Cooperation for Publication? An Analysis of Co-authorship Patterns in Leading Accounting Journals

Abstract

This paper examines patterns, antecedents, and effects of (international) cooperation in accounting academia based on 7105 papers published in 15 leading accounting journals. In particular, we investigate the dissemination of different forms of cooperation, identify author characteristics that are related to the propensity of cooperation and analyze whether cooperation is associated with research performance (in terms of research impact and output). We find that scholars from Asian countries tend to be more heavily involved in international cooperation than researchers from most European countries and the USA. A Ph.D. from a leading school, a scholar's previous publication experience and a past appointment as editor or editorial board member are positively associated with the propensity for cooperation, while a researcher's current affiliation has only limited impact. Surprisingly, our findings show that cooperation is not related to a greater research impact as measured by citation numbers per paper. Finally, we find a significant negative relationship between a scholar's share of co-authored papers and his or her research output in leading accounting journals as measured by the weighted numbers of papers per author.

Acknowledgements

We thank Laurence van Lent (the editor) and an anonymous reviewer as well as Christof Beuselinck, Hicham Daher and Timo Hövelborn for insightful comments. We also thank Mathias Hennemann, Ute Meretz and Julia Sartor for able research assistance and are grateful for helpful comments from participants at the 8th Conference on New Directions in Management Accounting (Brussels, 2012) and at the 36th European Accounting Association Annual Congress (Paris, 2013).

Notes

¹These journals are (in alphabetical order): Abacus (ABA), Accounting and Business Research (ABR), Accounting and Finance (AaF), the Accounting, Auditing and Accountability Journal (AAAJ), Accounting, Organizations and Society (AOS), Auditing: A Journal of Practice & Theory (AUD), Contemporary Accounting Research (CAR), the European Accounting Review (EAR), the International Journal of Auditing (IJA), the Journal of Accounting and Economics (JAE), the Journal of Accounting Research (JAR), the Journal of Management Accounting Research (JMAR), Management Accounting Research (MAR), the Review of Accounting Studies (RASt), and The Accounting Review (TAR).

²The study of journal publications implies that we analyze the output of the journal review processes and neglect the input in the form of submissions. As journals do not provide comprehensive information on papers submitted (e.g. Lukka & Kasanen, 1996), a differentiated analysis of the cooperation patterns of papers submitted and accepted does not appear to be feasible. Thus, we cannot exclude the possibility that our findings are somewhat distorted in a particular direction. For instance, the editorial reports of TAR indicate that the submission rates of non-US scholars are higher than the corresponding acceptance rates (e.g. Evans, 2014; Kachelmeier, 2011). However, the difference between submission and acceptance rates for various non-US regions is approximately proportionate to the overall difference between non-US submission and acceptance rates (Kachelmeier, 2011). Furthermore, we are comparing ‘successful’ co-authorships with ‘successful’ single-authorships – that is, research projects that resulted in a paper published in a leading accounting journal – and we are not aware of any evidence on differences between the acceptance rates of co- and single-authorships. Against this background, we argue that our focus on accepted papers allows for valid insights regarding cooperation in accounting research.

³Given that English is by implicit agreement the common language in the academic accounting community (Lukka & Kasanen, 1996) and in accordance with previous studies (e.g. Chan et al., 2013; Raffournier & Schatt, 2010), we focus on English-language journals. As there are important non-English-language accounting journals in countries such as France and Spain (Chan, Chen, & Cheng, 2006), we acknowledge this focus as a limitation of our study. Nevertheless, we expect that our approach refers to a set of journals that represent primary targets for publication and sources for a wide international audience among accounting academics.

⁴These enquiries were based on the Hasselback Accounting Faculty Directory (2013), the BYU (Brigham Young University) Marriott School accounting research rankings, web searches, and personal e-mail correspondence (Qu, Ding, & Lukasewich, 2009). Our database includes 5356 individual authors, equalling 14,704 name entries, as numerous authors have published multiple papers in the journals and period studied. We identified DDGIs – or the information that an author does not hold a Ph.D. – for 95.4% of the 14,704 name entries included in our data. Our findings do not change qualitatively, irrespective of whether we consider all papers – including those with missing information on the DDGI of individual authors or only those for which complete information is available.

⁵Note that we add a subscript to each variable indicating whether a variable refers to countries (subscript i), papers (subscript j), or authors (subscript m).

