Policy transfer routes: an evidence-based conceptual model to explain policy adoption

ABSTRACT Although studies on policy transfer have expanded, a general and comprehensive understanding of policy transfer is lacking. This study offers an evidence-based explanation of policy transfer processes. We extracted constraining and facilitating factors from 180 empirical studies using PRISMA (Preferred Reporting Items for Systematic reviews and Meta-Analysis) and aggregated these factors into a conceptual framework. We synthesize our findings in four “transfer routes”. We conclude that actors could shape a subset of those factors by taking certain decisions regarding transferability, adoptability and process design, albeit within the boundaries of the environment.


Introduction
"The Netherlands has its water management adequately organised. […]This approach is valuable to share with other countries and inspire the development of their climate adaptation strategies." stated Wim Kuijken, the Delta Commissioner (a special government commissioner) in the International Water Ambition (Ministerie I&M 2016). With this ambition, the Dutch government articulates its ambition to transfer the successful Dutch water governance policy to other countries. Such occasions of policy transfer are believed to be increasingly occurring nowadays due to the popularity of evidence-based policy-making (Marsden and Stead 2011;Legrand 2012) and the widespread use of modern information and communication means (e.g. the Internet) that allows access to information about policies elsewhere (Dolowitz 2006). However, a comprehensive explanation of how this process of policy transfer works is still lacking.
Existing research is limited to one or a few cases or uses deductive approaches to investigate limited elements of the transfer process. Most studies focus on single case descriptions (e.g. Dolowitz and Medearis 2009). This focus on individual cases limits the generalizability of findings. Furthermore, studies are usually deductive in nature: authors develop a theoretical framework and test this framework in a case study or apply an existing analytical framework to a different environment (e.g. Dixon 2007). Such research is valuable, but cannot provide general explanations of the process of policy transfer. Aggregating individual cases and explanations might provide such answers. Yet, several reviews have been conducted on policy transfer in the past decade, but none of them systematically documented the factors that influence the policy transfer process and thereby the outcomes of this process. We will elaborate on this in the next section.
We aim to contribute to the understanding of the process of policy transfer, by conducting a systematic review in order to develop a more evidence-based framework of policy transfer. We will identify and aggregate evidence-based factors (i.e. that were empirically identified) that have an influence on policy transfer processes. By focussing on empirical research rather than new conceptualizations, we will be able to answer the question why some mobilized policies are adopted and successfully implemented while others are not. Our contribution can serve as a reference point for scholars studying policy transfer processes and aid practitioners in enhancing the efficiency and effectiveness of policy transfer processes.
The remainder of this article consists of four sections. In "Conceptual demarcation of policy transfer" section, we will conceptually explore policy transfer, before introducing the Preferred Reporting Items for Systematic reviews and Meta-Analysis (PRISMA) method that we used to review the literature systematically in "Method" section. We present all identified factors in "Results: overview of presented factors and towards a conceptual framework" section by introducing a conceptual framework on policy transfer and discussing each factor's contribution to transfer. The main added value of this review lies in the subsequent deduction of policy transfer routes. In the "Discussion" section, we present four common routes that policy transfer processes can take from initiation to outcome. Finally, we conclude that factors early in the process may predetermine later transfer outcomes.

