What Matters for the Successful Policy Transfer? Empirical Evidence From South Korea’s Knowledge Sharing Program(KSP)

1. Introduction

With increasing global efforts in international development, foreign aid has evolved in its volume and design. Among them, technical assistance(TA) is receiving increasing attention as the partner country’s ownership and sustainability of aid have been emphasized. Technical assistance is commonly defined as the sharing of information, expertise, and other tactics involved in operating programs and achieving outcomes, but this definition takes transformative forms in various disciplines (West et al., 2012). It started as a simple transfer of ideas and practices for development, often referred to as a policy transfer (McMahon, 1997), but has transformed from a straightforward linear model to an adaptive and flexible one (Bhamidipati et al., 2019),^[1] incorporating recipients’ needs and making amendments as the project progresses. The definition of technical assistance has evolved consequently to encompass activities such as sharing of knowledge, providing training and technology necessary for solving development issues (Bazbauers, 2019). While it is difficult to reach a universally accepted definition, most of the literature defining technical assistance embraces the idea of capacity-building as the key element (Dunst et al., 2019; West et al., 2012).

The Republic of Korea’s Knowledge Sharing Program(KSP) is a technical assistance type aid program with increasing demand from many developing countries eager to learn from Korea’s rapid economic growth experience. Administered under the Ministry of Economy and Finance and managed by Korean Development Institute (KDI; a research institute affiliated to the Ministry of Economy and Finance), the KSP program is conducted in two types of cooperation: policy consultation (bilateral) and joint consulting with international organizations (multilateral).^[2] The KSP enables integrated policy research and consultation in a wide variety of socioeconomic development areas in partner countries, including macro-financial policies, science and technology, economic development planning, health and welfare, environment public administration, and more. Since 2004, over 626 KSP have been conducted to cooperate with 89 countries and 12 international organizations.

Contextual factors surrounding the process and outcome of policy advice have recently become a major research question related to the effectiveness of technical assistance. Unfortunately, much of the research so far is limited to theoretical discussion or case studies that are not enough to generate empirical evidence. One of the reasons for this limitation is the evaluation practice of these technical assistance type programs. The evaluation of technical assistance programs has been focused on the formative side: assessing whether the consulting output, usually in the form of workshops and documents, is delivered on time. The actual adoption and utilization of knowledge provided, which is the contribution of the program to generate developmental impact, is often left out of evaluation criteria and questions (Rondinelli, 1992). Such evaluation practices are not sufficient enough to generate lessons to improve the program nor to provide useful implications for the recipient, and recipient agencies have expressed dissatisfaction with the status quo (Bamberger, 1991). The Korean government’s KSPs have also been evaluated in a similar fashion, only assessing the consulting process and the delivery of consulting output (i.e., policy advisory report), while the actual adoption of policy advice or any relevant decision making after the project has been ignored (Choi & Kang, 2015).^[3]

First, the process of KSP output being reflected on the partner country’s policy formulation and decision making is usually complicated and time-consuming. Such process is also exposed to various contextual factors. The second reason is the limited availability of data. The final outcome of KSP can be identified as the actual adoption of the policy or institution advised by KSP. However, once the KSP is finalized and the final report is delivered, the data to assess the final outcome of the KSP is often not followed up by the partner country, and it is therefore difficult to collect them from the Korean government as well.

Under this backdrop, this paper aims to empirically evaluate the long-term outcome of KSP with the uniquely built database. After 2 or more years after each project, we have followed up on whether the policy recommendation provided by KSP has been actually reflected in the partner country’s policy for more than 300 projects carried out from 2004 to 2013. With this unique data appended to the basic information of each KSP and country-specific characteristics of the partner country, we aimed to analyze the outcome of KSP in relation to project and country-specific contextual factors. In terms of project-related factors, we separately identified characteristics of the policy being transferred and project management-related characteristics.

The empirical findings of this research have the potential to derive meaningful policy implications for technical assistance-type international development programs. At the same time, we aim to contribute to the previous studies on policy transfer theories. The article is constructed as follows: The next section identifies KSP as a policy transfer activity and explores theories of policy transfer and development, focusing on the contextual factors that affect the policy transfer outcome; Section 3 introduces major variables and methodology; Section 4 discusses the analysis result, and Section 5 concludes.

