A robust scholarship argues that policy entrepreneurs, change agents who work individually and in groups to influence the policy process, can be key in introducing policy innovation and spurring policy change (Arnold, 2020, 2022; Kingdon, 1984; Mintrom & Norman, 2009; Petridou & Mintrom, 2020). How to identify policy entrepreneurs empirically has received less attention. Yet being able to accurately and reliably identify these change agents is crucial for scholars seeking to understand when they emerge and why, and what factors make them more or less successful. This paper summarizes the data sources scholars use to learn about policy entrepreneurs, the process-based and event-based approaches scholars typically use to identify entrepreneurs, and ways scholars distinguish policy entrepreneurs from other political actors. It then highlights issues with these practices, including likely bias toward studying successful, elite actors and away from less successful, lower-profile advocates, lack of large-n comparative scholarship, and ambiguity around the markers that distinguish policy entrepreneurs. We then introduce a new technique for entrepreneur identification, applying it to the empirical case of unconventional oil and gas drilling in Pennsylvania.

How do scholars identify policy entrepreneurs for empirical analysis?

Scholars empirically identify policy entrepreneurs by finding evidence of them in source materials and distinguishing them from other policy actors. We focus on the former and then turn to the latter. In this effort, we review academic works on policy entrepreneurship which clearly document methods of data collection and analysis, while noting that there are many other works which develop theories of policy entrepreneurship but do not empirically test them (e.g., (Mintrom, 2019; Mintrom & Norman, 2009). We also emphasize that, although we refer to “policy entrepreneurs” for convenience, the status of an individual or organization as a policy entrepreneur is not immutable: An actor who entrepreneurially pursues X policy goal at time Y may take zero action concerning a policy in domain A at time B.

Policy entrepreneur data sources

A common approach to identify policy entrepreneurs is surveying elites or experts about influential actors in policy processes familiar to elites. Scholars have surveyed municipal clerks, asking them to point out and describe local policy advocates (Arnold, 2020, 2022; Arnold et al., 2017; Kalafatis, 2018; Kalafatis & Lemos, 2017; Kim, 1996; Schneider & Teske, 1992, 1993a, b; Teske & Schneider, 1994). Mintrom (1997, 2000) identified education policy entrepreneurs by polling education policy experts, while Anderson et al. (2020) surveyed state legislators about their contact with disaster preparedness policy advocates.

Researchers also interview elites and experts about policy entrepreneurs, often as part of case study efforts wherein researchers use news media content, advocacy materials, and/or policy documents to identify and characterize policy entrepreneurs (e.g., Crow, 2010a, b; Brouwer & Biermann, 2011; Petridou et al., 2021; Maurya & Mintrom, 2020; Guldbrandsson & Fossum, 2009; Font & Subirats, 2010; King & Roberts, 1992; Roberts & King, 1991; Cohen, 2012; Braun, 2009; Aviv et al., 2021; Botterill, 2013; Reimer & Saerbeck, 2017; Timmermans et al., 2014; Ruvalcaba-Gomez et al., 2020; Capano & Galanti, 2021; Arnold, 2015; Hammond, 2013; Oborn et al., 2011; Frisch-Aviram et al., 2018; 0- 2009; Petridou & Olausson, 2017; Lu et al., 2020; Petridou, 2018). Researchers in this vein often rely on in-depth knowledge of a case to help identify policy entrepreneurs, rather than specifying a precise rubric for distinguishing entrepreneurs from other advocates, experts, and political figures.

Some scholars identify policy entrepreneurs solely from secondary sources. For example, Weissert (1991) analyzes legislative records, identifying lawmakers who repeatedly introduced bills on a salient topic as policy entrepreneurs. Others use legislative records to define as entrepreneurs US congressional representatives who introduce and initiate action on foreign policy issues prior to any administrative action on the same issues (Carter & Scott, 2004). Analyzing policy documents and web materials, Chatfield and Reddick (2018) consider any Australian government unit who appeared to drive efforts to adopt open government data policies as policy entrepreneurs.

A small but growing area of research involves identifying and contacting potential policy entrepreneurs and asking questions to parse out who among them qualify for the classification. For example, Shearer (2015) analyzes policy documents and media content to identify members of the policy community concerned with integrated community case management in Burkina Faso, then surveys, and interviews these individuals to isolate the key entrepreneurial actor. Faling et al. (2018) survey members of global climate initiatives asking who among them were the entrepreneurs who drove the initiative's creation, following up with interviews with those actors.

Finally, some studies focus on well-recognized individuals whose classification as policy entrepreneurs appears a topic of general agreement, such as Erin Merryn, the key agent behind U.S. state adoption of child sex abuse prevention laws (“Erin’s Law”) (Fowler & Vallett, 2021; Vallett, 2021); former Arkansas Governor Mike Huckabee, who advocated for school-based policies to tackle childhood obesity (Craig et al., 2010); and the premiers of the Australian states of Queensland and Victoria, sites of major policy efforts to develop new knowledge economies (Mintrom et al., 2014). See as additional examples Oborn et al. (2011), Navot and Cohen (2015), Hammond (2013), Mackenzie (2004), and Palmer (2015). In sum, scholars find policy entrepreneurs by asking elites and experts to point them out, querying secondary sources, surveying possible entrepreneurs, and focusing on high-profile advocates.

Identifying policy entrepreneurs

How do researchers know a policy entrepreneur when they see one? A common approach is event-based, wherein researchers identify cases of policy innovation or success and then nominate as policy entrepreneurs one or a small number of actors or organizations that appear key in creating these outcomes. Scholars look for actors behaving in ways expected from policy entrepreneurs, like networking and strategically using expertise (see Aviram et al., 2020). This widely used approach has documented policy entrepreneurs involved in community cancer care (Petchey et al., 2007); state wetland management (Arnold, 2015); reform of the London healthcare system (Oborn et al., 2011); adoption of vertical greening policies in Shanghai (Lu et al., 2020); efforts to reduce government corruption in Israel (Navot & Cohen, 2015); development of China’s minimum livelihood guarantee system (Hammond, 2013); disability policymaking in Indonesia (Setijaningrum & Rahardian, 2022); innovating in local delivery of social services in Israel (Aviv et al., 2021); adoption of social health insurance policies in India (Maurya & Mintrom, 2020); and the development of a European Union biofuels policy (Palmer, 2015). Sometimes, this approach identifies entire units of government which achieved a policy innovation as policy entrepreneurs, including US states that adopt climate policies (Drummond, 2010; Rabe, 2004), countries that adopt a participatory budgeting model (Wampler, 2009), and city governments that implement a new approach to managing refugees (Garcés-Mascareñas & Gebhardt, 2020) and an open government policy (Ruvalcaba-Gomez et al., 2020).