Conceptual demarcation of policy transfer
Over the past decades, the body of literature on policy convergence has expanded and the number of terms describing policy spreading has increased concurrently. In this section, we will explore the most important concepts and position our study.
Several terms exist to denote processes of policy spreading, such as policy transfer, policy diffusion and policy mobility (Stone 2001;Prince 2012) and policy adaptation and policy translation (Mukhtarov 2014). These terms have overlapping meaning but nuances can be found in their understanding of convergence (Marsh and Sharman 2009). In this article, we are interested in the "action-oriented intentional activity" (Evans and Davies 1999) of spreading a policy and the use of "knowledge about policies, administrative arrangements, institutions, etc. in one time and/or place" to develop policies in another time or place (Dolowitz and Marsh 1996, 344). Furthermore, we consider policy transfer to take place between autonomous actors that can make sovereign decisions. To clarify this definition, we will compare two examples of policy transfer in the European Union (EU). According to our definition, policy transfer encompasses voluntary or pressured adoption of non-obligatory European norms. In contrast, our study disregards adoption of mandatory EU legislation by EU member states. This distinction is quite relevant, given that various bodies of literature have addressed the coercive imposition of standards on other, mainly developing countries. Examples include literature on institutional transformation and studies on conditionality, including the structural adjustments policy of the World Bank (Stone 2016). Although we acknowledge the existence and potential relevance of these studies, we will limit this review to studies originating from the policy transfer, diffusion and mobility literature. We elaborate on the choice for these fields in "Method" section. Summarizing, we position this study within the policy transfer tradition, but will incorporate studies using related terms that match our perception of action-oriented, intentional policy transfer.
The body of literature on policy transfer has been reviewed several times before, although these reviews did not provide satisfying explanations on how the process of policy transfer is affected by internal and external influences. Some early reviews focused on understanding the concept of policy transfer (e.g. Bennett 1991;Dolowitz and Marsh 1996). In 2000, this resulted in Dolowitz and Marsh's famous framework of seven questions about policy transfer, coming down to questioning who engages in policy transfer for what reasons, defining what is being transferred from where to where and describing the process of policy transfer (Dolowitz and Marsh 2000). This latter topic concerns the different degrees of transfer, what restricts or facilitates the process and how this process relates to "success" and "failure" of transfer. More recent contributions aimed to describe the conceptual refinements (Benson and Jordan 2011), innovations in the field (e.g. Peck and Theodore 2012;Stone 2012;Temenos and McCann 2012) or re-assess influential contributions to the literature (Stone 2016). Although these reviews provide a clear overview of the (conceptual) evolution in the scholarly field of policy transfer, they do not provide an overall explanation of policy transfer processes and outcomes.
Authors have listed (types of) factors that constrain policy transfer (e.g. Evans 2009), thereby addressing the question in Dolowitz and Marsh's framework on factors facilitating or restricting transfer processes. However, a comprehensive, empirical overview of these factors is lacking, which is surprising given the crucial role that these factors play in explaining the success of policy transfer (Marsh and Sharman 2009). In subsequent phases of the policy transfer process different factors play a role, as some authors identified (e.g. Kerlin 2009; Gullberg and Bang 2015). Moreover, certain factors become decisive during specific phases of a transfer process (Sugiyama 2016). Stone (2016) noted that some of these causes of failure or success are recurrent, such as the role of context in transferability, the role of actors to improve or complicate the policy transfer process and the role of learning to establish the transfer of knowledge. Consequently, we wish to provide a comprehensive overview of factors that affect the process of policy transfer and, based on these factors, connect internal and external influences on policy transfer processes.

Method
This article aims to provide an overview of empirical studies since 1996 (when Dolowitz and Marsh's seminal article was published) through a systematic literature review. This article makes use of the PRISMA method, an abbreviation of Preferred Reporting Items for Systematic reviews and Meta-Analysis (see Shamseer et al. 2015). In this section, we will elaborate on our application of this method, for a detailed explanation of this method please consult Shamseer et al. (2015).

Search strategy
We used the search terms policy transfer, policy diffusion and policy mobility to systematically review the literature. In the introduction, various alternatives to the term policy transfer were introduced, being policy diffusion, policy mobility, policy translation, policy convergence and policy adaptation. These terms are often used interchangeably (Stone 2001). However, the definitions of these six terms carry important nuances that increase or decrease the relevance of terms for our study. We compared definitions and randomly sampled 20 items per search term to evaluate whether this search term contributes to our cause. Policy translation mainly co-occurs with policy transfer. The remaining unique items address translation of research findings into policy or translation of policy decisions at a higher administrative level to a lower, executing administrative level. The search term policy adaptation returned essentially noise, as most items concerned policies of climate change adaptation. Policy convergence is "the tendency of societies to grow more alike, to develop similarities in structures, processes, and performances" (Kerr 1983, 3, through Bennett 1991 and lacks the intentional and action-oriented nature of policy transfer. In addition, policy convergence focusses on results, rather than the processes that are central in policy transfer. As a result, we excluded policy translation, policy convergence and policy adaptation as search terms. We applied four search strategies. First, we performed an electronic search in two online databases: Scopus and Web of Knowledge. We limited the searches to peerreviewed articles only, to ensure a certain level of quality of included items. Only English-language articles were included. Dolowitz and Marsh's seminal article (1996) unarguably has been influential (Benson and Jordan 2011) and is therefore taken as a starting point for this review. Studies from 1996 up to and including 2016 are thus considered. Second, we entered the search terms in the databases of eight journals with a non-Anglophone geographic focus to compensate for a domination of items from the UK, USA and Western Europe in the search results. The decision to include English-language articles only potentially caused this domination. We selected the journals that focus on Latin America, Africa, Asia and Eastern Europe and that occurred most frequently in the list of journals returned by the database search. The latter criterion was introduced in order to ensure the journal had published about policy transfer. Third, relevant books and book chapters were identified using the electronic databases Web of Science and Scopus. Finally, we asked experts in the fields of policy transfer, diffusion and mobility to examine our references and asked them whether they missed any item. Three experts suggested 56 novel items. Figure 1 summarizes the four search strategies and presents the number of books and peer-reviewed articles that were identified with each strategy.
The relevance of identified items was assessed using a list of five eligibility criteria. The first criterion is that only empirical findings are eligible for inclusion, given our goal to develop an evidence-based account of factors that affect policy transfer. The second criterion is that items should discuss agenda setting or policy formulation phases. Implementation or effect evaluation of transferred policies is out of scope. With similar reasoning, we exclude implementation, spreading or enforcement at lower government levels (e.g. local level) of policy issued by higher levels of government (e.g. federation or union). The third criterion is that a study should identify factors as independent variables. This review serves to identify factors that influence the policy transfer process, hence policy transfer is considered the dependent variable in this review. The review is thus limited to articles that investigate independent variables (i.e. the factors that explain policy transfer). The fourth criterion is that the transfer should be taking place or should have taken place, excluding studies on transferability (i.e. export) and suitability (i.e. import) potential of policy. Finally, the studied transfer should be the result of an intentional process. An unintentional policy transfer includes the convergence of policies following changed global policy paradigms.

