Authorship

To adequately assess publication counts and to attribute citation counts to individual authors, it is important to recognize contributions made in team-authored publications. Yet the prevailing definitions of authorship are broad and imprecise because of the wide variety of academic fields they need to apply to [33]. As a case in point, the prevailing modus operandi would identify scholars as authors if they made “substantial contributions” [34] to the publication. Admittedly, this definition of authorship is vague and leaves many degrees of freedom for the scholars involved. Consequently, there is a wide discrepancy between what authors should ideally attribute and what authors do attribute.

In the social sciences, this discrepancy is further exemplified by the authorship definitions of large societies whose codes of ethics or conduct state that individuals who contributed substantially to a publication should receive authorship [29–31]. Regrettably, the definition of a substantial contribution remains unclear in those guidelines. To solve this issue, editorial policies at an increasing number of prominent interdisciplinary journals that also publish social science articles have begun to implement more specific guidance on who should receive authorship and who should not. These guidelines are either based on [35,36], or very similar to [37,38], the authorship criteria defined by the International Committee of Medical Journal Editors (ICMJE) [39] (the authorship guidelines of Nature [37] and PNAS [38] are more relaxed as they only require responsibility for submission and one other task). According to the ICMJE, “Authorship credit should be based only on 1) substantial contributions to conception and design, or acquisition of data, or analysis and interpretation of data; [AND] 2) drafting the article or revising it critically for important intellectual content; AND 3) final approval of the version to be published” [39]. The ICMJE revised its criteria in 2013, adding the “agreement to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved” [40] as a fourth criterion. Every researcher fulfilling all four requirements is not only eligible for but must receive authorship [40]. It is not only interdisciplinary journals that refer to the ICMJE for details on authorship: the Committee on Publication Ethics (COPE), a multidisciplinary advisory body on ethical issues to which several large academic institutions, journals, and societies subscribe, advises its members to employ the ICMJE authorship standards as well [41]. This highlights the relevance of the ICMJE authorship criteria in the social sciences. Consequently, we employ these criteria as the benchmark definition in our empirical analysis.

Ghost authorship

Having established the definition of authorship based on the ICMJE criteria, we can elaborate on what constitutes authorship misconduct. The first type we will consider is ghost authorship. A ghost author contributes to a research paper, but, willingly or unwillingly, is not named as an author [42]. Ghost authorship shares with plagiarism the misattribution of intellectual work [43]. Yet, in plagiarism, the originators are usually not aware of someone else stealing their ideas.

Existing research provides three general explanations for the presence and prevalence of ghost authorship. First, pressure from co-authors, can lead to researchers declining or being declined authorship. The rising usage of fractional counts in assessing scholars’ productivity rs constitutes a potential reason for this because this motivates researchers to include as few authors as possible to maximize their own publication counts [44]. In such a case, perhaps the ghostwriter is a subordinate of one of the authors, with little power in the relationship [45]. Second, scientists may voluntarily decline author credits because they perceive the findings in the research as controversial, dubious, or weak; perhaps they fear that the publication of the paper may have negative consequences on their future career [46,47]. Third, researchers may evade authorship to disguise potential conflicts of interest. For example, a freelancer sponsored by a pharmaceutical company may approach life scientists with a biased bundle of articles and ask them to write a research paper based on it [48], hoping the paper will promote the official approval, and/or boost the sales, of a drug. As the freelancers do not receive authorship (and thus are ghost authors), the monetary commitment of the firm stays secret [49].

Honorary authorship

In contrast to leaving out contributors in published work, authorship misconduct may also result from adding individuals to publications that they have contributed to either insubstantially or not at all. This behavior is called honorary authorship [50], gift authorship, or guest authorship [51]. It is beneficial to those receiving it because they can get the credit for the publications without exerting the effort and time needed to conduct research and write articles. As with ghost authorship, pressure and power dynamics explain some of the occurrences of honorary authorship [52]: for example, senior scientists may demand authorship from their subordinates in return for employing them, from their Ph.D. students in return for supervising them, or to either in return for providing funding [53,54]. In other cases, the original authors voluntarily include honorary authors. The inclusion of a well-known researcher into the author list increases the publication chances, especially with single-blind review processes and for high-impact journals [55]. A famous co-author often boosts citations, too, which is beneficial to all authors [56]. Toward these ends, authors might ask well-known experts to co-author their papers without having participated in the research or writing processes [51]. A third type of honorary authorship happens when researchers engage in reciprocal relationships. The general increase in co-authored papers enables researchers to trade co-authorship: a scientist adds only a small, non-authorship-worthy contribution to a research project but still receives author credit; the original author also receives reciprocal co-authorship when the honorary author publishes his or her next paper. This behavior may be especially common within chairs or research groups, where reciprocal proofreading might lead to honorary authorship [57].