⁶We consider a time lag of five years for two reasons: on the one hand, it appears sufficiently long to approximate an author's previous publication experience. On the other hand, we assume that the number of the most recently published papers is of greater interest to researchers who decide to enter into cooperation. Therefore, the number of recently published papers by a potential collaborator may constitute an antecedent for different forms of cooperation. We consider different time spans comprising three and seven years to check the robustness of our findings and find qualitatively similar results (see also Sections 4.2 and 4.3).

⁷To check robustness, we consider a time lag of one or three years. As reported in Sections 4.2 and 4.3, these alternative variable definitions do not alter our conclusions substantially.

⁸Differences may emerge, for instance, due to different sizes of academic communities focusing on an accounting sub-discipline.

⁹We acknowledge that research impact is not the only dimension of a paper's impact. We consider, for instance, a paper's relevance for practice to be another dimension. However, our focus corresponds with the high importance attached to research impact in the context of research performance evaluation (e.g. Chan et al., 2009; ter Bogt & Scapens, 2012).

¹⁰The assumed association between a paper's research impact and citation numbers is subject to limitations, as ‘halo’ effects may occur when papers are cited because of an author's popularity or for negative reasons (Beattie & Goodacre, 2006; Moosa, 2011; Reinstein, Hasselback, Riley, & Sinason, 2011). However, Van Campenhout and Van Caneghem (2010) analyze the contributions of papers published in EAR and find that papers are cited because of their contribution rather than for author characteristics. In this way, they provide some justification for the considerable importance of citations in the context of research performance evaluation, as they are frequently used to rank academic journals, individuals or academic departments (Ballas & Theoharakis, 2003; Brown, 1996; Swanson et al., 2007).

¹¹SSCI provides citation numbers for 4313 papers in the journals and period studied, equalling 60.7% of the 7105 papers that are subject to our main analyses. We find that citation numbers from Google Scholar and SSCI are strongly correlated (ρ = 0.90; p = .01 (two-tailed)). We report findings from the supplementary analysis in Section 4.3.

¹²For the regression, we create dummy variables for each form of cooperation. Therefore, FORM_0_j equals 1 in case of single-authorships and 0 otherwise; FORM_1_j (FORM_2_j) equals 1 in case of national (international) collaboration and 0 otherwise.

¹³For our analyses, we assume that no common DDGI was involved if we were unable to identify the university from which an author received his or her Ph.D. and the remaining authors graduated from different universities.

¹⁴Given that some journals (e.g. AOS, JAR, TAR) publish articles in alphabetical order according to the last name of the first author, we manually checked the journal issues that are part of our data and set LEAD_ARTICLE_j to 0 for issues in which the running order follows an alphabetical order.

¹⁵As this variable assumes an even workload among co-authors, we compute OUTPUT_UNWEIGHTED_m. This variable does not factor co-authorships in and accumulates a scholar's papers published in the journals and period studied irrespective of the number of co-authors for supplementary analysis.

¹⁶In line with Hesford et al. (2007), we distinguish between analytical discussions and literature reviews: While literature reviews include syntheses and critiques of prior literature, analytical discussions provide conceptual frameworks that combine insights from multiple sources, ‘such as empirical facts, theoretical or practical observations, prior literature (in other areas or disciplines), supplemented with the authors’ own synthesis and perspectives’ (p. 7).

¹⁷Note that the following analyses focus on countries with at least 40 unweighted papers published in our data.

¹⁸While Hong Kong is a special administrative region of China, we follow related studies and consider Hong Kong a de facto country as the legal and political system differs considerably from that of China (Chan et al., 2007; Raffournier & Schatt, 2010). However, if we subsume Hong Kong under China for checking robustness, the FCA_RATIO_i for China equals 63.0%, while its COOP_INDEX_i is 65.6%. Thus, China remains one of the countries whose researchers are most strongly involved in international cooperation.

¹⁹Note that we focus on constellations of countries that have spawned at least 15 jointly written papers.

²⁰In the multinomial logistic regression, FORM_j with the attribution ‘0’ (single-authorship) serves as a reference category.

²¹Our findings neither change substantively when we rerun the regression using a Tobit model nor if the definition of LEAD_AFFIL_m or LEAD_PHD_m relies on the top-10 or top-20 lists instead of the top-50 list by Chan et al. (2007). However, in the latter cases, we detect that the coefficients for LEAD_PHD_m become significant.