2. Understanding KSP as a Policy Transfer

1) Defining KSP as a Policy Transfer

According to the Development Assistance Committee of OECD, Official Development Assistance (ODA) is capital transfer towards developing countries or international organizations, which ranges from grant and non-grants to long-term credits (OECD, 1969).^[4] However, over the past years, international donors have started to pay increasing attention to the role of knowledge, not just to the provision of capital resources (King & McGrath, 2004). This new interest is aligned with the increasing ODA in the form of technical assistance and knowledge sharing, and the Korean Government has made a full-fledged effort to introduce the Knowledge Sharing Program (KSP).

Among the various types of technical assistance aid, KSP is carried out in the form of policy advice or policy consulting with the objective of local capacity and institution building. If the success of KSP can be identified as the actual adoption or localization of the policy suggested, then the policy advice to adoption process should be examined through the theoretical lens of policy transfer instead of using the traditional evaluation standard of ODA, which usually measures the social effect of the intervention.

According to Dolowitz and Marsh (2000), policy transfer can be understood as “emulation and lesson drawing all refer to a process in which knowledge about policies, administrative arrangements, institutions, etc., in one time and/or place is used in the development of policies, administrative arrangements, and institutions in another time and/or place”(344). The concept of policy transfer is also widely accepted as “the use of a policy in a selected country to be ‘copied’ or ‘imitated’ in another country, whether it is within or between different political contexts” (Mokhtar & Rahman, 2020, p. 8). Such a definition of policy transfer corresponds to the characteristics of KSP in providing in-depth consulting and training regarding a specific policy and institution to developing countries for localization. Indeed, KSP has been understood and analyzed as a means and process for policy transfer in various studies (e.g. Chung et al., 2011; Hwang & Song, 2019; W. Kim & H., 2013; Park & Ko, 2013; Park & Park, 2013).

Most of the previous literatures have located KSP in the middle of the spectrum of Dolowitz and Marsh (1996, 2000) between policy transfer and voluntary and coercive nature. Dolowitz and Marsh have placed policy transfer on a continuum from “voluntary (rational)” to “coercive” transfer (1996, 2000). The catalyst of voluntary transfer is some type of dissatisfaction with the status quo (Dolowitz & Marsh, 2000) or uncertainty (Haas, 1980, pp. 377–403), which leads to searching for lessons and solutions (Rose, 1993). On the other hand, coercive transfer happens when one government or a supra-national institution forces another government to adopt a particular policy (Dolowitz & Marsh, 1996, 2000). KSP, an international aid program, is often carried out at the request of the partner developing countries. However, a country can also be indirectly pushed towards policy transfer if political actors perceive their country as falling behind its neighbours or the emergence of an international consensus to adopt similar policies (Dolowitz & Marsh, 1996). In addition, capacity and institution building through knowledge sharing are often regarded as a conditionality for development, which means that it’s regarded as optional but as indispensable (Y. Kim & Tcha, 2012). In that sense, KSP can be understood as located in the middle of this spectrum (Hwang & Song, 2019; Park & Ko, 2013; Park & Park, 2013).

2) Considering the ‘Context’ for Successful Policy Transfer

The literature on policy literature, mainly after Dolowitz and Marsh, can be largely categorized into three strands: (1) framework building or theoretical discussion; (2) case studies of specific policy transfer; and (3) examination of the network or actor dynamics of policy transfer (Hwang & Song, 2019). Although the national context, including the political environment and institutional capacity, is recognized as a major factor that may encourage or constrain successful policy transfer (Benson & Jordan, 2011; Hwang & Song, 2019; Karini, 2013), the policy transfer analysis has been limited to the approach and type of transfer, discussing what is transferred and the degree of transfer (Mokhtar & Rahman, 2020).

Understanding the “national context” becomes even more crucial when policy transfer is attempted targeting developing countries because a lack of attention to the differences in policy conditions and structures between the donor and recipient countries can lead to an inappropriate transfer (Cairney et al., 2012). The ‘national context’ includes the political environment as well as the institutional context (i.e., administrative capacity, existing institutions, etc.) that can become “facilitators” or “constraints” in the process of policy transfer (Common, 1998; Karini, 2013). When policy transfer is attempted in the form of aid by donors or aid organizations, it is crucial to adequately examine the “specific institutional, bureaucratic, political, cultural, and economic conditions” of the recipient for a successful policy transfer (Karini, 2013, p. 2).

