How Does A Seminal Article in Public Administration Diffuse and Influence the Field? Bibliometric Methods and the Case of Hood's “A Public Management For All Seasons?”

ABSTRACT:

Bibliographic methods offer an advanced microscope-like tool to probe specific fields of inquiry. In this article, we use these methodological tools to demonstrate how a seminal article diffused and influenced the intellectual structure of the public administration (PA) field. We do this by conducting citation (breadth and depth), co-citation (relationships between citing authors and journals), and citation network analysis (network of the most-cited authors citing the focal article) of Hood’s 1991 article “A Public Management for All Seasons?” The findings from the citation analysis of the 949 Hood-citing articles in the Web of Science suggest that Hood is typically studied within PA. Co-citation and citation network analysis points to continuity of themes and topics examined, but also identifies departures, divergences, alternate timeframes, and geographical associations that differ from Hood’s challenges and commentary on the evolution of the field. The extensive opportunities that bibliometric analysis presents for the discipline are discussed.

Notes

1 Bibliometrics include basic citation analysis to more advanced techniques, including bibliographic coupling, co-citation analysis, citation network analysis, co-author analysis, co-word analysis, the application of visualization techniques to quantify and portray different patterns of citation (e.g., overlay visualization using different color gradients to represent old versus new or highly versus least cited themes), and topic modeling.

2 This study is focused on citation to Hood (1991) and, therefore, does not capture discussion of Hood’s work on topics such as the administrative state, public service bargains, and regulation.

3 An example of data cleaning was dealing with data errors of authors’ names. Before we visualized the co-citation analysis, we conducted data pre-processing because bibliographic sources often contain errors, such as duplication or misspelling in the author’s names, the journal titles, or the reference lists. We examined the author names after plotting the initial co-citation networks, and found that the names of seven authors (i.e., Rhodes, Hodge, Boyne, Osborne, Pollitt, Warner, and Hood) were not used consistently. For instance, there were four versions for the author “Rhodes”: “Rhodes R. A. W,” “Rhodes R,” “Rhodes RAW,” and “Rhodes R.A.W.” To mitigate this difficulty, we used a simple programming technique that consisted of a dictionary for replacing all variant names of an author with a uniform version of the name. Therefore, all versions of Rhodes’ name were re-written consistently into one form: “rhodes, raw” (using lowercase letters). This process enhanced the data validity and improved the quality of visualization for the co-citation (and citation network) analysis.

4 We note that citation trends over time commonly show an upwards trajectory as the numbers of journals and scholars have grown, and practice of citation now carries more weight in the social sciences. However, this growth is typically consistent across disciplines. We note that articles citing Hood (1991) were found in 57 categories (or 22.89% of 249 categories made available by WoS in 2016) in the 1992 to 2016 period, as compared to 32 categories (or 12.85%) in the 1992–2004 period.

5 To perform co-citation analysis, a co-citation matrix was created from the “local” WoS dataset. This was done by drawing a connection between two articles if they are cited together by a number of other articles.

6 First authors are used to make the visualization more parsimonious. Otherwise, the co-citation visualization is not decipherable.

7 The citation networks can be depicted with or without the target article (i.e., Hood 1991). To aid understanding and better convey the meaning of author co-citation of a targeted cited article, we placed Hood (1991) at the center of the network.

8 Human coding reduced the nine original clusters that were computationally generated to seven.

9 Our approach is similar to Van Eck and Waltman (2014), who suggested that focusing on around 40 to 70 most important citations is optimal, and offers interpretable outputs.

10 We note that this produces a bias against more recent publications.

11 The remaining 18 articles did not fall into meaningful clusters.

12 We thank one of the anonymous reviewers for raising the issue of stock citations. Given the nature of the article, which primarily relies on quantitative approach to bibliometrics, we do not offer a constructivist/interpretivist explanation of the possibility of the stock citation problem. Consequently, future research can better integrate quantitative and constructivist approaches to doing bibliometrics.

Additional information

Notes on contributors

Yanto Chandra

Yanto Chandra ([email protected]) is Associate Professor at the Department of Public Policy, City University of Hong Kong (CityU). Chandra’s research focuses on social entrepreneurship, social innovation, impact investment, and nonprofits, particularly their strategy, governance, measurement, and their interactions with the public and social policy contexts. Methods of interest include computer-aided text analysis, corpus linguistics, text mining, and data science using R. His research has been published in, among others, World Development, Journal of Business Venturing, and Journal of International Business Studies.

Richard M. Walker

Richard M. Walker ([email protected]) is Chair Professor in Public Management in the Department of Public Policy, Director of the Laboratory for Public Management and Policy, and Associate Dean (Research and Postgraduate Studies) in the College of Liberal Arts and Social Sciences, City University of Hong Kong. Richard’s research interests are in the performance of public service organizations, innovation and performance information use, and the application of research methodologies that develop theory for public management.