Research questions

In the following, we aim to investigate to what extent these findings may generalize across disciplinary boundaries. We, therefore, extend prior work by examining authorship assignments in a field where journals only recently started encouraging scholars to disclose their contributions, where the number of authors on average is manageable and where there is (to the best of our knowledge) limited prior work on honorary and ghost authorship: the social sciences.

We believe this examination to be valuable as the results from the life and natural sciences may provide little guidance for the social sciences due to the many contextual differences in which publishing takes place. For example, the average acceptance rates of papers submitted to the respective field journals are quite different. In the social sciences, journals publish only around 25% of all submissions [58]. In nature and life sciences, journals publish approximately 40 to 50% of all submissions [59]. In the following, we study the prevalence of authorship misconduct using a large-scale survey of social science researchers.

Subsequently, we conduct exploratory research on ghost and honorary authorship using various possible determinants of authorship misassignment. We look at job positions because supervisor-subordinate collaborations seem susceptible to authorship misconduct [60–63]. We also look at time spent in academia, as it relates to the specific tasks that researchers take on [64] and thus might relate to authors’ eligibility according to the ICMJE criteria. (On the subject of tasks, it is worth noting that a recent study surveying “scientists who publish a paper every five days” [65]–which equals 72 or more published papers per year–showed that more than 70% of the respondents did not conduct at least one of the three required tasks of the ICMJE authorship criteria–excluding accountability–in at least every fourth paper.) In addition, we take into account researchers’ gender. Female researchers are arguably among the most vulnerable, as credit often tends to go primarily to their male colleagues. This tendency is known as the Matilda effect, named for the suffragist and abolitionist Matilda Joslyn Gage, who wrote about how women have been denied credit for their achievements throughout history [66]. Finally, we look at culture, as it, too, may influence authorship assignments. Salita [67] discussed the possibility that cultural differences might induce perceptions of authorship that could deviate from the ICMJE criteria.

We thus ask the question of what authors are more likely to confirm ghost or honorary authorship in their published team-authored work. In studying the determinants of these types of misconduct, we provide crucial information that can be used by university officials, journal editors, publishers, et cetera, to implement effective countermeasures.

Distribution

To address our research question, we draw on a large-scale survey of social scientists. We designed the survey in spring 2018 using the online survey tool Qualtrics. In May 2018, we asked colleagues for feedback. After incorporating their feedback, we conducted a test run by sending the survey link to 275 scholars who had presented at least one paper at the 2018 European Accounting Association Annual Congress. Data and feedback from the test run showed no need to adapt the questionnaire further. Therefore, the data gathered in the pilot phase is included in the analysis. We did not obtain approval for the survey from an institutional review board (IRB) as the survey was conducted at a point in time where both authors were employed at the RWTH Aachen University. At the RWTH Aachen University, only life science and psychological experimental studies go to the independent ethics committee. The authors contacted the independent ethics committee of the RWTH Aachen University with a request to review the study. Yet the ethics committee replied that they do not provide ethics oversight for empirical social scientific studies. In fact, neither the RWTH Aachen University, nor German state or federal agencies require or offer ethics oversight for empirical research in the social sciences.

To ensure broad dissemination of the questionnaire among researchers from various fields of research in the social sciences, we selected corresponding authors of published articles in well-known journals as well as of papers presented at conferences organized by large field-specific research societies between January 2007 and June 2018. S8 Table includes a list of the societies and journals. After deleting duplicates, we had a total of 126,480 unique email addresses. A random selection of half of these addresses led to an initial sample of 63,240. These scholars received an email containing a brief explanation of the purpose of the study and a Qualtrics URL link to the questionnaire in late August and early September 2018. The link to the survey was the same for everyone to ensure the anonymity of respondents. We ruled out multiple responses by the same individual through IP address restrictions. After sending out the survey, 15,573 emails automatically bounced back due to the email addresses being no longer in use. The contacted sample, therefore, contains 47,697 valid recipients.

