prestige-related variables are negatively associated with acceptance delay.Top corresponding authors, and their affiliation with top institutions, and the USA or the Netherlands also benefit from shorter delays than other authors and affiliations.Specific editorial policies and procedures also have a significant influence.Additionally, we find that acceptance delay has increased over time.Our results remain robust across different model specifications, including a proxy for article quality measured by future citations.Even when the sample is split into low-and high-article citations, the main results hold.The relationships between acceptance delay and article length, special issues, changes in journal workload, and most prestige-related variables are essentially the same.However, some differences exist between the results in both subsamples, indicating that qualitative factors not included in the model, such as random influences, may account for a substantial portion of the delay.
In the next sections, we formulate the empirical model and describe the data collection and sample, followed by an explanation of the results, a discussion, and concluding remarks.

Key points
• Acceptance delay is higher in social sciences compared to sciences, and it is particularly higher in accounting journals, where it increases over time in accounting journals.
• The results of this article are aligned with previous research suggesting the existence of bias in assessing research and, therefore, in acceptance delay.The assumption that the review process is objective and reliable is rebuttable.
• Prestige factors, such as corresponding author being a top author or affiliated with top institutions, are associated with shorter acceptance delay.
• Country affiliation is associated with acceptance delay.Specifically, we find that US and Netherland affiliation of the corresponding authors are associated with shorter acceptance delays.
• Some objective factors, such as article length and journal workload are positively associated with acceptance delay.

EMPIRICAL MODEL
This study formulates a model based on previous research on publication delay.The model states that acceptance delay (ACCDEL) is influenced by various factors, which can be represented by the following equation: where i refers to a specific article, and ε is the error term.The parameters β, γ, δ, and θ are to be estimated.The equation also includes n, m, and t dummy variables for country, journal, and year, respectively.The Appendix provides detailed definitions of all variables used in the equation.
The number of pages and cited references in the article and words in the title (NUMPAGE, NUMREF, and NUMWTITL, respectively) are indicators of the complexity and length of the article.Longer papers and those with more references are harder to read and, therefore, require larger revisions (Ellison, 2002).
Similarly, a longer title indicates a more specialized and complex paper.
A large number of authors (NUMAUTH) may also be an indicator of complexity, requiring the concurrence of various skills and thus larger reviews, but it may also add quality and achievement to the paper, which would be associated with less review effort.Therefore, the final sign of this variable is uncertain.
The importance of authors and institutions plays an important role in the editor's decision (Cole et al., 1981;Peters & Ceci, 1982).Although reviewers ignore the authors' identities in anonymised reviews, editors are aware of them.The corresponding author is usually assumed to be the main author of the manuscript, and the editor may be biased in accepting articles submitted by top corresponding authors or belonging to important institutions (dummy variables TOPCAUTH and TOPCAFIL, respectively), accepting them more easily and quickly than other manuscripts.This bias may arise because, on the one hand, editors from top journals may be more likely affiliated with important institutions or be themselves important authors, possessing a set of status characteristics and social identity membership influencing their decisions (Jackson & Smith, 1999).On the other hand, the Matthew effect explains that successful individuals or institutions are more likely to be recognized and rewarded, creating a self-perpetuating cycle of success (Bol et al., 2018;Petersen et al., 2011).In both cases, the editors may bias their decisions believing that manuscripts coming from top authors and institutions are of high quality.Therefore, we expect a negative sign for these variables.
Articles published in special issues (SPECISS) are usually subject to a predetermined schedule, especially if manuscripts, as is common, are submitted on dates close to the deadlines.As the authors have fewer incentives to send manuscripts to calls with delayed review terms than to brief ones, the editors of these special issues urge the reviewers to speed up and prioritize these reviews.Manuscripts containing research that has received grants (RESFUND) may have previously undergone more rigorous scrutiny than those with non-granted research, thus requiring less review effort.Therefore, we expect a negative sign for these two variables.
Changes in the number of published articles in a journal with respect to previous year (CHNUMART), control for a series of factors that are likely to influence acceptance delay in opposite directions.The number of published articles may relate to changes in submissions.When submissions increase, the editorial and review workload also increases, which can stress and lengthen the review procedure and the acceptance delay of the accepted manuscripts.On the other hand, more submissions may prompt the editors to take quick and perhaps less accurate decisions to alleviate the editorial workload.An increase in the number of published articles may also be related to higher quality submissions, more lax acceptance criteria, and other possible factors that results in shorter review periods and acceptance delays.Therefore, we have no expected sign for this variable.
The influence of a journal is a criterion followed by authors seeking to publish their research.Although top journals may be efficiently organized, they tend to have more demanding review procedures and higher rejection rates (Aarssen et al., 2008;Sugimoto et al., 2013).The authors may be resigned to follow lengthy review procedures to obtain acceptance from these journals.Fundamentally, journals with the highest JIF values are the most influential (Calcagno et al., 2012;DuBois & Reeb, 2001).Meanwhile, the behavioural theory of the firm states that performance below aspirations stimulates the exploration of new practices, whereas performance above aspirations conforms to their current inertia (Greve, 2003;Iyer & Miller, 2008).In this vein, we anticipate that a lower JIF in comparison to its historical aspiration will encourage shorter acceptance delays as an argument to make the journal more appealing for good research, whereas increasing the JIF will have the opposite effect of conforming to longer delays.We use the historical aspiration as calculated in O' Brien and David (2014) and adapt it to the historical aspiration of JIF (HISTASP) by Shijaku and Ceron Hurtado (2019).We also use the distance between the current performance and historical aspiration (DISTHASP).The two variables are more reliable than the raw JIF and the corresponding changes in JIF with respect to previous year, especially given the variability of accounting JIF.Gul et al. (2021) found a positive correlation between submission to acceptance and JIF.Therefore, we expected positive signs for these two variables.
We also added dummies for country affiliation (the generic variable COUNTRAF in Equation ( 1)) of the corresponding authors to our analysis.This is to control for any potential influence of predetermined opinions about the quality of research conducted in different countries, as well as to account for the impact that the quality of a country's research tradition and expertise may have on acceptance delay.We use the top eight country affiliations found in Argilés-Bosch et al. (2023), while the default and peripheral affiliation includes all other countries.Similar to the bias related to top journals and institutions, we expect to see a similar bias in the case of affiliations in the USA, given that the majority of authors and institutions, as well as the core of the discipline and journals, are located in the USA.Therefore, we anticipate that a USA affiliation, and to a lesser extent, the other eight country dummy variables, will have negative signs.
We also control for specific editorial and journal characteristics with seven journal dummies (out of eight journals), with JOURNAL as the generic variable used in Equation ( 1).We have no specific expectation for these dummies.
Finally, we control for specific temporary effects and likely time trends with dummies for the publication years (YEAR) or the continuous calendar year (YEARCAL), with values ranging from 2015 to 2020.According to some claims about the increasing complexity of articles published in economics (Ellison, 2002) and complaints about the increasing demand for journal requirements in accounting (Fogarty & Markarian, 2007;Moizer, 2009;Wood, 2016), we expect that delay will increase over time.
We winsorize all continuous variables at the 1% level in each tail to avoid influential cases.