Item assessment
The resulting items were assessed in a two-step approach. Firstly, we assessed titles and abstracts and excluded items that did not meet the eligibility criteria. Items were always included in case of doubt. Items that passed this first stage were subjected to a full-text read. A total of 78 publications ultimately failed the criteria and was rejected in this final stage after all. Figure 1 presents the flowchart of item selection.

Data analysis
All items were inductively coded and the final codes are the result of an iterative process of coding and re-coding in Atlas.ti software. We used a coding process that combined elements of selective and axial coding (Boeije 2010). An initial set of codes was pre-determined following Benson and Jordan (2011), identifying the policy (i.e. transfer object), type of actors, the mechanism of transfer (voluntary, conditional or coercive) and the outcome (non-adoption, imitation, adaptation or inspiration). We also collected data about the study itself: the number of transfers studied, the methods used, the country of the first author's institute and of the origin and destination of the transferred policy.
We used an open coding approach to code factors, basing the initial codes on formulations of the item's authors. In other words: factors that are included in the framework are not included because we found them relevant, but because they were identified by other authors in their studies. A phrase like "A lack of financial factors obstructed the transfer process." would thus receive the code "constraining effect" alongside the original code "Lack of financial resources". We checked for similar codes and aggregated these factors into one. The original code in our example was later aggregated into "Adoptability: resources". In subsequent iterative rounds, the original factor-codes were divided, aggregated or renamed.
The results section is primarily based on the analysis of these codes. We present both a quantitative description of the data analysis and an explanation of policy transfer success of failure in the form of a conceptual framework.

Results: overview of presented factors and towards a conceptual framework
After presenting general results in "Describing the research focus of included items" section, we move to the presentation and integration of factors into a conceptual framework in "Towards a conceptual model of policy transfer" section.

Describing the research focus of included items
A total of 180 articles is included in the review. See Appendix for the remaining references which are not referred to in the text. These articles originate from 124 different journals. Although journals from diverse fields of research are included, such as Urban Studies (4 times), most frequently cited journals relate to policy analysis. These other journals are Policy Studies (8 times), The Policy Studies Journal (5), Journal of European Public Policy (5), Governance (4), Public Administration and Development (4), Journal of Communist Studies and Transition Politics (4) and Journal of Comparative Policy Analysis: Research and Practice (4).
Policy transfer research in Western countries dominates the included items with Anglophone countries as its centre of gravity. A vast majority of the studies was conducted by first authors affiliated in Anglophone countries, headed by the UK (26%, N = 180), USA (21%), Canada (9%) and Australia (6%), and in Western-European countries such as the Germany (6%) and Netherlands (5%). Moreover, the empirical studies included in the review focus on transfer from, to and between these countries as well. The UK and USA are studied most often, both as source and as destination of a transferred policy.
According to Benson and Jordan (2011), the field of policy transfer research diverged from its initial focus on transfer between nation states. Although such a divergence is observable on a conceptual level, only a small proportion of actual empirical studies involves non-state actors. Out of 180, 173 items discuss at least one state actor, while only 34 discuss one or multiple non-state actors (such as researchers, consultants or NGOs). State actors are generally national governments or international governmental organizations (IGOs, such as the OECD, EU or World Bank) in policy diffusion and transfer studies. Policy mobility literature added studies on transfer between local governments. The results further suggest that national governments are generally senders, receivers and initiators of transfer processes. They are rarely facilitators or transfer agents, while IGOs primarily act as senders and facilitators rather than receivers of policy.
The included items most often consist of in-depth analysis of policy transfer. Most authors study a specific instance of policy transfer (80%, N = 180), although 20% of the articles compare multiple different transfer activities. The vast majority of studies is qualitative in nature, using interviews (57%), observations (15%) and document analysis (9%) as most reported research instruments. However, roughly one in four articles does not report on the methods used. Similarly, only 73% of the articles defines an analytical structure using a theoretical framework, model or concepts. One-third of the articles that specify their theoretical basis develops or tests a new theoretical framework, while most built on existing concepts. Eleven studies do not even mention neither their methods for data collection nor the theoretical underpinnings of their analysis.
Summarizing, the included body of knowledge in our review will be most representative to policy transfer between or originating from governments in Western cultures, although factors resulting from the presence of other actors and geographic areas are covered as well.