It is also important to understand the policy formulation and decision processes of the recipient countries for which the delivery of policy advice could actually be “transferred” and internalized. If KSP can be understood as a policy transfer, the output of KSP (i.e. the consulting process and the policy advisory report) has to influence the policy making process of the partner countries. Osman (2002) has pointed out that public policies in developing countries are peculiar in their characteristics due to the unstable political environment. Besides the country-specific context such as history, culture, and resources, developing countries share common features such as lack of pluralism, elites enjoying autonomous preferences, highly centralized decision making, and donor agencies as another dominant policy actor due to the scarcity of financial resources (Hai, 2008; Osman, 2002). The existing literature and case studies indicate that political and institutional context most crucially influence the actual uptake (or not) of research or policy advice (Court & Cotterrell, 2004; Young, 2004). Such an argument corresponds with Dolowitz and Marsh (2012)'s discussion of the relationship between governance and policy transfer. They suggested that comparing the process of policy transfer in different governance types(i.e., hierarchy, market, and networks) that involve different actors in different stages of policy transfer could be an interesting topic to be explored.

3) Project-Specific Characteristics affecting Policy Transfer

Attention should also be paid to how project-specific characteristics can affect the result of policy transfer. Project-specific characteristics can be separated into i) management characteristics of the project and ii) characteristics of the content of the project (i.e., policy being transferred). It can be intuitively assumed that the management or administrative characters of each project (i.e., the implementation agency, continuity of the project, etc.) can technically influence the result of policy transfer. For the content of the project, which is a specific “policy” or “institution”, public policy theories can provide useful insights to understand how the characteristics of a policy affect the policy transfer.

Theories of public policy formulation, and previous empirical studies of public policy identify influential factors that influence policy formation process such as the level of public interest, the expected level of conflict, the expected size of budget, and the need of legislation enactment. It has been argued that heightened public interest in specific issues results in a high level of support for adopting the policy being transferred (Swainson & Loe, 2011). Accordingly, if the interest of the general public in the policy is high (Chang, 2013), participation of non-regular participants in the policy formation process is encouraged, making the process more complicated and difficult to reach a consensus on (Gomley, 1986). Burstein’s (2003) meta-analysis of 30 studies revealed how the salience of policy magnified the relationship between public opinion and policy formation. The expected level of conflict induces similar effect on the policy formation process. The scope of impact of the policy recommendation, either targeted at a particular sector/group or the general public, could lead to varying degrees of conflict (Heritier, 1999). Moreover, the type of policy also plays an important determinant in predicting the level of conflict, as it differs depending on whether it serves a facilitating function or a corrective one (Lopez-Martinez, 2005) or whether it requires redistribution of resources (Wilson, 2013).

In discussing the re/distribution of resources, the budget size needs a consideration as it can cause distortion in other economic sectors (Timmer, 2004). According to the incrementalism theory by C. Lindblom (1959), policy decisions happen sequentially and incrementally, thus it is easier to adopt policies with a small change in budget than those requiring a radical budget change. The systematic change required due to the policy adoption also needs to be addressed, namely the enactment of law. Law and policy are closely related, where law functions as the purpose, means, product, or assumption of policy (Lim, 2014). Law also guarantees effectiveness of policy as well as limitation to it (Strempel, 1987). If a policy recommendation requires enactment of law, it may delay or impede the actual policy adoption as it requires additional consensus-making and the process of legislation.

3. Estimation Approach

1) Research Framework

Instead of drawing an individual hypothesis to test, we have developed a framework to explore the effects of diverse contextual factors that are assumed to have a relationship with policy transfer outcome. The framework is illustrated in Figure 1. We categorized the factors into three groups: country-specific factors; project-specific factors and policy-specific factors. The individual variables composing each group are drawn from previous literature reviewed in the second section.