A related, process-based approach considers as policy entrepreneurs individuals or entities involved in efforts to secure or implement innovative policies, even if they are not successful. Research may consider actors as policy entrepreneurs if they meet some minimum benchmark ranging from actively participating in the policy process at least once, to persistently putting forth policy suggestions. This includes actors active in climate policymaking in Norway (Reimer & Saerbeck, 2017), congressional representatives introducing foreign policy legislation before administrative action on the same topic (Carter & Scott, 2004); actors “personally and actively involved in developing” sustainability policy innovations in the Netherlands and Belgium (Timmermans et al., 2014); legislators sponsoring policy on a salient issue over multiple legislative sessions (Weissert, 1991); and individuals and organizations attempting to secure national-level drought policy (Botterill, 2013).

Distinguishing policy entrepreneurs

Scholars vary in the extent to which they use specific criteria to distinguish a policy entrepreneur from other actors in the policy process; some event- and some process-based studies, for example, appear to use a “I know it when I see it” standard. Researchers using policy entrepreneurs as their dependent variable, seeking to explain their emergence and/or functioning, tend to provide greater specificity.

Kingdon’s classic policy entrepreneur definition—“advocates who are willing to invest their resources—time, energy, reputation, money—to promote a position in return for anticipated future gain in the form of material, purposive or solidary benefits” (Kingdon, 1984, 179)—is a common starting point for empirical researchers. Lu and coauthors (Lu et al., 2020, 118), for example, note that the policy entrepreneur they identified “devoted considerable time and energy to continuously influence policymaking towards its desirable outcome.” Lavee and Cohen (2019, 479)’s policy entrepreneurs were “energetic … actors who were most active in their pursuit of policy change…” while Crow’s (2010a, b) policy entrepreneurs “fought” for a policy change. Arnold (2020, 980) and Arnold et al., (2017, 423) conceptualized policy entrepreneurs as individuals “most active” in trying to pass local fracking policies and found that more than 50 percent of these entrepreneurs “spend a lot of time on advocacy.” However, these studies rarely explicate terms like “considerable” and “most active” and “a lot.”

Policy entrepreneurs are sometimes identified by their influence. For instance, Lavee and Cohen (2019, 479) say that a policy entrepreneur is someone “whose influence on policy design was unquestionable.” Mintrom (1997, 746; 2000) asked education policy experts to indicate the “most important” person advocating for charter school policies. Similarly, Faling et al. (2018, 411) asked members of a climate smart agriculture alliance who they “considered to be most important in the establishment of the alliance.” These studies rely on qualitative assessments of those closest to the policy processes.

Scholars may also recognize policy entrepreneurs based on their motivation. On one hand, policy entrepreneurs may be identified as actors aiming to change existing policies or introduce new policies (Hammond, 2013; Brouwer & Biermann, 2011; Schneider & Teske, 1992; Teske and Schneider 2016; Schneider & Teske, 1993a, b; Kim, 1996; Petridou et al., 2021). Scholars following this conceptualization infer that entrepreneurs pursuing new or change-making policies intend to create innovation or change. In some cases, researchers may ask entrepreneurs to describe their aims (Petridou, 2018; Petridou & Olausson, 2017; Petridou et al., 2021). Alternatively, researchers may themselves select policies they view as innovative, new, or important policies, and identify any actors pursuing these policies as entrepreneurs. Studies using this approach have examined adoption of policies for managing recreational diversion of river water (Crow, 2010a, b), requiring or encouraging wetland managers to assess wetland functions and values using a rubric (Arnold, 2015, 2021), regulating high-volume hydraulic fracturing (Arnold, 2020, 2022; Arnold et al., 2017), addressing climate change or global warming (Kalafatis, 2018; Kalafatis & Lemos, 2017), preparing for natural disasters (Anderson et al., 2020), managing flooding and flood risk (Petridou & Olausson, 2017; Petridou et al., 2021), or facilitating charter school education (King & Roberts, 1992; Mintrom, 1997, 2000; Roberts & King, 1991). An unresolved issue that arises when using motivation as a criteria for entrepreneurship is whether someone whose job focuses on influencing policy (e.g., a lobbyist, a legislator) should be consider a policy entrepreneur, or whether the classification reserved by those motivated to act outside or beyond their professional scope.

Scholars often attribute a suite of characteristics (e.g., being knowledgeable, well-connected) and strategies (e.g., networking, framing problems, and solutions) to policy entrepreneurs, as analyzed in a comprehensive review by Aviram et al. (2020). There is little point in replicating their excellent work. Importantly, though, the review did not address which characteristics and strategies are necessary to qualify someone as a policy entrepreneur, probably because few empirical works address this dilemma head-on. Arnold (2020) makes progress by applying data classification techniques to survey data describing individuals involved in policy advocacy. That analysis derives policy entrepreneur “archetypes” as frequently associated suites of strategies, characteristics, and goals: Highly engaged activists, who display many entrepreneurial characteristics, deploy many strategies to influence policy, and pursue multiple policy goals; advocates, similar to activists except they deploy fewer strategies and pursue fewer goals; and concerned citizens, who display only a few entrepreneurial dimensions. Arnold posited that concerned citizens might not qualify as policy entrepreneurs at all, but noted that the literature does not lend itself to judgment.