Towards a conceptual model of policy transfer
The previous paragraphs described the results of the selective coding process. As described in "Method" section, axial coding of factors was based on the original authors' description of a factor and on the constraining or facilitating effect of this factor. We aggregated factors into four groups (see Table 1).
The aggregated factors form the building blocks for a simple conceptual framework, see Figure 2. Environmental factors create the context that delimits the playing field for factors of the other building blocks, namely Transferability of the policy itself and the sending actor, Process Design of the interaction between sending and receiving actors and Adoptability of the policy in the adopting context and eventually the adoption (or non-adoption) of the transferred policy. The arrows indicate that outcomes of factors in these building blocks influence factors in other building blocks. In the following section, we will discuss the building blocks of this model and the associated factors in-depth.

Environmental factors
The first building block concerns environmental factors. As discussed in the introduction, factors can be present in various phases of the transfer process. The environmental factors play a role in all phases. We distinguish between the policy arena, the subsystem and the general context. See Table 2.
Most environmental factors appeared in the policy arena. The policy directly shapes the freedom of movement of the key actors. This includes the Zeitgeist, existence or absence of competition with peers and the political climate (i.e. who forms the government). A policy might be at the right place at the right time (e.g. Cook and Ward 2012) or make use of a policy winding due to right timing (e.g. Busch 2005). A change of government can change the policy arena in favourable ways and open up a transfer process (e.g. Delpeuch and Vassileva 2016), but may suddenly terminate nearly completed transfers as well (e.g. Dussauge-Laguna 2012).
The subsystem relates to the availability of alternative policies and the institutional and political context. Policy transfer can provide an alternative to an actor. An example is how the EU's renewable energy policies provided an alternative to existing reliance on Russian gas in Eastern European countries (Ademmer 2014). However, a transfer process may be disturbed by the availability of alternative policies to the transferred policy (e.g. in the case of competing health policies, see Clarke 2013) or flourish in the absence of competing policies (e.g. in the case of IWRM, see Allouche forthcoming).
The general context sets the boundary conditions for the policy transfer actors and is formed by biophysical (e.g. Attard and Enoch 2011), cultural (e.g. Tsakatika 2012) and socio-economic (e.g. Edwards and Beech 2016) conditions. For example, the British and Australian Labour parties operated in similar neo-liberal systems but economic adversity required British Labour to adopt a dramatically different political position (Edwards and Beech 2016).  The table shows how many times the factor is mentioned in total and how many different items mention the factor. The percentage reflects the latter. a N = 180, please note that percentages might add up to above 100%.
Overall, nearly one in four articles mentions factors in the policy arena, which suggests that the policy arena might be the most important aspect of the environment to pay attention to in policy transfer processes.

Transferability
The second building block of the framework addresses the transferability of the policy. We identified several factors in the review data that relate to the transferability of the transfer object. Transferability-related factors dominate early stages of the transfer process, corresponding to the exploration phase. See Table 3.
Transferability is first determined by the ability of the source actor to convey policies. The source actor can have a positive image (e.g. Khirfan and Jaffer 2014) or a less positive one (Bok 2014), resulting in respectively stimulating and discouraging adoption of policies from this actor. Closely connected to a source actor's reputation is its legitimacy to transfer. Authors especially reported legitimacy issues faced by the EU when transferring to neighbouring countries outside the Union (Radaelli 2000;Xheneti and Kitching 2011;Vezirgiannidou 2015;Onursal-Beşgül 2016). Ademmer and Börzel (2013) provide yet an alternative explanation, namely that high compliance costs of adopting EU policies may outweigh benefits for non-EU countries such as Turkey.
Transfer processes are more easily established when there are existing relations between source, adopting and third-party actors. Relations increase the acquaintance with policies elsewhere through membership of an international organizations such as the EU or OECD (e.g. Ayoub 2014; Oanc 2015) and policy networks (Sloam 2005), through colonial history (Smith et al. 2002) or through trade and cooperation relations (Randma-Liiv and Kruusenberg 2012; Jinnah and Lindsay 2016).
The tolerance of the adopting actor further determines transferability. This ability is determined by the receptivity and decision-making power of this actor. Receptivity denotes the openness of the receiving actor to consider policies from elsewhere. A lack of openness limits the potential for transfer from the start (e.g. Dolowitz and Medearis 2009;Keating and Cairney 2012). Openness is necessary but not sufficient, as actors may lack decision-making power. Adopting actors can be dependent on other states or donors (e.g. Ohemeng 2010). Such dependency reduces their sovereignty and can benefit (coercive) policy transfer within or limit transfer to parties outside the spheres of influence (Ademmer 2014). Vezirgiannidou (2015) further demonstrated that actors are less vulnerable to external pressures when they are able to make sovereign decisions. Finally, policy features and normative fit determine transferability. Flexibility (Kerlin 2009; Lavenex 2014) and low context dependency (Jong and Bao 2007;De Loë et al. 2016) of policies increase the range of possible applications. Especially infrastructure policies may be tailored to specific biophysical conditions, reducing their transferability (Michaels and de Loë 2010;Attard and Enoch 2011). On the contrary, transferability increases when the policy matches the values (e.g. Chapman and Greenaway 2006) and political objectives of the receiving actor (e.g. Clavier 2010). Then again, policies with a reputation of proven effectiveness and success are popular transfer objects for policymakers because such policies are justified by their (perceived) success (e.g. Ovodenko and Keohane 2012;Metz and Fischer 2016).
Summarizing, normative fit and policy reputation appear more often as a factor in studies on policy transfer compared to, for example, the characteristics of the policy itself.