DATA AND SAMPLE
We selected journals that have been ranked in the top quartile (Q1) of the Journal Citations Report (JCR) in Web of Science (WoS) for the last 3 years with available data (2017)(2018)(2019) when we started this research in April 2021, and which have the necessary data available on their websites, including submission and acceptance dates.Eight journals meet this condition, and their names, abbreviations and corresponding dummies are provided in Panel B of the Appendix.While other studies may have used a different list of top accounting journals (see, e.g., Bujaki & Mcconomy, 2017), our selection is based on objective verifiable data on top journals in JCR, with the necessary data for our study.We scraped the data from their websites for the last 6 years (2015-2020) using Python libraries.Six years of data are necessary to calculate the 5-year JIF index.We also downloaded the impact factor data of these

RESULTS
As indicated by the Breusch-Pagan/Cook-Weisberg test, heteroscedasticity is present in our model, so we perform estimations with robust standard errors.
Column (1) of Table 4 presents the estimations of the base model, with an R-squared value of 16.2%.Objective article characteristics, such as NUMPAGE, exhibit a positive and significant coefficient, with the same sign as CHNUMART, indicating the prevailing workload effect on this variable.Special issues and variables related to top corresponding authors and affiliations are related to lower acceptance delays, as evidenced by the significant negative coefficients of SPECISS, TOPCAUTH, and TOPCAFIL.
Although HISTASP is not statistically significant, the negative coefficient of DISTHASP is contrary to expectations.Our findings suggest that changes in journal efficiency, particularly in acceptance delay, may influence changes in the JIF, rather than active editorial policies influencing acceptance delay.Shorter publication delays may result in higher-quality manuscripts, which can raise the JIF and vice versa.
The corresponding author's affiliations with institutions in the USA and the Netherlands are significantly associated with shorter acceptance delays, as expected.However, none of the other six dummy variables used for country affiliation show significant shorter acceptance delays than the default variable.
Our results indicate that specific journal editorial policies and review procedures influence acceptance delay, with JAR, AAJ, and BAR significantly associated with lower delay, while TAR and AOS are significantly associated with more delay than the default JAE.
As expected, all year dummies show positive and significant coefficients, indicating higher delays compared with the default first year 2015.2).
We re-estimate all three columns in Table 4 removing HISTASP and DISHASP and instead incorporating them into the untransformed lagged JIF and the changes from current JIF with respect to the lagged value.The results, which are not tabulated, are essentially the same as those in Table 4, except that the change in JIF is not significant (at p < 0.1) in Column (3).
The negative relationship between acceptance delay and TOPCAFIL and TOPCAUTH, including a USA affiliation of the corresponding author may be due to editor bias in accepting articles.However, an alternative interpretation is that top authors and authors affiliated with top institutions are more skilled in research, producing better manuscripts that require less review time.A similar situation arises when the corresponding authors are affiliated with USA institutions.
We assume that better articles will receive more citations after publication, and we control for this characteristic by adding a variable to our model that measures citations received by an article relative to the number of days that it has been available since publication or early access (CITESPERDAY).Table 5 presents the corresponding estimation results.As expected, the coefficient of CITESPERDAY is significant and negative.Assuming that the reason for this coefficient is that higher-quality manuscripts may require less review attention and, therefore, less review delay, after controlling for manuscript quality, most results are essentially the same as those displayed in Table 4. Variables measuring objective manuscript characteristics, journal workload, distance to historical aspirations, specific editorial policies, and dummies for years maintain signs and levels of significance.The coefficient of the variable YEARCAL has become insignificant.More importantly, for this analysis, the coefficients of TOPCAFIL, TOPCAUTH, USA, and NETHERLAND remain negative and significant, suggesting that after controlling for manuscript quality, the relationships probably reflect the influence of editors' bias in acceptance decisions.
To perform analyses with more homogenous samples of article quality, we split the sample by the median of CITESPERDAY, and repeated estimations for both subsamples.Results are presented in Table 6.The coefficients of TOPCAFIL and TOPCAUTH remain negative and significant, but with slight differences.