An emerging strand of scholarship uses social network analysis (SNA) to identify policy entrepreneurs by their relational attributes. Scholars elicit the information sharing, resource sharing, or other types of networks among policy participants, often contextualizing SNA results with qualitative insights. Analysis suggests that policy entrepreneurs have high betweenness centrality, meaning they occupy the shortest path between actors in a network and thus can influence flows among them (Christopoulos & Ingold, 2015); see also Shearer (2015) for similar results in health policy in Burkina Faso. Research also shows that policy entrepreneurs possess ties to both central and peripheral actors and bridge structural holes in networks, connecting actors with otherwise limited interplay (Christopoulos & Ingold, 2015); see also Christopoulos (2006) and Christopoulos and Ingold (2011). McIntyre et al. (2018) find that policy entrepreneurs in Canadian food insecurity policy connect regionally isolated subgroups. Analyzing local government flood risk mitigation efforts, Petridou et al. (2021) identify an actor with extensive connections to central (influential) actors and a high score on a composite index of centrality as a policy entrepreneur. Petridou (2018) emphasizes that while network measures point to actors well positioned to act as policy entrepreneurs, actors must still deploy entrepreneurial behaviors to be considered as such. To sum up, policy entrepreneurs are typically identified by their involvement in policy events or processes, active engagement in advocacy and pursuit of innovative policy goals, and individual or relational characteristics and strategies.

Problems with existing approaches for eliciting and characterizing policy entrepreneurs

Problems with data sources

Using elite or expert input to identify policy entrepreneurs may overrepresent the prevalence of policy entrepreneurs (i.e., informants’ peers or contacts) while overlooking less connected or less expert individuals trying to influence policy. When researchers ask other people to make inferences and judgements about policy entrepreneurs, the informant’s own perceptions color their responses, which may be quite different than perceptions the subjects themselves would give. First-hand accounts from policy entrepreneurs can answer questions about their motivations in ways that second-hand approaches cannot.

While there are studies that ask policy entrepreneurs about themselves and their work (Arnold, 2015, 2020; King & Roberts, 1992; Petridou et al., 2021; Roberts & King, 1991), there is a need for larger-N and comparative research. Existing studies tend to interview a small number of policy entrepreneurs rather than survey a larger number, yet the latter is important for testing theories developed in small-n qualitative work and helping detect systematic and potentially generalizable trends. Research comparing how policy entrepreneurs characterize their activities and aims with how other policy participants perceive these same elements could offer insights to help policy entrepreneurs reshape and improve their advocacy success.

Problems with policy entrepreneur identification and related research design

It is a problem that case studies of successful policy entrepreneurship, typically involving an event-based entrepreneur identification approach, dominate the literature. Fewer scholars examine failure cases to assess whether policy entrepreneurship was absent, yet such analysis is crucial for establishing that policy entrepreneurship actually influences outcomes. There are important exceptions, like Crow’s (2010a, b) case studies of Colorado communities considering applying for a new water right, where policy entrepreneurs were active in communities that pursued the innovation and absent in those that did not, and Mintrom (1997, 2000) finding that states were more likely to fail to adopt an education policy innovation when no policy entrepreneur championed it. Kalafatis and Lemos (2017), Kalafatis (2018), and Kim (1996) observe similar results in municipal-level climate policymaking, and Hoeijmakers et al. (2007) explain the difficulties four Dutch municipalities experienced trying to adopt local health policies by, in part, the absence of a policy entrepreneur.

Some of these analyses suggest that policy entrepreneurship is not determinative: Thailand adopted a new vaccine program without a clear policy entrepreneur (Munira & Fritzen, 2007), efforts to change degree requirements for nurses failed in three US states where a policy entrepreneur pushed for the change and succeeded in a fourth state without an entrepreneur (Smith, 2010), and multiple entrepreneurial efforts failed to achieve a national-level drought management program in the USA, whereas Australia adopted such a program without a policy entrepreneur (Botterill, 2013). The literature needs more careful analyses of cases with varied outcomes to understand whether policy entrepreneur presence or absence correlates with these outcomes in theoretically anticipated ways.

Even fewer studies examine cases where policy entrepreneurs were present but unsuccessful and try to explain why they failed. If a jurisdiction fails to adopt a path-breaking policy, the failure may not be solely attributable to lack of policy entrepreneurship: Policy entrepreneurs might lack facilitative characteristics or be using outdated or bad strategies to try and advance policy goals. Again, there are exceptions which point to paths for future research: Arnold (2015) describes how state wetland officials fail in their attempts at policy change due to deficits of political, social, and intellectual capital and that wetland managers who innovate exhibit greater constancy, have access to more resources, use more politically sensitive strategies, and focus more on cultivating intragovernmental relationships (Arnold, 2021). Arnold (2020) correlates expert evaluations of policy entrepreneur success and failure with entrepreneur archetypes, and composites of entrepreneur characteristics, strategies, and goals. Understanding what doesn’t work for policy entrepreneurs, when, and why, is as important as understanding what does work.

Problems distinguishing policy entrepreneurs

The literature lacks clear criteria for distinguishing policy entrepreneurs from other types of policy participants (e.g., concerned citizen, grassroots activist). Indeed, at the extreme, either an event- and process-based approach could result in the broad designation as policy entrepreneur of anyone involved in advocacy at any point. It is more sensible to understand policy entrepreneurism is a role that actors can embrace to varying degrees at various times (Capano & Galanti, 2021). However, scholars still need markers to guide their search for individuals or organizations practicing policy entrepreneurism.

There are promising efforts on this front, including recent applications of social network analysis which assign something akin to “entrepreneurship potential” scores to policy participants and consider as entrepreneurs those who breach a theoretically motivated threshold (Petridou et al., 2021). See for example, Frank et al. (2012)’s index of policy advocacy, measuring the extent to which scientists engage in “policy-oriented behaviors” to influence climate policy; and Arnold (2020)’s minimum threshold for policy entrepreneurship, considering individuals as such if they exhibit one of ten policy-targeting activities. These approaches need not be tied to quantitative analysis: Rubrics comparing one policy actor’s level of advocacy to another’s can be completed by the researcher using insights from interviews and focus groups, for example, as easily as they can be informed by numeric survey or secondary source data.

In sum, the data sources and approaches scholars use to identify policy entrepreneurs often lead them to examine successful, elite actors rather than less successful, lower-profile advocates. Scholars often infer rather than assess policy entrepreneur motivations, fail to examine cases of failed entrepreneurship, are vague about the markers distinguishing policy entrepreneurs from other actors, and do not rigorously test whether policy entrepreneur activity influences outcomes.