Process design
The third building block concerns the interaction between transfer actors, shaped by a particular process design. Process design concerns the set-up of interaction between actors exchanging knowledge and in adopting and implementing the transferred policy (Table 4).
Relations to various actors in the form of building coalitions and engaging all key actors are essential in the process design. Policy consensus enhances the ultimate success of a transferred policy, whether this support is built around a broad coalition of domestic stakeholders (e.g. Müller and Slominski 2016) or external support (e.g. De Loë et al. 2016). An adopted policy can fail in the final implementation phase as well, because support from executive officials was not secured earlier phases (Šimić Banović 2015). Key players, such as policy entrepreneurs and political leadership, can control this transfer process. This leadership can follow from existing leaders' characteristics, such as charisma (e.g. Ohemeng 2010), or can be managed through strategic human resource management. An example of such strategic management is the employment of experienced West-German policy officials in East-Germany after the latter adopted West-German institutions after reunification in the nineties (Welsh 2010).
In the previous section, we introduced the need of openness to external policies by the adopting actor. While exchanging knowledge, actors engaged in the process need to be open to mutual understanding of and adaptation to values, practices and beliefs of the other actors. Transfer agents should look for both similarities and differences (Hoyt 2006, 238). Insufficient adaptation of policies to the local context may result in inappropriate policy transfer. Jong and Bao (2007) argue that mutual understanding of cultural differences is in fact more important than having a similar culture. This understanding can be achieved by two-way instead of one-way communication as Chung, Park, and Wilding (2016;Park, Wilding, and Chung 2014) demonstrated. Language is an important factor in (preventing) miscommunication (Welsh 2010;Xheneti and Kitching 2011;Fawcett and Marsh 2013). The role of language is disputed though, as Stadelmann and Castro found no positive influence of having the same language on climate policy diffusion in a large-N study (2014).
The management of the transfer network is important (Dolowitz and Medearis 2009;Timms 2011). The absence of a clear hierarchy may constrain the transfer (Vinke-De Kruijf, Augustijn, and Bressers 2012). On the other hand, a flat organizational structure with high autonomy for transfer agents (Khirfan and Jaffer 2014) can be beneficial as it may stimulate innovation (Khirfan and Jaffer 2014). Nonetheless, several studies stress the importance of having a dense policy network. Characteristics of such density include informal relations (e.g. Chien and Ho 2011) and face-to-face interaction (Vinke-De Kruijf, Augustijn, and Bressers 2012). The existence of a dense network ensures that resources can be mobilized (Rodgers 2014). When reserving too little time for the exchange process, the resulting transfer may be superficial only (Pojani and Stead 2015).
Regardless of how formal the exchange process is organized, the process evolves in a certain transfer type. Following existing typologies (Rose 1991;Goldfinch and Roberts 2013), we distinguish between imitation, adaptation and inspiration as adoption mechanisms. Imitationalso referred to as copying, mimicking or harmonizationis considered a "quick fix" for policy-makers in urgent need of a solution and is associated with several forms of failed transfer (Toens and Landwehr 2009;Crot 2010). Adaptation refers to the incorporation of the basic model with changes and includes emulation along with the more recent terms of translation and assemblage. Bulmer and Padgett (2005) suggest that bargaining results in transfer that is the synthesis of several policies and relies less strongly on one source. Inspiration results in the creation of new policies that are based on (elements of) policies from elsewhere, such as the "hybrid system created that drew on US and Australian examples for inspiration and copied selected aspects of policies and statutes" (Michaels and de Loë 2010, 501). Learning is often associated with successful transfers (e.g. Biesenbender and Tosun 2014). Additionally, we include negative lessons in the category of "other" adoption models. Negative lessons as outcome refer to the decision to seek alternatives because of the limited success of the originally considered policy. Finally, the level of coercion has an influence on the policy transfer process. External pressures can enhance the acceptance of certain policy norms by other countries, for example, when these norms are part of a trade agreement (Jinnah and Lindsay 2016), but may initiate transfers that are inappropriate for the objectives of the adopter (Parnini 2009) or transfers that are not completed (Webber 2015). Conditional transfers are formally voluntary but in practice the result of external pressure. An example concerns the conditional loans from the World Bank (Larmour 2002). We will discuss the relation between the level of coercion, the exchange mechanism and the adoption or non-adoption of the transferred policy in the subsection "Policy (non-)adoption".
Outlining the process design, it seems that actor relations received the most attention in the body of literature studied, in terms of both networks, key individuals and coalitions formed.