TOPCAFIL in the subsample of citations below the median is insignificant (p < 0.1).Moreover, the coefficient of a USA     affiliation is insignificant for the subsample of citations above the median.The coefficient of NETHERLAND is negative and significant in both columns.These estimations essentially reinforce the previous results regarding the existence of editorial bias in acceptance delay caused by corresponding authors' characteristics and affiliations.However, there are differences in the significance levels of the coefficients of some other variables between the two subsamples.The results are volatile, suggesting that other factors, whether small or big, and perhaps complex, including even qualitative or random factors, could account for a substantial portion of the delay.The smaller size of the subsamples compared with the whole sample may also influence the loss of significance of some variables.

DISCUSSION
Although the negative associations between acceptance delay and top authors and institutions may be controversial, and such relationships may be attributed to high-quality research produced by these authors and institutions, the additional analyses performed with a variable controlling for article quality and with homogeneous samples suggest the existence of bias in assessing research, and therefore in acceptance delay.Some previous research (Cole et al., 1981;Inglis & Mejia-Ramos, 2009;Peters & Ceci, 1982) is in accordance with our findings and casts doubt on the assumption that the review process is objective and reliable.
Even though random factors may also play an important role, the manuscripts from top authors and institutions may be less rigorously evaluated by editors than those of less prestigious authors and institutions.Editors and reviewers may perceive high current or potential quality in the manuscripts submitted by top corresponding authors and institutions, making them lenient in their judgements and sensitive to avoiding false-negative evaluations by rejecting and delaying such manuscripts.The cut-off point for acceptance and the subsequent delay may consequently be shortened.On the other hand, editors may have more reservations about accepting manuscripts submitted by less eminent authors or institutions, requiring more evidence, tests, and rigorous statements, which produce lengthier reviews.Editors may be concerned about avoiding the false-positive error of accepting flawed research in these cases.The implications of such behaviour are significant.If editors are less rigorous towards prestigious institutions and authors, the outcome may be the publication of lower-quality articles.Given the limited space for publication, the acceptance of manuscripts submitted by top authors and institutions would be at the expense of less known authors and institutions, which will have fewer opportunities, even though they may be of higher quality and deserve publication.Moreover, manuscripts submitted by authors and institutions perceived to be less important, still receiving acceptance, will bear longer acceptance delays and subsequently fewer opportunities to engage in new research projects.Such behaviour is unfair, does not allow for equal career advancement opportunities, and is detrimental to knowledge advancement and dissemination.
Ensuring unbiased decisions is not the only recommendation for editors.Acceptance delay is also significantly related to journal characteristics.Therefore, journals may consider improving the efficiency of the review process by providing detailed reviewer guidelines, encouraging reviewers to provide timely and constructive feedback, as well as expanding the pool of potential reviewers to avoid overload of a few individuals, which may reduce acceptance delays.
However, article characteristics, such as length and quality, also influence acceptance delay.Authors should pay close attention to presenting their research in a structured and logical way, using clear and concise writing to highlight the contributions, research question, key findings and the overall quality of the research performed without unnecessarily enlarging the paper.This paper contributes to the almost non-existent research on factors influencing publication and acceptance delays.As mentioned, to the best of our knowledge, few previous studies have performed correlation analyses between publication delay and JIF (Khosrowjerdi et al., 2011;Pautasso & Schäfer, 2010;Shah et al., 2016) or article citation (Fiala et al., 2016;Lin et al., 2016;Shen et al., 2015), whereas only two studies have analysed factors influencing or related to publication delay.One of these studies (Lotriet, 2012) performed a qualitative analysis, whereas Ellison (Ellison, 2002) performed a quantitative multivariate analysis.In this vein, we also contribute methodologically with a thorough multivariate analysis measuring the influence of various factors.Furthermore, we contribute to the accounting and business fields, where no previous study has performed a similar study despite their extant long publication and acceptance delays.