Figure 1.Research Framework

Among the country-specific factors, we paid close attention to the level of democracy, while the rest of the country-level variables are controlled. As explained in the data section, the democracy index is comprised of components that address the political and institutional environment and capacity (EIU). Although the effect of political environment as well as the institutional context are regarded essential for policy transfer outcome (Common, 1998; Karini, 2013), previous studies have not provided enough empirical evidence regarding this matter.

Previously, the studies that evaluated policy transfer in an international development context have been limited to theoretical discussion (Dolowitz & Marsh, 2000; Hwang & Song, 2019) or individual case studies employing qualitative analysis (Cornell, 2013; Knack, 2001; Sam et al., 2017; J. Zhang et al., 2021). We aimed to contribute to these previous studies by generating empirical evidence regarding the effect of various contextual factors on successful policy transfer.

2) Analysis Model and Methodology: Bivariate Censored Probit Model

We found that simply applying a conventional Logit or Probit model can generate a problem. Unobserved characteristics that influence the probability of following up on KSP outcome may be correlated with unobserved characteristics that influence the probability of actual adoption/implementation of KSP output. At the same time, the successful adoption of a policy/institution recommended by KSP can only be observed in the case where the final outcome is followed up. There are two reasons why a conventional Logit or Probit will fail to produce an efficient and consistent estimator: i) it does not solve the potential correlation with the error term in the policy adoption probability equation; and ii) it estimates a probability of success of KSP, which is conditional on the probability of “follow-up”. The standard Heckman’s two-stage procedure is not applicable for this study as our model has a binary dependent variable (Boyes et al., 1989). An alternative methodology is the Bivariate Censored Probit model, which resolves both the selection problem and the correlation between error terms by allowing three different types of probabilities in the model.

The estimation equations for the probability of follow-up and the probability of KSP success can be specified as the following bivariate probit models:

$F^* = \beta_1 X_1 + \varepsilon_1$

F = 1 if $F^*$ > 0, 0 otherwise

(F* = continuous latent index that measures the probability of being followed up)

(F = bivariate variable for being followed up)

$S^* = \beta_2 X_2 + \varepsilon_2$

S(KSP Success) = 1 if $S^*$ > 0, 0 otherwise

(S*= continuous latent index that measures the probability of success of KSP)

(S = bivariate variable for the success of KSP)

$$\begin{aligned} &E\left(\varepsilon_1\right)=E\left(\varepsilon_2\right)=0 \\ &Var\left(\varepsilon_1\right)=Var\left(\varepsilon_2\right)=1, \\ &Cov\left(\varepsilon_1, \varepsilon_2\right)=\rho\\ &\left(\mathrm{S}, \quad X_2\right) \text{observed only when } \mathrm{F}=1 \end{aligned}$$

There are three cases in which ‘Follow-Up’ and ‘KSP Success’ are observable: (1) Follow Up = 0 (i.e., the outcome is not followed up); (2) Follow Up = 1 and KSP Success = 1 (i.e., the outcome if followed up and KSP output is adopted as a policy/institution); (3) Follow Up = 1 and KSP Success = 0 (i.e., followed-up and KSP is failed to be adopted as a policy/institution). Unconditional probabilities for each of the three cases can be estimated as follows (Greene, 2012, pp. 786–787):

$$\begin{aligned} &\operatorname{Pr}(FollowUp =0)=1-\Phi\left(\beta_1 X_1\right), \\ &\operatorname{Pr}(Success =0, Followup =1)=\Phi_2\left[-\beta_2 X_2, \beta_2 X_2,-\rho\right], \\ &\operatorname{Pr}(Success =1, Followup =1)=\Phi_2\left[\beta_2 X_2, \beta_3 X_3, \rho\right] \end{aligned}$$

, where the marginal probability for Follow Up is just $\Phi(\beta_1 X_1),$ whereas the conditional probability is $\Phi_2 (\cdots) / \Phi(\beta_1 X_1).$

Finally, the specified bivariate probit model for the success of KSP is the following:

$$\begin{aligned} &Prob \left[FollowUp = 1, KSP\ Success = 1 \mid \mathrm{X}_F, \mathrm{X}_S\right]\\ &\quad = \Phi_2\left(\mathrm{X}_F^{\prime} \beta_F, \mathrm{X}_s^{\prime} \beta_s, \rho\right) . \end{aligned}$$

As explained, the dependent variables in the model are

F = Whether the outcome of KSP is followed up

S = Adoption of KSP recommendation as a policy or institution

$X_F, X_S$ are composed with the vectors(z) of independent variables listed in Table 1.