Adoptability
The final building block is adoptability of the transferred policy and Table 5 presents its factors. These factors mainly occurred during later phases of the policy-making process.
Suitability of the transferred policy plays a key role in policy adoption. In the first building block, we introduced the normative fit of a policy as part of its transferability. Besides this normative fit, the institutional fit plays an important role as well. A transfer object is adopted in a certain institutional context. If certain pre-requisites are met, a smooth policy integration may be possible. For example, transfer of educational norms and reforms from the EU to Turkey was possible due to the America-based model of higher education in Turkey (Onursal-Beşgül 2016). When actors fail to meet these pre-requisites, implementation failure is inevitable, as was the case in transfer to India of industry reforms due to the unforeseen lack of a regulatory agency in India (Xu 2005). The flexibility of a policy can reduce the mismatch to a certain degree. Policies with a fixed core but high flexibility in implementation will be adopted easier at destinations with a reduced normative or institutional fit (Kerlin 2009). Moreover, simple or simplified policies will require less organizational capacity and are therefore less prone to failure in this phase (Lepinard 2016).
This adopting capacity consists of the expertise to search and implement external policies and the organizational capacity to evaluate policies. Extensive policy evaluation will ensure that policy learning takes place and that only policy with a good fit is transferred (Fawcett and Marsh 2013). Such evaluation also enables negative lesson drawing (Timms 2011). An important requirement is that destination actors are sufficiently equipped to organize the process of searching and implementing external policies (Randma-Liiv and Kruusenberg 2012). Such organization "requires significant commitment by politicians and, especially, public servants to investigating its operation ( … )" (Fawcett and Marsh 2013, 184).
Sufficient resources are needed to adopt and integrate a transferred policy. Such resources could be time and human or financial resources. These resources are required in all phases. However, a lack of resources is often mentioned to cause failure in the adoption phase, especially lack of time and financial resources (e.g. Marsden et al. 2012). Even with sufficient organizational capacity and favourable policy characteristics, the adopting actor should be able to change the policy course. Previously made policy decisions create path dependency. Path dependency can be beneficial to the transfer (e.g. Gullberg and Bang 2015) or abort the transfer at any time due to inability to change the policy course (e.g. Zhang 2012). In the final phases of the transfer, a lack of support from decision makers might be catastrophic. The transfer process itself may not have been hampered by this lack, but restricts the outcome of policy transfer. An example is the fruitful exchange of ideas between Dutch and Japanese train operators where the Japanese decision-makers had decided on a different policy before the transfer was even completed (van de Velde 2013). Involving or ensuring access to decision-makers early in the process facilitates the process (Kerlin 2009).
The most cited factors to influence policy transfer in this building block concern the compatibility of the transferred policy, both with respect to integrating this policy in institutional structures and to its ability to intercalate with existing policy paths.

Policy non-adoption
The final element of our model concerns the outcome of the policy transfer process in terms of adoption or non-adoption. We distinguish between successful adoption, formal adoption and non-adoption of the transferred policy. Successful adoption refers to a completed transfer process where the receiving policy-maker adopted the policy. Ultimately, successful transfers achieve political goals as well but that element is left outside the scope of this study. Non-adoption includes all transfers that were considered but never initiated or that were initiated but aborted along the way. Formal adoption finally describes policy transfers where the policy was formally adopted but was not implemented or enforced.
We counted how often adoption, formal adoption and non-adoption were specified in the studies and plotted the outcomes against types (Table 6) and the level of coercion of these transfers (Table 7). We based the distinctions between transfer types and between levels of coercion on the process design factors with the same labels. As could be seen in Table 7, most included articles described cases of successful policy transfer and the majority of review items concerns more voluntary transfer. A similar trend can be observed for the transfer type. Imitation and adaptation are most commonly studied,  but imitation results more often in non-adoption when compared to adaptation and inspiration.
There are two explanations for this majority of successful and voluntary transfers in the studies. The first explanation is that such cases receive more attention, making them more feasible as study object and therefore create overrepresentation of such cases in empirical studies. A second explanation could be that voluntary transfers are more likely to result in successful adoption. Our data support previous claims (e.g. Ogden, Walt, and Lush 2003) that coercive or conditional transfers are more likely to result in unsuccessful or formal transfer than voluntary transfers. Webber (2015) even draws a direct relation between the coercive nature of the transfer and the resulting incomplete transfer. An explanation is that the receiving actor is merely interested in complying with the conditions for other purposes, rather than adopting policy out of genuine interest. However, especially in developing countries a lack of compliance might be the result of lacking infrastructure to implement a certain policy, rather than lacking the willingness to do so (Bennett et al. 2015).

