DIFFUSION STUDIES IN POLITICAL SCIENCE

We begin with a broad definition: diffusion occurs when one government’s choices are influenced by the choices of other governments. This definition encompasses a wide range of topics, including the spread of conflict across borders, socialization and the spread of norms facilitated by international organizations, and the spread of specific policies or governmental structures across polities, to name some of the most common categories. Using a broad set of search terms, we built a database of all 781 diffusion articles published between 1958 and 2008 in 53 top political science journals.^{Footnote 1}

We then coded each article using a number of criteria, including the subfield of study and “what is diffusing”—that is, the feature of politics that is diffusing (e.g., conflict, norms, and policies). To identify the subfield of study, we categorized each article as AP, CP, IR, or other, based on the subject matter, the author’s main subfield affiliation, and the journal in which the article was published. Two of us separately categorized each article, and the third author resolved any differences. For “what is diffusing,” the largest category by far was “policy,” with substantial numbers of studies on norms or ideas, conflict, democracy or other forms of government, innovations (such as technology), and conditions (such as convergence of economic or environmental outcomes).

We then collected information on which of these diffusion articles cited one another, thereby forming the basis for a network analysis. The details of our qualitative and quantitative assessments of this literature are offered elsewhere (Graham, Shipan, and Volden Reference Graham, Shipan and Volden2013; see also Shipan and Volden Reference Shipan and Volden2012). For our purposes here, we illustrate the connections of these studies across pairs of political science subfields in figures 1figures –figures 3.^{Footnote 2}

Figure 1 Diffusion Studies in American and Comparative Politics

Figure 2 Diffusion Studies in American Politics and International Relations

Figure 3 Diffusion Studies in Comparative Politics and International Relations

Figure 1, which includes articles from AP and CP, shows a tight central cluster of AP articles focused on policy diffusion. These works tend to focus on the spread of policies across the American states (with some additional studies of localities and of the centralization or decentralization of policy making); and they cite one another extensively, as represented by their many connections and close proximity. In contrast, the diffusion studies in CP are more scattered. Policy and norm diffusion studies within CP are interspersed with one another and are loosely connected to the AP literature.^{Footnote 3} The CP literature on the spread of democracy and other governmental forms, shown in the upper left of figure 1, is mostly separate from other CP diffusion literatures as well as AP policy diffusion studies.

Figure 2, which focuses on studies in AP and IR, again shows a tight cluster of AP articles about policy diffusion. The IR literature also features fairly tight clusters of articles, studying the diffusion of conflict, norms, and policy. There is some overlap between studies of norms and policy diffusion within IR, but very limited overlap with AP in these areas, even though both subfields extensively examine policy diffusion. Notably, policy diffusion research in IR is more closely connected to AP diffusion studies through the IR work on norms than directly. This illustrates a substantial disconnect, indeed.

Figure 3, which depicts the literatures in CP and IR, presents the greatest overlap across subfields. In particular, the CP and IR literatures on the diffusion of policy and norms are completely interspersed, and we were unable to isolate distinct clusters between CP and IR on these topics. As in figure 2, however, once again there is a separate cluster of IR studies that investigate conflict. More generally, relative to the AP clusters shown in figures 1 and 2, the nodes here are spread farther apart, indicating fewer citation connections across articles on average.

Together, these three figures show patterns both across and within subfields. We find that certain topics—for example, democratization in CP or conflict in IR—tend to be studied in isolation from topics in other subfields (and even, to a lesser extent, from studies within the same subfield). At the same time, for the topics of policy diffusion and norm diffusion, overlaps exist, indicating that there are substantial opportunities for conversations across subfield lines. Now, we turn our focus specifically to these two types of diffusion.

In tables 1a and 1b we restrict our analysis to the 311 articles that focused on norm and/or policy diffusion. Table 1a provides a breakdown of these articles by topic. Studies of policy diffusion are more common and appear with similar frequency across all three subfields. Studies of the diffusion of norms occur less frequently and appear in only the IR and CP subfields, with roughly equal frequency in these two subfields.