Sample

The dissemination of the survey resulted in 2,817 responses. This constitutes 4.45% of all contacted email addresses and 5.91% of all valid e-mail address. These rates are comparable to other recently conducted non-incentivized online surveys among scientists investigating academic misconduct [68,69]. Moreover, the demographic characteristics of our sample (available in S1 and S2 Tables) are comparable to the summarized database of more than 30 million Zippia.com profiles affiliated as social scientists [70]. For example, the share of women is 33.2% in our study and 37.2% in Zippia.com [70]. Of those who completed the survey, 2,223 were social scientists. We then deleted one respondent with nonsensical answers who claimed to have been working in academia for 100 years despite being only 67 years old. Therefore, our original sample consists of 2,222 respondents. This corresponds to a response rate of 3.51% among all e-mail addresses and 4.66% among valid email addresses. Our drop out rates are comparable to existing research on academic misconduct [71]. Nevertheless, we investigate whether these dropouts induced potential sample selection bias in the robustness checks. However, the samples used for the following analyses differ because some conference attendees had not yet published in journals and some respondents selected “N/A” for one or more responses. Respondents who had not published a journal paper did not receive the questions on their last published paper but were still able to respond to two vignettes in the last section that described projects and asked respondents to assign authorship to the researchers involved. Respondents stating that their last published paper did not include any non-author contributors were excluded when comparing the actual and perceived rate of ghost authors in model 5 of Table 1 because of divide by zero errors.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Regression results of actual and perceived prevalence of ghost and honorary authorship.

https://doi.org/10.1371/journal.pone.0267312.t001

Survey design

The survey consists of three parts. In the first section, respondents answered questions about their demographic and job characteristics. The second part of the survey asked about the distribution of authorship and contributor acknowledgments in the latest published paper that names the respondents as authors. The third section covered two vignettes, as noted above, which described research projects and asked respondents to assign authorship to the researchers involved.

Statistical analysis

We conducted the entire data analysis using Stata 16. For the analyses of authorship assignments in respondents’ last published papers, our first set of dependent variables indicates whether the author list of a paper suffers from Ghost Authorship and/or Honorary Authorship. Our second set of dependent variables employs the exact number of Ghost Authors and Honorary Authors per paper. Our third set includes the indications of the extent of Perceived Ghost Authors and Perceived Honorary Authors in respondents’ last published papers. For the analyses of the vignettes, our binary dependent variables employ respondents’ authorship assignments by each individual respondent.

We employ different empirical estimation strategies for each set of dependent variables. We apply logistic regressions for the dichotomous variables of Ghost Authorship and Honorary Authorship. For Ghost Authors and Honorary Authors, we employ negative binomial regressions due to the overdispersion of the count variables. We utilize Poisson regressions for the equidispersed perception-based dependent variables of Perceived Ghost Authorship and Perceived Honorary Authorship. Last, we employ logistic regression again for the authorship assignments associated with the vignettes.

Actual prevalence of authorship misconduct

Based on the ICMJE authorship criteria, out of 1,878 papers with full data on the authorship tasks, one ghost author participated in the creation of 43 (2.29%) papers and two or more ghost authors participated in the creation of 21 (1.12%) papers (Fig 1A). Moreover, regardless of the definition of authorship, we can clearly identify 34 ghost authors who have contributed to all tasks of a research project but did not receive authorship. Honorary authorship, as defined by the ICMJE requirements, occurs much more frequently, with 418 papers (22.22%) containing one honorary author, 234 papers (12.44%) containing two honorary authors, 107 papers (5.69%) containing three honorary authors, and 57 (3.03%) containing four or more honorary authors (Fig 1B). In addition, regardless of the definition of authorship, we can clearly identify 134 honorary authors who did not contribute at all to a research project on which they were named as authors.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1.

Rates of ghost (A) and honorary (B) authorship. N. of obs. are 1,878 for (A) and 1,881 for (B).

https://doi.org/10.1371/journal.pone.0267312.g001

We compare the rate of ghost authors that we identified based on the ICMJE guidelines to the degree that respondents believed that authorship was inappropriately withheld from one or more contributors. Fig 2A show that social scientists perceive ghost authorship to be more prevalent than it is: the perceived rate of ghost authorship (light blue) exceeds the identified rate of ghost authorship (dark blue). The difference in the means (4.65% for actual and 13.09% for perceived) is significant according to a two-sided t-test (t = -8.0874; df = 812; p<0.00).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2.