Discussion and conclusions
General remarks on the framework To address the lack of an evidence-based explanation of policy transfer processes, we based our review of empirical studies on policy transfer. These studies are dominated by Anglophone transfers and were usually conducted using interviews and document analysis to study one or multiple cases of policy transfer. These cases predominantly involved state actors rather than non-state actors and we observed a concentration of voluntary transfer that resulted in policy adoption. We identified four clusters of factors (transferability, process design, adoptability, environmental factors) that influence this policy (non-)adoption. In Figure 3, we present the full conceptual framework based on the results, with more details regarding factors.
Two results stand out in particular. First, factors related to process design play a crucial role in policy transfer processes and the ultimate success or failure of such a process. Especially the selection of the right actors, both individuals and coalition-wise, plays a key role. Second, previous studies attribute a considerable value to similarities or differences in context (e.g. Stone 2016). The review supports this claim, but also indicates that the policy arena is more often a decisive enabler or barrier for policy transfer. Related concepts such as normative and institutional fit are more important than the general context and so is policy reputation. These results suggest that the emphasis on context as explanation for policy transfer outcomes should be accompanied by actor selection, institutional and normative fit and the present-day political situation. More importantly, the involved actors can control these factors to a certain extent, in contrary to (general) context. Some factors are an important barrier when they are absent and a key facilitator when present, or vice versa. An example of such a factor is having sufficient resources. Control over such factors is important for those involved in policy transfer, and the insights of our model pave the way for purposeful management of policy transfer processes.
We refrained from a phase-based model, but conclude that most factors in the transferability building block played key roles during earlier phases of the transfer, while factors related to adoptability became decisive during later phases. The temporal role of factors was suggested before (e.g. Kerlin 2009; Gullberg and Bang 2015) and this review systematically mapped them. This temporal element is interesting, because they may direct transfer agents' attention during, for example, agenda setting.
In summary, we can conclude that some factors are more important than other factors and that actors can control the nature of these factors to a certain extent. Furthermore, conditions at the early phases of policy transfer may eventually affect the final results in terms of transfer and adoption mechanisms. As policy transfer processes require extensive resources, such as time, money and human resources, the framework can also help identify challenges in policy transfer that can be used to enhance the efficiency and effectiveness transferring policies, thus reducing the risk of inappropriate, incomplete or uninformed transfers (Dolowitz and Marsh 2000). Coming research could further analyse the nature of the relation between and the relative importance of factors.