Tackling problems of policy entrepreneurship research

The effort we describe below seeks to locate low-profile policy entrepreneurs in addition to elites and less successful entrepreneurs in addition to successful ones. It asks policy entrepreneurs themselves about their own motivations and proposes specific markers to distinguish policy entrepreneurs. We pilot an approach for identifying policy entrepreneurs that combines automated text-mining and manual content analysis of local news media. We survey the resulting policy entrepreneur candidates, allowing them to explain for themselves what they did and why. Using survey data, we develop a replicable approach for distinguishing policy entrepreneurs from other policy actors. We then describe attributes of the policy entrepreneurs we identify and what appears to help them succeed, though we do not comprehensively evaluate whether these are the only or most important factors predicting success. Because our primary focus is explaining our methods, the data analysis is mainly intended to suggest promising avenues for future research.

Substantive focus: Unconventional oil and gas development

The empirical case is policy advocacy around unconventional oil and gas drilling (UOGD, colloquially known as “fracking”) and associated industries in Pennsylvania, overtop the highly productive Marcellus and Utica shales. The US UOGD boom that began in the mid-2000s transformed the energy industry, making the US the world’s leading producer of natural gas and dramatically increasing its oil production (Rapier, 2017). UOGD is controversial, with opponents alleging that it pollutes the environment, harms public health, and leads to crime and social ills in surrounding communities (Klasic et al. 2022). Policy advocates opposing UOGD seek local, state, and national bans and restrictions on the industry, while advocates seek to loosen and limit regulation (Arnold, 2020, 2022; Arnold & Long, 2019; Arnold et al., 2017). This is thus a fruitful case for identifying policy entrepreneurs.

Identifying policy entrepreneurs: data sources

Our aim in using local news media as a data source is to find policy entrepreneurs who might lack the political connections or expertise of entrepreneurs nominated by elite informants, yet nonetheless try to influence policy. Policy scholarship analyzing news media often queries national or state-level papers (e.g., Olofsson et al., 2018; Weible & Heikkila, 2017), whose large scope means that their writers may only have space for interview quotations from a small number of actors involved in or affected by an issue, and so target the most high-profile actors and/or easiest to track down. By contrast, the delimited scope of local media gives more opportunity for detail. Local reporters often know their communities well and can be attuned to community undercurrents, subtle local political dynamics, grassroots engagement, and the participants therein. We examine local news media to try to find those participants.

Of course, newspapers do not offer a comprehensive account of all actors involved in policy processes. Individuals who primarily act behind the scenes may escape journalistic scrutiny, as may individuals whose lack of skill or expertise causes their policy entrepreneurship efforts to fail. Reporters are likely to interview people more active in policy affairs. Journalists trying to demonstrate impartiality may seek quotes or information from actors on multiple sides of an issue, even if some of these perspectives are not widely represented in the population (Foreign Press USA, 2022). Thus, our newspaper-based identification strategy may identify as potential policy entrepreneurs individuals who would not be considered credible, and likely would not be nominated, by elite informants.

After extracting the names of individuals and organizations from local news media with the potential to be policy entrepreneurs, as more fully described below, we searched for email addresses for these actors. We emailed them a survey, soliciting descriptions of their own policy advocacy efforts to determine whether they should count as policy entrepreneurs. The survey asked about their motivations for pursuing advocacy, tackling an issue arising when informants or secondary sources characterize policy entrepreneurs: These sources cannot directly speak to the policy entrepreneur’s goals, but the policy entrepreneur themselves can.

Identifying policy entrepreneurs: case selection

Rather than beginning with policy successes and trying to identify individuals involved, we began with a substantive domain (UOGD) in which policy was discussed, debated, and made and re-made over a number of years by local and state governments. We consider policy processes in this domain over more than a decade, hoping this expansive scope will encompass multiple policy advocates experiencing successes, partial successes, and failures at different times in different places. We posit that any individual involved or interested in or affected by UOGD in the study region has the potential to become a policy entrepreneur. This inclusive approach may help us capture individuals less successful at influencing policy, or who have a less decisive role or exert influence yet have a lower profile. It also may help avoid some bias associated with asking elite or expert informants to identify policy entrepreneurs.

Distinguishing policy entrepreneurs

We use data from multiple survey measures to establish theoretically grounded empirical thresholds for a survey respondent to qualify as a policy entrepreneur. First, drawing on Kingdon (1984), we operationalize the resources a policy entrepreneur invests in advocacy as their frequency or duration of time or financial investment in advocacy, or their investment of reputation. Specifically, we consider as policy entrepreneurs, individuals who report engaging in advocacy a few times a month or once a week or more,Footnote 1 or engaging in advocacy for five or more years, or spending more than 5% of their annual budget on advocacy,Footnote 2 or whose name or organization is identified by at least one other survey respondent as an entity with whom they work the most to oppose (support) UOGD. The latter condition assumes that popular advocacy partners are such because of their strong reputation as advocates. The research team selected the investment thresholds based on our own experiences with advocacy and related judgments about what constitutes significant commitment.

Second, an individual’s motivation can help determine their policy entrepreneur status. An individual may qualify as a policy entrepreneur if they aim to get government officials to adopt policies concerning UOGD (an area in which policies can be generally considered novel or innovative because they involve a drilling practice not previously deployed at scale), get government officials to pay attention to UOGD, or get government officials to share the respondent’s views on UOGD. This is a simplifiedFootnote 3 version of a set of goals previously used to identify policy entrepreneurs in local UOGD policymaking (Arnold, 2021, 2022; Arnold et al., 2017).