3) Data

(1) Dependent Variable

The data was collected in 2015 during a separate research conducted by the authors with the Korean Development Institute (KDI). Most evaluation reports of KPS are not publicly available. For some evaluation reports that are publicly shared, there is a limitation in data as the outcome is often measured in output level (the delivery of report to partner countries). Due to the lack of access to the KSP information as well as the limited publicly shared information, we could not collect the outcome data of KSP after 2015. The dependent variable of this study is the success of KSPs, which is identified as the actual adoption of KSPs as a local policy or institution. Traditionally, the success of KSP has been evaluated based on the timely delivery of policy advice or consulting output (i.e., a report, workshop, or seminar), while the ultimate outcome of KSP, whether the KSP is actually adopted as a policy by the partner country or not, has never been evaluated. Following Dolowitz and Marsh (2000), defining the success of policy transfer as the extent to which a policy was adapted properly, we identified the actual adoption of the KSP as a local policy or institution as the dependent variable: “success of KSP.”

As this study attempts to evaluate the policy outcomes of KSPs from 2004~2012, we encountered a challenge in following up the policy outcome KSPs, which exposed the data to the possibility of censoring. The policy outcomes of KSPs are followed up both directly and indirectly: some of the projects are followed up by the KDI staff by contacting the partner country, and some are followed up by going through the press releases regarding the project outcome. Some projects were unable to be followed up if the counterpart staff/official/team was changed or if there was no relevant information available. The projects that are followed up but have not achieved the policy outcome are coded as 0, while the projects that are not followed up are also coded as 0. As the value 0 can have different meanings, the data inevitably has a potential censoring problem, and we have developed a model that embraces this problem.

The descriptive statistics of ‘Follow-up’ demonstrate 57.8% of the total sample (196 out of 339). Among the followed-up KSPs, the success rate was 35.7% (70/196)

(2) Independent Variables of Interest: Degree of Democracy

We used the Democracy Index published by the Economist Intelligence Unit(EIU) to measure the political environment. The index ranges from 0 (Authoritarian) to 10 (Full Democracy) as a continuous variable, where it can also be categorized into 4 levels: Full democracies (score 8.01 to 10), Flawed Democracies (6.01 to 8), Hybrid Regimes (4.01 to 6), and Authoritarian Regimes (0 to 4).^[5] In our estimation, continuous democracy variable is employed in Model (1) and (2), while categorical variables are employed in Model (3) and (4). The descriptive analysis of the democracy index is provided in Table 2. As illustrated, the partner countries for KSPs are mostly developing countries, and there was no country with full democracy.

(3) Country-level Control Variables

Other country-level variables are included in the model to control their influence, following similar quantitative analysis using macro-level variables (Benmamoun & Lehnert, 2013) as well as literature on international development (e.g., Burnside & Dollar, 2000; Easterly et al., 2004; Svensson, 2000). Since the level of corruption might affect the policy making process, the Corruption Perception Index (CPI) is considered in the analysis. As the economic status of the recipient country is highly associated with its responsiveness to international assistance, GDP and GDP per capita are included in the model. We included both GDP and GDP per capita as GDP assesses the size of country’s economy while GDP per capita controls for distribution and allocations (Dissanayake & Tahmasebi, 2021). As demand for the KSP program as well as the response to KSP results could be varied depending upon the level of ODA or policy fulfilled by other donors, the level of ODA is also controlled to capture ODA competitiveness, using total ODA volume and ODA per capita. ODA implies total inflow into the country while per capita controls for possible bias rising from country size and aid fragmentation (E. M. Kim & Oh, 2012).

(4) Project-level Policy Factors

Besides the country’s characteristics, the research included four factors as policy characteristics that would potentially influence policy transfer based on the literature review: (1) expected level of public interest, (2) expected level of conflict, (3) expected volume of budget, and (4) need of enactment of legislation. Each factor is measured based on the heuristics by the researchers based on the information provided by each KSP final report.