Route varieties of policy transfer
Based on the insights from our conceptual framework, we will discuss four policy transfer routes that policy transfer can take from initial conditions to final outcomes. The framework provides a comprehensive overview of factors that other authors have found, but this section reflects the deeper insights we gained by scrutinizing their studies. When we combine the factors discussed in the framework with the policy transfer routes, we cannot neglect that the initial conditions of a policy transfer process influence the outcomes of that trajectory. These "route varieties" include opportunistic, branded, pressured and learning policy transfer. In the next paragraphs, we describe these routes in detail. These details are summarized in Table 8   Opportunistic policy transfer is characterized by bounded searches, considering a single external policy that mainly serves to justify policy measures "at home". The adopting actor is usually the initiator of the transfer attempt. Political urgency to act can trigger opportunistic transfer, of which the imminent threat of a financial crisis is an iconic example. These "quick fixes" usually rely on imitation as exchange mechanism, as time is highly constraint and a limiting factor (O'Hara 2008). Constraining factors are usually found in the process design building block, as the exchange is characterized by unidirectional flow of information accompanied by limited knowledge about key issues and poor policy evaluation. Opportunistic transfers might result in inappropriate or uninformed transfer. Adoptability seems to be a strength of opportunistic transfers, making use of environmental factors, but poor process design limits its successes.
Branded policy transfer is initiated after policy marketing or relies on existing (bilateral) relations between source and adopting actors. Such strategies enhance the occurrence of transfer, but these strategies are not without risk. Actors' reputations can be improved by lists of excellence, for example by city ranking on various policy topics or policy promotion. The source actor plays a dominant role and is likely to act as a transfer agent to influence formal adoption. These transfers depend on imitation and limited adaptation. Such framing can enhance the perceived normative fit and fuel diffusion, although such marketed policies risk ending up as inappropriate transfer due to the limited evaluation of the policy. Similarly, Wood (2015) warns that existing contacts can increase the transferability but not necessarily result in the most suitable policy. Transferability is artificially propelled and challenges are mainly encountered in the adoptability building block.
Pressured policy transfer ranges from transfer through peer-pressure  to transfers based on limited sovereignty or full dependency of adopting  (2012) Ogden, Walt, and Lush (2003) Cohen-Vogel and Ingle (2007) Wood (2014) Note: We also present an exemplary reference per route.
actors. Conditional transfers often involve third-party actors, such as IGOs. Such transfers benefit harmonization and are a means for more powerful actors to influence decisionmaking elsewhere. These transfers risk inappropriate transfer and associations with neo-colonialist influences. Several authors (e.g. Jinnah and Lindsay 2016) mention limited financial resources as a major constrain, thus suggesting that financial support might be more effective than imposing financial inducements to laggards. Transferability is enforced in pressured transfer and major challenges are found in the adoptability block, although elements of the process design play a distinct role as well.
Finally, a process of mutual learning may result in policy transfer. Any actor can initiate learning transfers, but decision-making remains exclusive to the receiving actor. The search for policies is usually bounded and extensive policy evaluation or piloting is part of the process, resulting in well-considered decisions and broad support coalitions. Consequently, this route requires considerable resources and is therefore not always attainable. A full learning process results in adaptation, inspiration or mutual influencing and is less vulnerable for incomplete or uninformed policy transfer. Learning can make pressured transfers successful, although a "learning paradox" exists (Toens and Landwehr 2009;Evans and Barakat 2012). Unsuitable knowledge may be internalized, resulting in inappropriate transfer. Nonetheless, learning is generally associated with improved transfer because learning internalizes procedures of policy formation. The process design is the strength of learning transfer and requires attention throughout the exchange.

Limitations and a future research agenda
Our framework is largely based on successful, voluntary transfers. This skewness may be the result of the eligibility that excluded studies on implementation and enforcement of adopted policies, which may describe unsuccessful cases. This study could be extended by including articles that focus on these steps of the policy cycle to increase our understanding of the relations between factors and outcome by sharpening the framework and routes that we introduced here.
The PRISMA method proved to be effective in identifying a wide range of factors described in existing studies. The prescribed selection procedure leads to the inclusion of articles that would have remained unnoticed with other (systematic) review methods. However, we acknowledge that there are other bodies of literature "out there" that may address the same phenomenon but use different labels, such as literature on transformation of institutions. The inclusion of non-English literature could further add to this review, as Anglophone studies now dominate our results.
During this review, we noticed that the term transfer mechanism can refer to different phenomena: the model of reproduction (e.g. inspiration, Theobald and Kern 2011), the voluntary or coercive nature of transfer (e.g. Keating and Cairney 2012), the mode of exchange (e.g. policy learning, see Nicholson-Crotty and Carley 2015) and the channels through which policies spread (e.g. Nazif-Muñoz 2015). As described in the introduction, the same goes for the various terms for policy spreading. These terminological voids are problematic, especially given the continuous assimilation of policy transfer studies in other research fields (Benson and Jordan 2011). We distinguished between transfer types and level of coercion to describe respectively the model of reproduction and the degree of coercion in transfer and encourage future studies to use consistent terminology.
Finally, actors engaged in transfer make decisions that result in distinct circumstances. These circumstances eventually induce various factors that line up for a certain trajectory. This study has indicated that several factors shape the circumstances of policy transfer and thereby leads transfer process to line up for certain policy routes. Thinking in terms of policy transfer routes highlights that there is a relation between the transfer type, mode of exchange and the ultimate policy success. Future studies can build on our framework and route varieties by further operationalization of factors and by establishing the added value of this framework in various case studies. Alternatively, research may focus on the ability to shift between policy transfer routes. The identified learning route further confirms the preference for "policy translation" or "learning" (e.g. Stone 2001) over imitation and adaptation. A question that remains is how scientists and policymakers in practice can use these insights to actively anticipate these circumstances in a policy transfer process.
This study increases our understanding of how the process of policy transfer relates to success or failure and what constraints or facilitates these processes. We conclude that actors could shape some of the identified factors by taking certain decisions regarding transferability, adoptability and process design, albeit within the boundaries of the environment. Professionals may use these insights in managing transfer processes and form a conceptual departing point to study how these processes can be steered more consciously.