Shares of the identified rate of ghost authors and perceived rate of ghost authors (A) as well as the shares identified rate of honorary authors and perceived rate of honorary authors (B). N. of obs. are 813 for (A) (because we only can assess the rate of ghost authors among the papers including at least one non-author contributor) and 1,842 for (B).

https://doi.org/10.1371/journal.pone.0267312.g002

Fig 2B compares the rate of identified honorary authors in each paper to respondents’ assessments of the rate of honorary authors in their last published paper. The difference between the identified (dark blue) and perceived (light blue) occurrences is not as stark as for the perception of ghost authorship (the mean of actual is 23.86% and the mean of perceived is 17.26%). Nevertheless, using a two-sided t-test, we find that the perceived rate of honorary authors is, on average, significantly lower than the identified rate of honorary authors (t = 7.5946; df = 1841; p<0.00).

Potential antecedents to authorship misassignment

To identify the reasons for the mismatches between scholars’ perceptions of the prevalence of authorship issues and the actual occurrence of these issues, we investigate their antecedents and correlates. Table 1 depicts the regression results using a host of explanatory variables. Models 1 and 2 report the results from a logistic regression using a binary dependent variable that takes on the value of one if the researchers’ last papers include at least one ghost author (1) or honorary author (2) based on the ICMJE’s definition of authorship. Models 3 and 4 report the results from a negative binomial regression employing the number of ghost and honorary authors in each paper as dependent variables.

The results show that women do not exhibit more ghost authorship in their papers than men. Yet while the positive coefficient pointing towards women facing honorary authorship more often in their papers is not significant at the conventional p < 0.05 level, this effect is marginally significant at p < 0.1 (S5 Table). Furthermore, we find strong effects that scholars in Anglophone and Continental European countries report fewer ghost and honorary authors than scholars from other world regions. Respondents who are Ph.D. students indicate 51.3% more occurrences of honorary authorship and 127% more occurrences of ghost authorship in their last published paper. In contrast, respondents who are professors indicate fewer occurrences of honorary authors but do not differ significantly from the baseline category of junior faculty members regarding ghost authorship. The more papers a scholar has published in the last three years, the more honorary authors he or she reports, though the coefficients are equally small. Concerning differences across research fields, we find that scholars in economics, finance, and political science are about 40% less likely to report the presence of honorary authors in their last published paper than are members of the baseline group of interdisciplinary social scientists. By contrast, psychologists report even more instances of honorary authors in their published papers than interdisciplinary social scientists do.

Models 5 and 6 depict respondents’ perceived assessments of the rates of ghost and honorary authors in their last published paper as the dependent variables. The results show that higher the identified rates of ghost and honorary authors according to the ICMJE criteria are, the higher also the respondents’ perceived the occurrences of these forms of authorship misconduct. Yet the effect for ghost authors is not significant at the conventional p < 0.05 level but only marginally significant at p < 0.1. Nevertheless, the results further attest to the discrepancy between the perception of authorship misconduct and its actual occurrence due to the relatively low explanatory power (as measured in Pseudo-R²) of the regressions.

Hypothetical assignments of authorship

The results on the prevalence and the perception of authorship misconduct revealed some preliminary evidence that the survey respondents seem to be applying an authorship definition that is fairly wide and does not adhere to a common point of reference such as the ICMJE definition. However, it is unclear whether the occurrences of these authorship issues derive from a lack of knowledge of common authorship criteria or from other considerations (whether conscious or unconscious) not mentioned in the survey questions.

To explore the two explanations further, we randomly assigned respondents into two groups and presented them with two vignettes. Table 2 depicts respondents’ authorship assignments for these two hypothetical scenarios.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Authorship assignments in the vignettes split by treatment group.

https://doi.org/10.1371/journal.pone.0267312.t002

The first vignette lists the same scenario for both groups: A professor and a postdoc collaboratively write a paper and a student assistant supports the data collection. By design, the answers concerning who should be listed as an author between the two groups should not differ. The results indicate that the answers within the groups are nearly identical and the respondents similarly assign authorship. Almost all respondents (917 in Group A and 923 in Group B) award authorship to the professor and the postdoc, in concordance with the ICMJE authorship definition. Yet a substantial number of scholars (284 in the first group and 292 in the second group) deviate from the ICMJE criteria by awarding author credits to the student assistant who did not participate in the writing, revision, or submission process. This finding corroborates the prior conjecture that social scientists appear to have a broader conception of authorship.