Table 1a Type of Diffusion Articles by Subfield

Table 1b shows the number of citations by each article type (e.g., AP policy) to each of the other article types (e.g., CP policy), again restricted to the set of 311 articles on policy or norm diffusion. The first row illustrates the strikingly insular nature of the AP policy work. Together, AP studies of policy diffusion cited 295 other articles about policy or norm diffusion, and fully 96% of these citations—283 out of 295—were to other AP policy articles, while only 2% of citations were to IR articles and another 2% were to CP articles. The second row, which focuses on citations by CP policy diffusion articles, shows a greater openness by CP policy scholars who cite IR policy and (especially) AP policy diffusion studies in combination nearly as frequently as their own works. The fourth row shows that IR policy diffusion scholars are equally open; in fact, there are more citations to AP and CP policy diffusion articles combined than to other IR policy diffusion articles. In addition, IR policy articles cite IR norms diffusion articles nearly as frequently as they cite other IR policy diffusion articles.^{Footnote 4}

Table 1b Citations by Type of Article and Subfield

Note: Cells indicate the number of citations to this group of articles (columns) by this group (rows). Values in parentheses indicate the number expected if each type of article were equally likely to cite each type of article (parenthetical row and column totals may not add up due to rounding). The null hypothesis of such equal citations is easily rejected (χ2=1914.7, p < 0.001).

Table 1b also demonstrates that AP policy diffusion articles are the most cited of all policy diffusion articles (as shown in the first column). Despite comprising only 27% of all policy or norm diffusion articles (i.e., 83 of the 311 articles included in table 1a), they receive the majority of citations by all policy and norms diffusion articles across subfields. Specifically, table 1b shows that 66% (349 of 526) of the citations found in these articles are to AP studies of policy diffusion.^{Footnote 5} CP and IR policy diffusion articles, in contrast, receive only 19% (101 of 526) and 8% (40 of 526) of the citations, respectively. The overall bias toward citations within subfield and within topic is evident down the main diagonal in figure 1b, accounting for 73% of all citations.^{Footnote 6}

These general tendencies conform to previous characterizations of the subfields. For example, CP has been described as lacking well-defined research programs (Mahoney Reference Mahoney2007, 116) and as a “fragmented discipline, one that continually struggles to reconcile broad, transportable general theory with context-specific, contextually relevant knowledge” (Caporaso Reference Caporaso2000, 699–700). But one benefit to the subfield is its diversity and ecumenical approach, often drawing on literature in other subfields to build and test arguments. In contrast, AP diffusion scholarship is built on a common base yet is open to Pierson’s (2007, 146) charge that “Americanists have much to sell, but it would be far healthier if they were also very active buyers in the intellectual market place.”

COMMON CULPRITS: TERMINOLOGY, METHODOLOGY, SUBFIELD JOURNALS?

The analysis thus far has demonstrated that substantial connections exist between the IR and CP diffusion literatures dealing with norms and policies, with IR articles especially likely to cite CP articles. In addition, CP articles frequently cite AP policy diffusion articles. Connections between AP and IR policy diffusion studies, however, are almost nonexistent. What might explain these patterns? Is the methodological reliance on quantitative research in AP a barrier to exchange with subfields that feature more qualitative research? Do differences in terminology across these literatures keep them isolated from one another? Does the degree of overlap across journals play a role? In this section, we explore each of these potential culprits that might be responsible for the lack of cross-subfield work, opening a window into divisions in political science that may extend well beyond the diffusion literature.

Regarding terminology, as noted previously, we identified articles by various search terms, and here we focus on the five that are most prevalent in the study of the spread of norms and policies: diffusion, convergence, race to the bottom, policy transfer, and harmonization. Table 2a shows the number of each type of article that predominantly uses each of these five terms. As this table shows, terminology certainly has the potential to be divisive across subfields, as some terms are used nearly exclusively in a single subfield (“race to the bottom” in AP, “policy transfer” in CP, and “harmonization” in IR). Moreover, “convergence” is the most commonly used term in CP and IR, but is dramatically outpaced by “diffusion” in AP.

Table 2a Articles by Terminology Used

Note: Cell entries indicate the number of articles of this type (rows) that mainly used this terminology (columns). The totals add to 310, rather than 311 from Table 1a because one CP: Norms article primarily used the term “contagion.”

Table 2b reveals the extent to which chosen terminology matters in connecting scholars and their articles. The results are dramatic. Relative to the values in parentheses, reflecting expectations of equal citations across terms, we find a strong bias toward the citation of articles that use the same terminology. For example, as shown in the first row of table 2b, 92%—that is, 355 of 387—of citations by articles using “diffusion” terminology are to other articles that do the same. This is remarkably high, especially because diffusion-terminology articles only comprise 42% of all articles (129 of 310) shown in table 2a. Therefore, diffusion-terminology articles draw on insights from articles using other terminology (i.e., convergence, race to the bottom, policy transfer, and harmonization) only 8% of the time. Articles using other terminologies exhibited similar (although less severe) biases.^{Footnote 7} In sum, 78% of citations across all of these categories were within-terminology, compared to an expected 31% that would have occurred if there were there no within-terminology bias.^{Footnote 8} Such a bias, combined with the different terminologies used across subfields, matches our finding of a closer alignment between CP and IR than with AP. It also explains further anomaly—for instance, the AP articles on the lower right of figure 2, outside of the main cluster, are almost entirely studies of a potential “race to the bottom,” mainly in US state welfare policy, separated from the others seemingly only by a terminological choice to not describe their phenomena in terms of “diffusion.”