Next, an individual’s entrepreneur status could be decided by the strategies, or the suites of actions, pursued to achieve a goal. We consider four strategies outlined in an oft-cited paper by Mintrom and Norman (2009): defining problems in ways favorable to a desired policy goal (in the survey we termed this “raising public awareness” about UOGD, on the logic that entrepreneurs channel awareness along paths favoring their goals but that the words “framing” or “defining” might cause respondent aversion); “building teams or networks”; leading by example, often using “pilot or demonstration projects” to build a base of evidence supporting the entrepreneur’s preferred policy; and deploying social acuity, which we termed “people skills” in the survey. We add a fifth strategy, “drawing ideas and resources from diverse sources,” frequently described in the literature (Aviram et al. 2020; Kingdon, 1984; Giambartolomei et al., 2021; Frisch-Aviram et al., 2020; Brouwer & Biermann, 2011; Meijerink & Huitema, 2010; Mintrom, 1997, 2000).Footnote 4 Respondents also had the option of describing another strategy they use in the policy process.

To connect with scholarship arguing that a policy entrepreneur can be defined by their relational attributes (that is, where they operate in a policy network, how, and with whom), we consider the extent to which an individual pursues advocacy with partners. This approach diverges from other social network analyses (SNA) of policy entrepreneurship because it considers only the alters (connections) of a focal individual (the ego), rather than the whole network (see, e.g., Arnold et al., 2017; Petridou et al., 2021; Christopoulos & Ingold, 2015). However, a key SNA measure of policy entrepreneurism, network centrality, can be measured in ego networks as degree centrality, the number of connections an ego has to alter. We posit that a minimum threshold for policy entrepreneurship is an individual working on UOGD advocacy with at least one alter.

Myriad characteristics are linked to policy entrepreneurs (Aviram et al., 2020). We consider four: possessing relevant expertise or knowledge and being well-connected to government officials, other policy advocates, and members of the media. This is a simplified version of characteristics previously associated with policy entrepreneurs in local UOGD policymaking (Arnold, 2021).Footnote 5 Survey respondents could choose to report that these factors increased their advocacy efficacy. We posit that a minimum threshold for policy entrepreneurship involves an individual selecting at least one. Future research should consider a wider range of entrepreneurial characteristics.

Importantly, we do not equate acting as a policy entrepreneur with being a successful one. We do ask respondents to rate their success in achieving their policy goals, and the resulting data do show that individuals experienced varied levels of advocacy success.

The empirical thresholds we specify for qualifying as a policy entrepreneur could be set or configured in a variety of ways, and we welcome theoretically grounded suggestions for alternate approaches. We consider as a policy entrepreneur an individual who meets one of the investment thresholds; reports at least one entrepreneurial goal; reports at least one entrepreneurial strategy or one characteristic; and has at least one network connection. We allow a trade-off between strategies and characteristics because in some cases they have conceptual and practical overlap (e.g., a strategy of networking is likely linked to the characteristic of being well-connected; a strategy of drawing ideas and resources from diverse sources is likely linked to a characteristic of having expertise or knowledge).

Methods

Selecting and querying newspapers

We queried three local newspapers in areas of Pennsylvania affected by UOGD, satisfying three criteria: (1) They are published daily; (2) their archives are online and searchable; and (3) they charge nothing or only a small fee for article searches and downloads. Although we had hoped to collect articles published from the mid-2000s, coinciding with the onset of the UOGD boom, we were not able to identify any newspapers that satisfied these criteria prior to 2010. For two of the three newspapers, we collected articles from 2010 through 2021, a period in which the UOGD industry experienced cyclic booms and busts. For one paper, we were only able to collect articles from a four-year interval within this period.

The Herald Standard, based in Uniontown, Pennsylvania, provides the majority (84%) of articles; see Table 1. Uniontown, located in southwestern Pennsylvania’s Fayette County, ranks among Pennsylvania’s ten counties with the most UOGD drilling (Busby and Mangano 2017). The other two papers, contributing nearly 7% and 9% of the corpus, respectively, are the Williamsport Sun Gazette and the Lewiston Sentinel. The Williamsport Sun Gazette is based in the city of the same name in northeastern Pennsylvania’s Lycoming County, also in the top ten counties for UOGD activity (Busby and Mangano 2017). This paper’s coverage only spans 2010–13 because the newspaper blocked our download attempts after 2013. The Lewiston Sentinel is published in the city of the same name in Mifflin County. Mifflin County, in Central Pennsylvania, is not a top site for UOGD but is proximate to Dauphin County, the seat of state government. Given the major impact of UOGD on Pennsylvania, we suspected that Lewiston Sentinel might contain useful UOGD coverage despite a relative lack of local UOGD operations. The bounds of our corpus likely make it best suited to identifying policy entrepreneurs in southwestern Pennsylvania, potentially skewed toward those active in the early 2010s.

Table 1 Newspapers
Full size table

We queried the newspapers for search terms associated with UOGD, casting a wide net. Specifically, we downloaded articles that (1) referenced Pennsylvania or surrounding states also heavily affected by UOGD (West Virginia and Ohio primarily) and (2) contained at least one of: oil, oil and gas, gas, natural gas, fracking, fracing, fracturing, horizontal, shale, Marcellus, Utica, directional, industry, hydraulic fracturing, hydraulic fracking, hydraulic fracturing, hydrofracking, or drilling.

Team members reviewed a random sample of 220 articles to ascertain their relevance, defined as describing (1) actual events, activities, or decision-making processes related to UOGD, as opposed to industry overviews or descriptions of trends; (2) occurring in Pennsylvania or surrounding states substantially affected by UOGD; and (3) mentioning specific people, companies, organizations, or groups directly involved in or affected by the events, activities, or decision-making processes. This check showed that 0% (0/22) articles with only one keyword were relevant (failed to meet all three criteria); 8% (9/107) of the articles with two keywords were relevant; 18% (7/39) of those with three; 17% (5/29) with four; 30% (3/10) with five; 75% (3/4) with six; 75% (3/4) with seven; and 100% (2/2) with nine keywords were relevant. (No article of the 220 had exactly 8 keywords.) We decided to analyze only articles including at least two keywords, leaving a corpus of 3746 articles.

We next explored whether natural language processing (NLP), implemented using the R package GoogleNLP (Weinstien 2018), could effectively identify the people and organizations with proper names in each article; additional notes on this process are in Appendix. Using a random sample of 31 articles, team members manually extracted these terms and compared them with the terms extracted using NLP automation. Because the manual coders did not identify any entities that the NLP overlooked, but the NLP identified some entities that manual coders missed, we decided to use the NLP.