To measure the public interest, Gormley’s (1986) concept of “salience”, the factor that motivates non-traditional players to participate in policy making, is utilized. Applying the heuristics reflecting Gormley’s theory, the predicted level of public interest of the KSP policy recommendation is identified in the three-level scale: ‘3-high(interested),’ ‘2-intermediate(slightly interested),’ and ‘1-low(indifferent).’^[6]

Based on Lopez-Martinez (2005) and Wilson (2013), a four-level decision tree (heuristics) was constructed for the level of conflict. With each decision node representing distinct levels of conflict-level choices, the heuristics lead to a three-level scale of conflict: “3-high conflict”, “2-intermediate conflict”, and “1-low conflict”. The following four standards are included in the heuristics for the measurement: “presence of policy advice”, “the scope of advice supplied at macro-level or sector-specific”, “the extent of conflict based on the functional role of KSP policy advice”, and “a need for resource allocation”.^[7]

The heuristics to measure the expected level of budget are constructed based on the review of KSP final reports. We classified each project as a three-level scale of “3-high (large scale infrastructure projects/policies)”, “2-intermediate (small scale infrastructure projects/policies)”, and “1-low (support to research and/or model development)”. Lastly, the need of legislation enactment is determined according to the policy recommendation stated in each KSP final report.

(5) Project-level Management Variables

As mentioned, certain characteristics or features of KSP projects can also influence the outcome of the project, therefore project-specific variables are controlled in the analysis. Firstly, whether the project is directly managed or outsourced is included as a control. Another project-specific variable considered is the continuity of the project, as some KSPs are carried out consecutively one after another with the same counterpart. Lastly, whether or not the recipient countries are assigned as priority partners according to the 5-year ‘Strategic Plan of International Development Cooperation’ of Korea is included as a control.^[8][9]

4. Results and Discussion

1) The Effects of Country-Specific Contextual Factors

(1) Effect of Democracy

The degree of democracy, which represents the political environment of the KSP recipient, turns out to have a versatile effect on the probability of KSP adoption while exerting a negative effect on the follow-up activity. When included as a continuous variable, the effect of democracy score on FollowUp is negative and statistically significant, but the effect on KSP adoption is not statistically significant (Model (1)) when other country-specific and project-level variables are controlled. The model with a categorical democracy variable provides an explanation for this statistical weakness, as the possibility of follow-up the KSP outcome decreases statistically significantly in the ‘Hybrid Regime’ compared to the ‘Authoritarian’ regime (Model (2)),^[10] which is the reference group . The possibility of KSP success statistically significantly decreases in the ‘flawed democracy’ (democracy score 4.1-6) compared to the ‘Authoritarian.’ This result implies that an ‘Authoritarian regime’ has a relative advantage in following up the KSP results as well as in the probability of actual adoption of KSP.

Modernization theory provides a useful interpretation of this result related to the effect of the political environment on the adoption of KSP, which can be regarded as a developmental policy transfer. According to the modernization theory, economic development and the probability of democracy have an “S-shaped” nonlinear relationship, as illustrated in Figure 2. Explaining the positive relationship between democracy and economic development, the theory also implies how economic development can happen under the limited level of democracy (Geddes, 1999), such as in the case where an authoritarian or hybrid regime pursues centrally driven top-down development. It is indeed argued that strong leadership that promotes property rights is the most suitable for rapid economic development because the leaders will take the initiative to amend what belongs to them (Charron & Lapuente, 2011; Olson, 1993). Likewise, a single-party government can be more responsive to citizens because it has shortened the mechanism for directing the voices of citizens straight to the party, thus shortening the process of reaching consensus (Charron & Lapuente, 2011; Irvin & Stansbury, 2004). In terms of developing countries, where the administrative function is not sufficiently developed to manage diverse stakeholders and perspectives fairly and justly, the democratic political decision-making process can be inefficient to promote development.

Figure 2.Modernization Theory

Note. From Geddes, B. 1999. What do we know about democratization after twenty years?. Annual review of political science, 2(1): 115-144.