Importantly, Models 1, 2, and 3 in Table 3 show that the authorship assessments do not differ significantly between the groups when we control for demographic and job-related factors, as the coefficient of Group is insignificant. This helps to establish a baseline behavior for the subsequent analysis when each group judges slightly different scenarios to explore the misconception of authorship criteria among social scientists in more detail.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Regression results of hypothetical authorship assignments in the vignettes.

https://doi.org/10.1371/journal.pone.0267312.t003

In the second vignette, the assignment of authorship to the primary researcher (a postdoc for Group A and a professor for Group B), who takes over most of the tasks, differs only very slightly in comparison to the assignment of authorship in the first vignette. 950 respondents in Group A and 963 respondents in Group B assign authorship to the primary researcher. These assignments concur with the ICMJE authorship criteria. Yet respondents in both groups award authorship less often to the second researcher (a professor) who only performs the minimum number of tasks (revising the research design, revising the paper, and approving the submission of the paper) that would fulfill the ICMJE authorship criteria. In Group A, 607 respondents assign authorship to the professor. In Group B, 700 respondents assign authorship to the professor. Again, the responses deviate substantially from common conventions of authorship. More importantly, however, they even tend to withhold authorship. Differences between the groups remain highly significant even when controlling for various demographic and job-related factors. Model 5 in Table 3 shows that respondents in Group B were 42.9% more likely to award authorship to the second researcher.

Summing up, we find differences in the authorship assignment practices not only between the two vignettes but also across the two groups. The respondents less often assign authorship to the second-mentioned researcher if the primary researcher was a postdoc rather than a professor.

Robustness of results

We conduct several robustness and endogeneity checks. First, while we are unfortunately not able to conduct a non-respondent analysis due to the preservation of respondents’ anonymity, we analyzed the levels at which respondents disengaged from the survey. We compare the characteristics of those respondents that disengaged from the survey after filling out the insensitive questions on demographic, job and academic positions to the sample respondents. S8 and S9 Tables show the descriptive statistics as well as the results from a Wilcoxon rank sum test (we use this non-parametric test because there are 137 respondents who solely answered the insensitive questions). There exist only two significant differences: Those who only answered to the insensitive questions had on average more co-authors as well as more contributors on their last published paper. This points out that respondents most likely disengaged from the survey because the survey constituted too much work for them but not because of the existence of sensitive questions. On the page following the insensitive questions, respondents had to indicate all of their co-authors’ and contributors’ participation for each task. Consequently, the second stage of the survey was more extensive for participants with more co-authors and/or contributors, thus leading to them quitting more often. However, this deviation does not seem to heavily impact the composition of our sample because our sample average of 2.83 authors per paper (S3 Table) lies within the social scientific field averages [73].

First, we run all models with robust standard errors. The levels of significance remain invariant. Second, we estimate all models using OLS regressions and calculate variance inflation factors (VIFs) to assess whether our models suffer from multicollinearity. All VIFs are below the conservative threshold of 5 [74] and thus we have no reason to believe that our models suffer from multicollinearity. Third, we apply firthlogit, a special form of a logistic regression that considers rare events, to correct for the relatively low number of 64 ghost authorship observations [75]. The results remain invariant. Fourth, the average team size differs across research fields [76]. Therefore, we estimate the same regressions with the percentages of honorary and ghost authors in the supplementary material. The only difference is that psychologists no longer report a higher rate of honorary authors anymore. Fifth, we run all analyses including only respondents of papers with a maximum of five authors and contributors to avoid any issues arising from the approximations of respondents stating that their last papers included more than five authors/contributors. The significance levels and implications do not differ from the findings above. Sixth, to detect potential unobserved heterogeneity within the groups, we create interaction terms with the product of Group and all exploratory variables for Models 4 and 5 in Table 3 and again run logistic regressions. The results are available in S7 Table. The coefficients of Group change only slightly, and the significance levels remain invariant. Last, Model 5 in Table 1 might suffer from sample selection, since honorary authorship increases the number of authors and lowers the number of contributors. The inclusion of only papers with at least one contributor in the analysis of the relationship between the perception and the occurrence of ghost authors may, therefore, suffer from endogeneity. To investigate this issue, we employ a Heckman two-step regression by applying Honorary Authorship as the selection variable. This regression returns an insignificant inverse Mills ratio. Therefore, our results do not suffer from sample selection bias [77].