Team members manually inspected each term in its textual context to ensure that it was used (in at least one instance) in a manner satisfying the relevancy criteria noted above. Examples of terms determined ineligible include the name of a company that produces gas stoves, the university affiliated with a geology professor studying UOGD (the professor himself was considered eligible), a women’s empowerment nonprofit which received a grant from Chevron’s foundation arm but otherwise had no connections to UOGD, a deceased person, and duplicate terms tagged as distinct because of variations in spelling, title (e.g., Mrs. vs. Ms.), or similar. In total, 2,068 terms were judged irrelevant.

Creating a database of potential policy entrepreneurs

The next step involved turning the 4,681 remaining terms into a usable database of individuals to be surveyed. When an organization was noted in an article without mention of a specific individual within or representing the organization, we searched for other mentions of the organization in the corpus. If those mentions included specific individuals, we eliminated the organization-only term and retained the entry with a specific contact. If the corpus did not link any individual to the organization, we used web searches to identify the individual in charge of the organization (e.g., an executive director) or responsible for a relevant division of the organization (e.g., Earthjustice’s Fossil Fuels Program), and included this person in the dataset as the organizational contact. We reviewed duplicativeness again, taking care to aggregate terms substantively but not meaningfully different (e.g., Ed Rendell, Edward Rendell, and Edward G. Rendell, all referencing the former Pennsylvania governor). We removed terms describing national or international political figures lacking strong ties to the Marcellus/Utica region or UOGD, like Kamala Harris, Lindsey Graham, the UN Secretary General, and Barack Obama.Footnote 6 After this filtering, the database contained 1166 potential survey respondents.

Next, team members searched the Internet for email addresses for these individuals. For some, like government officials, contact information was readily accessible. For many others, particularly individuals in private sector oil and gas operations or “average citizens” identified in an article only by their area of residence rather than occupation or organization, information was more difficult to find. We extracted non-paywalled content from specialized online databases which aggregate personal information (e.g., Spokeo, PeopleFinder), attempting to triangulate freely available information across databases to cobble together one or more email addresses potentially associated with an individual. We also used business-to-business “lead” websites, like RocketReach and Aeroleads, to identify typical email address formats for a company (e.g., jane.doe@company.com, jdoe@company.com, doej@company.com) and used these to make educated guesses about employee email addresses. When we could find no specific contact information for an individual or make reasonable guesses as to their email, we used general organizational email addresses (e.g., info@company.com) when available. There are 33 individuals who we could find no way to contact via email, leaving 1,133 individuals in the survey sample. For 353 of these individuals, we identified two or three possible email addresses.

Administering the survey

We emailed a link to a survey programmed in the online platform Qualtrics to 1955 email addresses for the 1133 individuals in the sample, plus an additional 12 individuals (one email each) to whom initial contacts referred us. A total of 1258 emails appeared to have reached a recipient, while 697 emails bounced due to non-existent addresses, spam filtering, or some other problem. This large number of invalid addresses is not surprising given our use of educated guesses and error-ridden public information databases to source some addresses. The 1258 valid email addresses were associated with 996 unique individuals.

Members of the sample received an initial survey invitation plus three follow-up email reminders over roughly a two-month period in 2022. A total of 121 individuals consented to the survey and answered the first question concerning engagement in UOGD policy processes, for a response rate of 121/996, or 12.15%. This rate, while low, is consistent with falling survey response rates across a range of social science disciplines (Stedman et al., 2019). More importantly, in this project the response rate is not particularly meaningful: We do not know the size of the larger population of interest, UOGD-focused policy entrepreneurs in Pennsylvania and environs overtop the Marcellus and Utica shales. Our strategy for selecting individuals into our sample very likely included many non-entrepreneurs who skipped the survey because it did not appear relevant to them; we also likely failed to detect other “true” policy entrepreneurs who for one reason or another were not identified in news media. Our results characterize 121 individuals who had the potential to act as UOGD policy entrepreneurs and demonstrate methods we hope can be used to understand more policy entrepreneurs in other domains.

Measuring variables

We measure a policy entrepreneur’s investment in advocacy four ways. First, we asked respondents when they began UOGD-related advocacy and when or if they stopped, producing a duration variable indicating time in years. We collected ordinal data on frequency of advocacy by asking respondents to consider the year when their advocacy was most active and report whether, in that year, they engaged in the policy process once a week or more, once a month, once every few months, or once or twice a year. To create an ordinal variable measuring personal financial investment, we asked respondents to think of their most active year of advocacy and report the proportion of their personal annual budget spent on advocacy that year: 0–5%, 6–10%, 11–15%, 16–20%, or more than 20%. To measure reputation, we assigned a 1 (on a 0/1 binary variable) to a respondent whose name or organization was identified by at least one other survey respondent as an entity with whom they work the most to oppose (support) UOGD.

We asked respondents to indicate the goals they pursued in their policy advocacy, choosing among seven options. In this paper, we focus on three goals specifically targeting government officials’ policy agendas, constructing each as a binary variable (an individual did or did not pursue the goal). Respondents rated their success at achieving each goal they pursued, on a 1–5 scale where 5 is extremely successful and 1 is not successful. Binary variables capture a respondent’s use of each entrepreneurial strategy, their reporting of entrepreneurial characteristics, and whether they had at least one network partner. We also consider continuous variables counting the number of goals, strategies, and characteristics reported by a respondent, and their number of network partners (degree centrality).

Distinguishing policy entrepreneurs

Following the protocol detailed above, we identified 37 of the 121 survey respondents as policy entrepreneurs (Table 2). This approach is meant to ensure that the policy entrepreneurs we identify have representation across a range of important dimensions identified by policy entrepreneur theorists.

Table 2 Stepwise process for distinguishing policy entrepreneurs
Full size table

Analysis

Having identified the policy entrepreneurs, we now seek to understand them better and to learn whether their success can be explained by particular levels or types of investment, goals, strategies, individual or relational (network) characteristics, or demographics. The analysis thus speaks to the literature’s interest in understanding whether and how policy entrepreneurs impact policy processes and outcomes (Arnold, 2020). Although the exploratory nature of this work means we do not test specific hypotheses, it seems reasonable to expect that policy entrepreneurs who do more—invest more in advocacy, have more goals, pursue more strategies, and have facilitative individual and relational characteristics—may be more successful in achieving their goals.