Table 2b Citations by Terminology Used Being Cited

Note: Cells indicate the number of citations to this group of articles (columns) by this group (rows). Values in parentheses indicate the number expected if each type of article were equally likely to cite each type of article (parenthetical row and column totals may not add up due to rounding). The null hypothesis of such equal citations is easily rejected (χ2=1200.2, p < 0.001).

With regard to methodology, we code articles as using mainly a quantitative or a qualitative approach.^{Footnote 9} As shown in Table 3a, 72% of AP policy diffusion articles are quantitative in nature, whereas 65% of CP and 80% of IR diffusion articles are qualitative. Perhaps these differences explain the scholarly divide, most pronounced between AP and IR. To explore the extent of any within-method bias, Table 3b shows the number of citations across these articles both by and to each type of method. As shown in the first row, quantitative articles exhibit an in-method bias, citing other quantitative articles 91% of the time, despite comprising only 40% of the articles being examined. Interestingly, however, qualitative work, too, exhibits a quantitative bias, citing within-method only 46% of the time. When we look at these results in conjunction with subfield affiliation, we find that 60% of all cites are “within category” such as quantitative AP articles citing one another or qualitative IR norms articles citing one another. This is four times the rate expected if articles exhibited no biases in citing other articles, and instead cited other articles randomly, without regard to subfield or method.^{Footnote 10}

Table 3a Articles by Methodology Used

Note: Cell entries indicate the number of articles of this type (rows) that used this methodology (columns). The totals add to 267, rather than 311 from Table 1a because some articles use both methodologies (mixed methods) or neither (such as formal modeling work).

Table 3b Citations by Methodology Used

Note: Cells indicate the number of citations to this group of articles (columns) by this group (rows). Values in parentheses indicate the number expected if each type of article were equally likely to cite each type of article (parenthetical row and column totals may not add up due to rounding). The null hypothesis of such equal citations is easily rejected (χ2=628.5, p < 0.001).

Finally, we considered the effect of journal outlet in producing or entrenching subfield divides. To do so we coded each of the 53 journals in two different ways. First, we coded each journal strictly as AP, CP, or IR based solely on which subfield published the most norm or policy diffusion articles in that journal. Second, we recoded many of these journals as “multi-subfield” if they are widely regarded as “general journals” or if they are known for publishing mainly in one subfield but instead publish diffusion articles mainly from a different subfield.^{Footnote 11} Tables 4a and 4b show our analysis based on the first coding scheme. As seen in Table 4a, and as expected, most articles are published in journals dominated by their own subfield, such as 96% of AP policy diffusion articles published in predominantly AP journals. That said, 18% of CP articles are published in AP journals and 15% in IR journals, while 18% of IR policy diffusion articles are published in predominantly CP journals.

Table 4a Articles by Main Subfield of Journal

Note: Cell entries indicate the number of articles of this type (rows) published in journals most frequently publishing this subfield’s diffusion work (columns).

As shown in Table 4b, citations largely stayed within subfields, based on journal placement, with other AP journal articles accounting for 81% of citations in AP journal articles, and with within-subfield journal citations accounting for 76% of all citations overall.^{Footnote 12} The largest crossovers in the first row are AP journal articles citing those in CP. The middle row shows about equal crossover from CP to AP as to IR, whereas the bottom row shows robust crossover to CP from IR, but not to AP.

Table 4b Citations by Journal’s Subfield

Note: Cells indicate the number of citations to this group of articles (columns) by this group (rows). Values in parentheses indicate the number expected if each type of article were equally likely to cite each type of article. The null hypothesis of such equal citations is easily rejected (χ2=1243.7, p < 0.001).

The second analysis, using “multi-subfield” journals is somewhat less informative along these lines, because a majority of AP and CP policy diffusion articles are published in these journals, and because “multi-subfield” journals receive 81% of the citations.^{Footnote 13} This indicates, unsurprisingly, that publications in top general journals are more likely to be noticed and attract citations. More relevant for our present purposes, and consistent with the first coding scheme, we find that of those articles published in subfield (rather than “multi-subfield”) journals, 85% of citations are within-subfield journal, with zero crossovers from other subfields to AP journal articles or vice versa. IR and CP scholars pay little attention to articles that, although they might be relevant, are published in AP journals; and AP scholars unfortunately return the favor.