Implications

Our results highlight a high prevalence of ghost and honorary authorship and a broad deviation in authorship assignments from such as the commonly employed ICMJE criteria. These findings highlight the necessity to introduce authorship criteria that are better tailored to the needs, preferences, and perceptions of social scientists. Hence, we call upon large research societies such as the Academy of Management, the American Economic Association, the American Psychological Association, and the American Sociological Association, as well as on the most prominent publishers such as Elsevier, Taylor & Francis, and Springer [23], to revise their existing guidelines on authorship. The revision should focus on establishing clear, precise, and specific criteria that should be also discussed and taught at their annual meetings. This would increase knowledge and awareness of authorship and would ease the job of journal editors as they could rely on accurate guidelines when making attributions of authorship. In turn, these attributions would lead to fairer comparisons of authors for career decisions. Though, one might account for more multi-dimensional criteria in these comparisons anyhow. As a case in point, Moher et al. [87] highlighted several methods of assessing scientific performance including creativity, openness, transparency, and addressing and solving societal problems that go beyond pure publication counts.

The differences in authorship attributions in our study highlight the need to rethink contribution disclosures. Contribution disclosures allow insights into the workload distributions among author teams. Even if researchers’ understanding of authorship varies, contribution disclosures would give an outside observer a better chance to assess who did what regarding a specific publication. This would not rule out misconduct, of course, but it would make it much harder to add individuals who did not contribute at all. If journals require contribution statements from each author and contributor, scholars will face high coordination efforts to submit factually wrong but congruent contribution disclosure statements. Hence, mandatory contribution disclosure statements raise the barrier for submitting falsified author lists. The introduction of contribution disclosures has already reduced instances of honorary authorship in the life sciences [33]. Moreover, contribution disclosures might also reduce instances of benevolent discrimination, as uneven divisions of efforts are accounted for in the contribution statements.

Of course, the introduction of contribution disclosures is not a panacea. It requires editors and authors to be aware of authorship criteria and the consequences of potential misconduct. To ensure this, we recommend that academic societies, research institutions, and publishers establish social scientific authorship criteria in their guidelines and provide tutorials and workshops on authorship.

Limitations and future research

Every research has to be understood and interpreted in light of its limitations. Perhaps the greatest limitation of this study is that existing authorship criteria in the social sciences are overly vague, leaving researchers with very little guidance. In our work, we applied the ICMJE authorship criteria as a common point of reference to better understand authorship assignments in the social sciences. Although universities and societies should apply these criteria (through associating with COPE, for example), social scientists’ perceptions about what constitutes authorship clearly differ from the ICMJE criteria. We attribute this among others to inherent differences in the research and authorship attribution processes of social and life sciences [21]. Consequently, though we find that ghost and honorary authorship are common, some social sciences researchers might argue that this does not represent misconduct but rather is reflective of common authorship practices in the social sciences.

Also, the results of our study may not necessarily generalize. Our findings reported a relatively low number of identified ghost authors. While the application of firthlogit indicates that the findings are robust and valid, the small number of observations might be obfuscating further effects. The quantitative survey contained sensitive questions that, despite the fact that respondents were assured of anonymity, could lead to the understatement of actual wrongdoings–a common problem in research on questionable research practices and academic misconduct [68]. For this reason, the actual prevalence of ghost and honorary might be higher than what our study uncovered, especially when using survey designs that explicitly deal with sensitive items.

Also, the unequal distribution of researchers among geographical regions and research fields might reduce the applicability of our results. Some fields might exhibit higher or lower levels of ghost and honorary authorship. More field-specific studies could determine the prevalence more precisely.

Several fruitful areas for future research can be derived from this study. First, future research could discuss whether common authorship practices in the social sciences like automatically assigning Ph.D. students first authorship should be continued or replaced by more merit-based mechanisms. Second, comparing research-field-specific authorship criteria could explain why scholars from different disciplines vary in their authorship assessments. Third, closely examining the authorship assignments of extremely prolific social science scholars could clear up doubts on whether they really exhibit higher productivity levels or receive honorary authorship more often. Fourth, a qualitative study surveying researchers who experienced ghost and/or honorary authorship could provide a better understanding of the motivations and consequences of authorship misattribution. Last, the use of anonymity-preserving survey measures such as item-sum techniques [88] could increase respondents’ perceived anonymity and therefore lead to a more accurate assessment of ghost and honorary authorship.