Table 3 characterizes policy entrepreneur demographics using variables constructed from survey data, and Table 4 describes salient entrepreneurial features (their investments, goals, and so on). Job sector was constructed from responses to an open-ended query. Inductive coding yielded eight sectors: academia; environmental advocacy; government; nonprofit/community (non-environmental); oil, gas, and energy; private sector (legal services and consulting); and other. The data suggest that the policy entrepreneurs are primarily White males in their early 1960s with postgraduate education. Most engaged the policy process on multiple issues, not just UOGD. They are largely affiliated with the Democratic party, oppose UOGD, and work in the environmental advocacy sector.

Table 3 Policy entrepreneur demographics
Full size table
Table 4 Policy entrepreneur features
Full size table

Because of the modest number of policy entrepreneurs we identify, and because our primary purpose is to review methods used in policy entrepreneurship research and describe a novel methodological contribution, the statistical analyses are limited. They do not examine the relative impact of contextual factors versus entrepreneurial features or demographics on policy outcomes. Rather, we examine associations between features or demographics and policy entrepreneurs’ self-perceived level of goal achievement using Fisher’s exact test. This approach is more appropriate than a chi2 test when more than 20 percent of cells have expected frequencies < 5 (Kim, 2017), as often occurs in our data. By comparing actual and expected frequencies, we evaluate the direction of the relationships between variables when Fisher’s exact test shows statistical significance.

Table 5 records how policy entrepreneurs rated their success in achieving the three government-focused advocacy goals (recall that pursuing at least one of these goals was required for someone to be considered a policy entrepreneur). Success was measured on a scale from 1 to 5, where 1 means “Not successful” and 5 means “Extremely successful.” Responses suggest our elicitation strategy achieved its aim of identifying entrepreneurs with varied levels of success. Mean success is fairly similar across the first two goals, getting government officials to adopt UOGD policies and getting government officials to pay attention to UOGD; for both, success falls between “Not too successful” (2) and “Somewhat successful” (3). Success is lower for the goal of getting government officials to share the policy entrepreneur’s views on UOGD, between “Not successful” (1) and “Not too successful” (2).

Table 5 Policy entrepreneurs’ perception of their level of success by goal
Full size table

Table 6 displays p-values from Fisher’s exact tests considering whether policy entrepreneurs’ investment in advocacy, goals, strategies, characteristics, and networking appear to vary in systematic ways with their success in achieving policy goals. Most tests show no statistically significant patterned relationships. We do observe a few differences concerning the goal of getting officials to pay attention to UOGD. Comparisons between observed and expected frequencies (see Appendix) indicate that policy entrepreneurs who build teams or networks, use people skills, and draw together ideas and resources from diverse sources report higher levels of success in directing government officials’ attention. There is also some evidence that policy entrepreneurs who collect ideas and resources perceive themselves as more successful at getting government officials to share their views on UOGD.

Table 6 Fisher’s exact tests evaluating relationship between goal success and entrepreneur features
Full size table

We use the same approach to explore whether policy entrepreneur demographics vary in patterned ways with an entrepreneur’s self-perceived success. Most variables do not exhibit systematic relationships, with two exceptions: policy entrepreneur age and job sector. Comparisons of actual versus expected frequencies (see Appendix) suggest that younger policy entrepreneurs tend to report more success than older, with the exception of the highest age category, where there is also skew toward success. Policy entrepreneurs from academia, the energy sector, the private sector (legal services and consulting), and an “other” job sector tend to report more success than would be expected by chance, whereas policy entrepreneurs who work in environmental advocacy, government, and community organizations (non-environmental) tend to report less success than expected by chance (Table 7).

Table 7 Fisher’s exact tests evaluating relationship between goal success and entrepreneur demographics
Full size table

Discussion

Policy entrepreneur scholarship is hampered by a dominance of event-based, small-n case studies of success, studies lacking clear parameters for identifying and distinguishing policy entrepreneurs from other policy actors, and research characterizing policy entrepreneurs from the descriptions of others rather than from policy entrepreneurs themselves. These dimensions make it harder to understand factors associated with entrepreneurial success versus failure (or degrees of either), test the generalizability of policy entrepreneurship theories, and understand policy entrepreneur motivations.

Recognizing these issues, we developed a novel approach for identifying policy entrepreneurs, hoping it could reveal, in ways other techniques have not, advocates experiencing varied levels of policy advocacy success and those who might lack the political connections or expertise of entrepreneurs nominated by elite informants, yet nonetheless try to influence policy. We also developed a protocol for distinguishing policy entrepreneurs from other actors in the policy process, defining them as individuals who invest in advocacy in a non-trivial way, seek to influence the choices and perceptions of government decision-makers around a focal issue (which we can measure because we query policy entrepreneurs themselves), exhibited at least one strategy or characteristic attributed to policy entrepreneurs by existing literature, and engaged in a threshold level of networking in pursuit of policy goals. Further, we examine a subset of potential entrepreneur characteristics that may influence policy advocacy success. Future research should consider a wider range of entrepreneurial characteristics.

Our elicitation approach relied on local newspapers for granular data on local policy and political processes, leveraging the close knowledge of local reporters and presses about community dynamics into insights about policy entrepreneurs who might otherwise be overlooked. However, this strategy may not be viable for long, as local newspapers are rapidly closing (Simonetti, 2022). If future scholars wanted to replicate this work in an area where local newspapers still exist, we suspect that a few changes might make the elicitation more efficient without a large impact on efficacy: Researchers could require eligible articles to have a larger number of keywords and could employ automated text analytics for some or most of the verifications of term relevance. We also suspect that this project could have been done more quickly had we been willing to pay online databases for personal email addresses, but we cannot speak to whether paywalled email addresses are more likely to be valid than ones identified without payment. If researchers interested in looking “under the hood” of local policy processes cannot find adequate local news coverage (but nonetheless wish to rely on secondary/archival sources rather than interviews or surveys due to concerns about respondent fatigue or non-response bias), we suggest considering public meeting minutes or draft planning documents (e.g., comprehensive plans) from local government, social media postings and pages, blogs, and local organization planning documents or websites. Many of the techniques we employed here could be adapted to these types of sources.