There are a number of studies implicating the feasibility of a developmental policy transfer under the less democratic policy decision-making process (of the recipient country) represented by a centrally controlled decision-making process. Zhang and Yu (2021) describes how the policy transfer with the China-Europe Public Administration(CEPA) program was possible because the Chinese counterparts were composed of the Chinese bureaucratic elite, who were interested in administrative reform and had enough influential power in the policy decision-making process (Y. Zhang & Yu, 2021). In their case study of a policy transfer program carried out in Nigeria, Sam et al. (2017) also highlights that the strong political will of major political actors or a group is crucial for successful policy transfer, while the existence of conflicting political goals or ideals between politicians and citizens can constrain successful policy transfer. Foli et al. (2018) explicates how instrument constituencies, the collective policy actors who are bound by an interest in a particular policy instrument or solution, dominate the decision-making process of policy adoption in developing countries. They found that instrument constituencies promote particular policy instruments while legitimizing their jurisdictions. Similarly, Bhamidipati et al. (2019) addressed how actor constellations with political influence, regarding the renewable energy policy transfer program, could accelerate the policy transfer within the specific political and institutional context in Uganda.

(2) Country-Level Control Variables

Some of the country-level control variables exhibited statistically significant effects on both the follow-up and the final outcome of KSP. The level of GDP per capita and ODA per capita had a negative effect on the possibility of follow-up and actual adoption of policy. GDP per capita may represent the economic condition of the country, while the ODA per capital indirectly reflects the effect of other donors and ODA (Acharya et al., 2006; Kimura et al., 2012). It can be interpreted that countries with more developmental needs (with a low level of economy) and less donor competition^[11] (and therefore less possibility of overlapped policy advice) can lead to better follow-up and a better outcome of KSP. The possibility of better aid effectiveness with less donor competition has been empirically supported from previous literature (Acharya et al., 2006; Kimura et al., 2012), which is once again proved from the analysis.

2) Effect of Policy-Specific Variables

The analyses convey the decreased possibility of follow-up when the expected level of public interest gets higher, and the negative effect of expected policy budget size on the final adoption of KSP when other contextual factors are controlled. The difficulty of the adoption of an “expensive” policy or institution demonstrated in the analyses can be intuitively comprehended, verifying the previous discussion made by C. Lindblom’s incrementalism theory. Also aligned is the theory of gradual institutional change by Mahoney and Thelen (2010) where institutional change happens in small and gradual ‘layers’, making amendments and additions to old institutions rather than abrupt and substantial change.

The negative impact of high public interest on the follow-up of KSP can be interpreted in two ways. One explanation is the possible delay in policy discussion and formulation suspending the follow-up due to the further needs to improve the policy or to coordinate different opinion regarding the policy drawn from the heightened attention from the public (Birkland, 2001; Kingdon, 1995). The public can also easily become interested in highly complex policies, in which case, high salience does not necessarily lead to an intervention or policy without clear expression of the public for a certain policy area (Spence, 1997). Another possible explanation is that the policy process is intentionally undisclosed by the policymakers and administrative stakeholders, especially when the policy has public salience. According to Gormley’s Salience-Complexity Model of Regulatory System (1986), the location for policy discussion and decision for publicly salient policies are either ‘Hearing Room’ or ‘Operating Room,’ where only selective stakeholders (i.e. elected officials, upper level bureaucrats and, technical experts and professionals) participate. Making a national decision with high public salience involves technical difficulties and high level of responsibility, and thus policy making authorities can be reluctant to share the discretion (Neshkova, 2014). For KSPs that involve potential salience, it is possible that the recipient decision-making process is concealed and it’s difficult to follow up.

3) Effects of Project-Specific Variables

At the same time, projects that are directly managed by KDI (not outsourced) tend to have a lower possibility of being followed up, but have a higher possibility of actual adoption of the advised policy or institution. Perhaps the projects directly monitored and managed have less need to be followed up on as the possibility of actual policy adoption could have been anticipated during the process of KSP. With the same logic, the outsourced KSPs might become the target to be followed up on as the process of KSP is not directly monitored and the final outcome of KSP is more difficult to conjecture. Moreover, the outsourced projects often involve external experts with a management that better understand and track information that is not redundant with the organization’s existing management knowledge (Espedal, 2005). From this perspective, the possibility of being followed up on KSPs that are internally managed might be lower than the outsourced KSP.