CONCLUSIONS
This study conducts an empirical analysis of the factors influencing acceptance delays in academic journals, using a sample of top accounting journals.
We found that some manuscript-related factors are associated with acceptance delays.Specifically, longer manuscripts tend to result in longer acceptance delays, while special issues and article quality are associated with shorter acceptance delays.
Journal-related factors are also associated with significant differences in acceptance delays, revealing differences in editorial policies and dynamics.Specifically, we find that articles published in TAR and AOS have longer acceptance periods, while those published in JAR, AAAJ, and BAR have shorter acceptance delays than the default journal, JAE.CPA and MAR are journals that do not present significantly different delays.While there is weak association between the historical aspiration of JIF and acceptance delay, the distance to historical aspiration is negatively associated with such delay, suggesting that changes in editorial policies influence changes in JIF.The number of published articles is also associated with acceptance delay, suggesting a prevailing workload effect.
We also confirm previous findings of an upward trend in acceptance delay over time.
An important finding is that author and institution-related factors persistently affect acceptance delay.Specifically, top authors and institutions, as well as those with affiliations in the USA and the Netherlands, benefit from significantly shorter delays, suggesting the existence of editors' bias related to social identity and the Matthew effect.These results are robust to model specifications, controls for proxies for research quality, and estimations with subsamples of homogeneous articles in terms of quality.
These findings can be used by researchers to better understand the publication process and improve their chances of timely publication.They also provide useful insights for editors to improve the review process and enhance the dissemination of academic research, with special attention to accounting research.Moreover, they highlight the need for continued research in this area.
We acknowledge some inconsistencies in our results.The delays that we study appear to behave randomly, to some extent, or depend on complex pieces of qualitative and as yet unobserved factors that are difficult to model using quantitative variables in an equation.Conceivably, qualitative analysis would provide insights into this phenomenon.Qualitative analysis may also provide more accurate insights for a clear distinction between the influence of quality and bias on acceptance delay.
An additional limitation of this study is that it only considers papers accepted for publication.Papers that were rejected are not available in our database, and consequently, factors affecting delays in articles submitted but not accepted are not studied in this research.Moreover, considering that manuscripts may be rejected by several journals before acceptance, the final acceptance lag may be considerably longer, a fact that we do not analyse in this research.These are issues deserving future research.

YEAR
Generic variable for the publication year, specified in the following six variables

Y2015
The article was published in 2015 (the default year)