We hope that this attempt at a systematic and replicable policy entrepreneur elicitation protocol will stimulate conversation about what should or should not be included in such a protocol, and how. The appropriateness of the thresholds we chose (e.g., spending more than 5% of one’s annual personal budget on policy advocacy in one’s most active year of advocacy) can certainly be debated. There may also be weaknesses in some ways we operationalized variables: For example, we asked respondents whether they sought to raise public awareness about UOGD as a way of assessing whether they engaged in problem and solution framing, but directly using the word “framing” might better capture this quantity. We also proxied an actor’s investment of reputational resources in advocacy with evidence that other study participants identified them as a popular collaborator, but someone could have a strong reputation without being collaborative. Additionally, even if collaborator popularity effectively speaks to reputation, respondents may have collaborated with individuals who did not answer the survey and thus could not highlight the respondents as popular partners. Different policy domains or issues might require different operationalizations and thresholds. Our point is that policy entrepreneurship scholars should try to establish and employ transparent and specific elicitation techniques rooted in policy entrepreneurship theory. This is how scholarship in our field can advance rather than just accumulate.

With respect to the statistical results: It is intriguing that few of the variables drawn from policy entrepreneurship scholarship appear to explain success in UOGD-related advocacy in Pennsylvania and environs. Other scholarship finds policy entrepreneurs are more successful when they use more strategies or pursue more goals (e.g., Arnold, 2021); we do not reach the same conclusion. Conversely, the fact that some of the strategies typically attributed to policy entrepreneurs do appear to explain efficacy in drawing government officials’ attention to a policy issue affirms existing scholarship. All of these results, however, are tentative, produced from exploratory, uncontrolled, bivariate analysis. Although the number of policy entrepreneurs we studied (n = 37) is larger than many samples of such individuals in existing literature, it is not large enough to support more sophisticated analyses comparing the relative impact of different entrepreneurial strategies, characteristics, or demographics. Configurations of these variables may better explain levels of success than any one individually, a possibility to explore in future research. Future research should also be conducted across different policy issues with different dimensions (e.g., different levels of complexity, salience, political partisan polarization, and so on). It should examine policy entrepreneurship at different levels of government, since entrepreneurs may behave differently when targeting different levels, and the types actors who engage in policy entrepreneurship may vary across levels. We might observe similar dynamics when the magnitude of the change potentially delivered by a sought-after policy is more transformative or affects more people (e.g., the US Affordable Care Act) versus is more minor or affects fewer (e.g., a change in local zoning practices affecting UOGD siting in a city).

A particularly important path for future scholars is understanding to what extent our existing theories and assumptions about who policy entrepreneurs are and what they do can be validated with a large-n approach. The limitations on the present study’s analysis prevent us from claiming such validation. However, it is intriguing that the policy entrepreneurs we identified tended to be older White males with high levels of education, anecdotally consistent with accounts of policy entrepreneurs found in seminal works like King and Roberts (1992), Teske and Schneider (1994), and Mintrom and Vergari (1996) and consistent with a large body of scholarship that links being White, male, and educated with greater political self-efficacy and participation (e.g., Beaumont, 2011; Verba & Nie, 1972). Larger samples of policy entrepreneurs will afford more opportunities for comparison and potentially validation.

Conclusion

The technique we used to identify potential policy entrepreneurs was fairly arduous. Scores of research assistants spent nearly two years combing through texts and searching the Internet for thousands of people who might be involved in UOGD policy advocacy, ultimately resulting in survey responses from just 121 individuals. There certainly may be more efficient ways of finding policy entrepreneurs who experienced varying levels of success and were not necessarily elite or elite-adjacent, perhaps leveraging new developments in machine learning and analysis of text as data. We hope our effort can begin conversations about fruitful approaches.

Efforts to use quantitative, large-n approaches to studying policy entrepreneurship are needed. Policy entrepreneurs are a key feature in the extensive literature on the Multiple Streams Framework, but investigations of policy entrepreneurs specifically are much rarer. When they exist, they often involve case studies. Case studies are excellent for developing theory and have produced a rich body of ideas about policy entrepreneurs that require testing. Large-n quantitative studies are now important in assessing the validity and generalizability of theories across contexts and policy issues.

A large-n quantitative research program on policy entrepreneurs will succeed only if researchers can concretely define and agree upon what a policy entrepreneur is and does. By now scholars should recognize that we need to do more to identify policy entrepreneurs than beginning with a policy success and tracing back to find individual(s) who helped achieve it. Not only does this ignore the many individuals or groups who probably tried to engage in entrepreneurship, but with less success or visibility; it also blunts efforts to develop nuanced understanding of these advocates. Hopefully our research community can make progress in this area.

Understanding policy entrepreneurs is practically important because practitioners can use research results to help choose strategies and activities. Providing these insights requires us to study factors that vary with greater or lesser entrepreneurial success, which presupposes we know how to measure success. Here, we measure success using the respondent’s own ratings. Other studies ask informants to rate the success of policy entrepreneurs of whom they have knowledge, or measure success as the achievement of policy goals entrepreneurs pursued, like passage of particular laws. The choice of which success measure to use seems to come down to feasibility or data accessibility. However, we urge scholars to make an effort to measure success multiple ways in order to learn how well these measures match one another, or where and how they diverge. Such analysis could help researchers understand if some success measures are better suited than others to particular contexts, policies, or policy entrepreneurs.

A final area ripe for progress is learning from other literatures and disciplines studying how people and groups influence policy processes. Political science scholarship on interest groups is extensive, but studies of policy entrepreneurship rarely connect with it. Organizational and management scholars examine how leaders of firms and organizations emerge and behave, and presumably some of those leaders try to influence policy. A leadership lens likely could enrich our understanding of policy entrepreneurs, as could perspectives from scholarship on social movements, where researchers examine how advocates and advocacy groups function. The policy entrepreneurship literature often argues that entrepreneurs are most efficacious when they build networks and coalitions, span boundaries, and bridge structural holes. Policy entrepreneurship scholars should do the same.