IMPLICATIONS AND CONCLUSIONS

We set out to explore whether and why subfield boundaries separated scholars who ostensibly should be drawing on one another’s ideas, approaches, and findings. To do so, we chose an area of inquiry—diffusion studies—in which similar topics stretch across subfields. We found, as should be expected, that when the topics of interest varied from one another, such as conflict versus democratization, there was little overlap in citation networks. Put simply, scholars working on these different topics were not highly connected to one another. However, even when scholars were studying similar phenomena, such as the spread of policies or of norms across governments, the network ties were inconsistent.

A recent turn in policy diffusion scholarship emphasizes the importance of specifying the mechanisms of diffusion—that is, not simply when, but how diffusion occurs (e.g., Berry and Baybeck Reference Berry and Baybeck2005, Braun and Gilardi Reference Braun and Gilardi2006, Shipan and Volden Reference Shipan and Volden2008). Learning is one prominent mechanism of diffusion whereby policy makers take note of effective policies tried elsewhere and subsequently choose to adopt those policies, thus facilitating diffusion. Our analysis suggests that subfield boundaries, reinforced by differences in terminology, methodology, and journal placement, prevent scholars from learning from one another.

Scholars couching their research in terms of “policy transfer,” a “race to the bottom,” or “harmonization” may be speaking with one another; but they are neither engaging nor subsequently being engaged by the larger group of scholars who describe their work mainly with the term “diffusion.” Elsewhere we offer a humorous list of 104 separate terms that scholars have developed to describe different diffusion processes (Graham, Shipan, and Volden Reference Graham, Shipan and Volden2013). Here our research suggests that scholars may instead benefit from seeking to come to a common understanding of a much smaller number of groupings, and to discuss them with a common terminology.

On methodological grounds, we found a large divide between the quantitative studies in AP and the qualitative studies in CP and IR. Coupled with a bias toward citing studies that use similar methodology (especially among quantitatively-oriented scholars), it seems clear that methodological divides are keeping scholars from building on one another’s findings. In our view, more could be done to bridge these methodological divides. For instance, the Empirical Implications of Theoretical Models initiative has sought to break down methodological divides between empirical and theoretical approaches. In so doing, more articles presenting formal theory results tend to discuss their testable implications. In a similar way, qualitative scholars (and referees of their works) could take further steps to concretely lay out whether general, systematic, and testable expectations follow from their qualitative findings. And quantitative scholars could highlight areas for future qualitative work, or could offer more accessible and thorough qualitative analyses of their own to motivate their hypotheses or to fill in where available data may fall short.

While subfield-specific journals may serve a variety of useful purposes, they also clearly reinforce disciplinary boundaries. This suggests a useful role to be played by referees and editors from other subfields, even in subfield-specific journals. The practice may be particularly appropriate for topics like diffusion studies, which lend themselves to cross-subfield scholarship. The selection of outside referees has the potential to enlighten both author and reviewer by broadening the scope of feedback for the author and providing referees with insights into alternative ideas and perspectives on familiar political processes.

In sum, we find diffusion scholarship to be prone to these barriers despite its status as a most likely case for integration. Unlike some other topics, this is an area where all three subfields are focused on similar questions related to the political conditions that facilitate the spread of policies and norms across governments. If integration is far from complete in this case, we would not expect high levels elsewhere, where various connections could prove useful even if they are less obvious.

There are clear costs associated with these divisions. Without an awareness of developments in other subfields, scholars miss opportunities to spur new ideas, counter old arguments, and build on one another’s accomplishments. And as a discipline, we may be more subject to reinventing the wheel, with what appear to be new and novel contributions containing ideas found elsewhere—only using different terminology or appearing in journals in a different subfield.

We conclude with one glimmer of hope, however, based on the increasing use of coauthorship in political science today. Scholars who “sit at multiple tables” or those who work with coauthors who straddle these divisions can overcome some of the problems described previously. For instance, Harvey Starr’s (1991; along with Siverson and Starr Reference Siverson and Starr1990) work on contagion effects associated with conflict and with democracy provides an example of the former, as the arguments draw on and tie into literatures prevalent in both IR and CP. And Michael Mintrom’s studies with coauthors in AP (e.g., Mintrom and Vergari Reference Mintrom and Vergari1998) and IR (e.g., True and Mintrom Reference True and Mintrom2001) provide an example of the latter. Indeed, greater integration between IR and CP in diffusion studies and between international and comparative political economy more generally has been facilitated by such scholars (e.g., Milner Reference Milner1998; Simmons, Dobbin, and Garrett Reference Simmons, Dobbin and Garrett2006), in an encouraging trend that we hope to see continue.