Y2016
The article was published in 2016

Y2017
The article was published in 2017

Y2018
The article was published in 2018

Y2019
The article was published in 2019

Y2020
The article was published in 2020 YEARCAL Continuous calendar year with values ranging from 2015 to 2020 CITESPERDAY Number of citations received relative to the number of days that the article has been available since publication or early access number of articles published by the journal in the current year with respect to previous yearHISTASPHistorical aspiration of JIF, calculated as, where JIF is the journal impact factor.DISTHASPDistance between the current impact factor and HISTASP, calculated as DISTHASP t = IF t À HISTASP t Panel B: Dummy variables: Indicator variables that are equalin to 1 if the observation meets the characteristic, and 0 otherwiseTOPCAFILThe corresponding author is affiliated with one of the top 25 institutions of the articles published between 2015 and 2020 in the 8 journals of the sample, calculated using the sample dataTOPCAUTHThe corresponding author is one of the top 66 authors of the articles published between 2015 and 2020 in the eight journals of the sample, calculated using WoS recordsSPECISSThe article was published in a special issueRESFUNDThe research published in the article received research funding COUNTRAF Generic variable for country affiliation of the corresponding author.It is specified in the following nine dummiesUSAThe corresponding author is affiliated with a USA institutionCHINAThe corresponding author is affiliated with a Chinese institutionCANADAThe corresponding author is affiliated with a Canadian institutionUKThe corresponding author is affiliated with a UK institutionAUSTRALIAThe corresponding author is affiliated with an Australian institutionNETHERLANDThe corresponding author is affiliated with a Dutch institutionSINGAPOREThe corresponding author is affiliated with a Singaporean institutionGERMANYThe corresponding author is affiliated with a German institutionOTHERThe default country affiliation.the corresponding author is affiliated with an institution from a country other than those specifiedJOURNALGeneric variable that indicates the journal that published the article.It is specified in the following eight dummies JAR The article is published in the Journal of Accounting Research AAAJ The article is published in the Accounting Auditing and Accountability Journal TAR The article is published in The Accounting Review CPA The article is published in the Critical Perspectives on Accounting JAE The article is published in the Journal of Accounting and Economics (the default variable) BAR The article is published in the British Accounting Review MAR The article is published in the Management Accounting Research AOS The article is published in the Accounting Organizations and Society 531 Factors influencing article acceptance delay Learned Publishing 2023; 36: 517-532 © 2023 The Authors.Learned Publishing published by John Wiley & Sons Ltd on behalf of ALPSP.www.learned-publishing.org Panel B: Dummy variables: Indicator variables that are equalin to 1 if the observation meets the characteristic, and 0 otherwise

TABLE 1
Number of observations and acceptance delay by journal.Panel B: Median (mean in the last row for all years) number of months from submission to acceptance Panel A: Number of articles with data on submission and acceptance dates (yearly mean in the last column) www.learned-publishing.org © 2023 The Authors.Learned Publishing published by John Wiley & Sons Ltd on behalf of ALPSP.Learned Publishing 2023; 36: 517-532
cles.Panel B of Table1shows the winsorized median acceptance delay in months by journal and year.Three of the eight journals (JAR, TAR, and AOS) increased the delay from the first to the last year; two of them decreased (JAE and AAAJ; however the latter presents only 2 years of data), and three showed little change (CPA, BAR, and MAR).However, some journals present volatile figures over the years, and even slightly U-shaped or inverted U-shaped patterns.Therefore, the overall year medians do not provide conclusive results.The substantially higher mean values, with respect to median values, suggest the existence of high maximum values despite winsorization.Panel A inTable 2 presents descriptive statistics for the winsorized continuous variables, while Panel B presents those for the categorical variables.The mean acceptance delay of 650.68 days in Table 2 corresponds with the 21.7 months in Panel B of Table 1.Among the continuous variables, CHNUMART and DISTHASP exhibit the highest volatility (see Panel A).The mean HISTASP of all the journals in our sample is 2.99.As can be

Table 4
(Sheather, 2009)2005) with year dummies replaced by a continuous calendar year variable (YEARCAL) in Column (2).The significantly positive coefficient of this variable suggests an upward trend over the study period.The results for the remaining variables are essentially the same as those in Column (1).Despite the Pearson coefficients not being high, the variable inflation factors (VIF) are concerning, with values as high as 9.34 or 6.83 for dummies Y2020 and Y2019, respectively, which are between the conventionally recommended thresholds of 10(Kutner et al., 2005)and 5(Sheather, 2009), indicating likely collinearity problems.Even with the estimation in Column (2) using the continuous variable YEARCAL, the VIF is still high at 5.15 and 5.95 for this variable and DISTHASP, respectively.Therefore, we removed all YEAR variables and estimated the equation, with the results displayed in Column(3) of Table4.The maximum VIF is 3.59 for the variable NUMPAGE, which eliminates any concerns for collinearity.The results in this column are essentially the same as those in

TABLE 5
Factors that influence acceptance delay, with the addition of a proxy for article quality in Equation (1).

TABLE 6
Factors that influence acceptance delay.Estimates are based on the sample split by median CITESPERDAY.