For the continued KSPs and KSPs conducted in prioritized partners, the results exhibit the higher possibility for the follow-up. It can be understood that the close relationship built with the partner country could make tracking the result more feasible. It was demonstrated from individual cases where continued KSPs had taken place in priority countries. For instance, the Partnership for Skills in Applied Sciences, Engineering and Technology (PASET) was launched in 2019 and 2020, continuing from the successful completion of the KSP from 2016 to 2018 in Rwanda and Ethiopia.^[12] Setting up the coordination and collaboration mechanisms through regular meetings and workshops based on the previous relationship already established, the South Korean government was able to have easier access to follow-up the project results (Korea Development Institute, 2020).

Projects in priority countries, however, do not seem to guarantee the final adoption of KSP. One of the reasons can be found in the possible political instability and low institutional quality to internalize the policy recommended. Political stability and institutional quality are likely to play a vital role in achieving the intended objectives of foreign aid (Hongli & Vitenu-Sackey, 2020; Islam, 2005). We aimed to control the political environment and economic conditions in the model, but they are not enough to control the country-level instability, whereas the institutional quality is partially controlled through the democracy index. As primary partners tend to exhibit political instability (Sohn et al., 2011), prioritized partner countries may struggle to adopt KSPs under their political circumstances.

5. Implications and Conclusion

The KSP is a representative development cooperation project of Korea that was initiated to meet the learning demand of developing countries for the rapid economic development experience of Korea, called the “Miracle of the Han River.” Unlike other bilateral aid projects, KSP aims to help the recipient country grow on its own, and the ultimate purpose is to have the advice provided by KSP be reflected in the recipient country’s policy decisions. There is an increasing demand for KSP as well as similar policy consulting and capacity-building types of ODA from partner countries. Against this backdrop, this study aims to contribute to previous literature exploring the effect of various contextual factors on successful policy transfer. Using more than 300 KSP data, we attempted to provide empirical evidence on the effect of the country, policy and project level factors on the actual adoption of KSP.

This study categorized country-specific factors, policy-specific factors, and project management factors to study the relevant contextual factors that influenced the process of knowledge transfer and adoption. Among various factors, the type of governance, which is measured with the democracy index, is discussed in-depth as the study discovered how the possibility of policy adoption in a country increases as the degree of democracy increases until the country becomes a “Hybrid Regime”, but it decreases when a country achieves “Flawed Democracy”. The statistical insignificance of the effect of the aggregated democracy score seems to be attributed to the offsetting effects exerted by the different degrees of the democracy score. The KSPs are largely developed based on Korea’s economic development experience under the benevolent dictatorship of Park, which manifested a hybrid regime type of governance. On that note, the results from this study might imply that the KSPs are perceived as more attractive and therefore actively adopted by political leaders of countries that share similar political environments to Korea back in the `70s. Besides the type of governance, the overall economic condition, the volume of ODA, the expected level of public interest, as well as the budget size of the advised policy turned out to be influential factors that can affect the success of policy transfer.

Besides the academic implications, this study has the following implications for knowledge sharing or technical assistance projects such as KSP. First, the political system of the recipient country needs to be considered when selecting a partner country for a policy transfer project. Most of the countries that demand knowledge sharing may not have achieved full democracy yet, but are in the process of democratization. Therefore, the type of political environment should be considered with more attention when selecting partner countries, not treating democracy as a dichotomous variable. For instance, KSP designed to strengthen the democratization process can be suggested to the countries that are assessed to be in the process of democratization. It would be effective if specific policy recommendation could be provided targeting the specific element of democracy that is evaluated to be lacking, such as government administrative function or civil society. Second, the contents of policy advice should include plans and alternatives that are realistically consistent with the political system of the recipient country. More specifically, the type and characteristics of the policy or institution to be recommended with KSP (i.e., the expected level of public interest, conflict, and budget) should be carefully reviewed and customized considering the economic and political environment of the recipient. Third, extra attention is needed for KSPs that handle salient and high-budget policies and institutions. For highly salient policies, additional policy alternatives or strategies to handle the public attention and disputes should be provided together. In the meantime, KSPs that recommend costly institution building could be linked with additional ODA projects. Forth, as the continued projects do not guarantee the final adoption of KSPs, the implementing agency should not be satisfied with the “continuity” itself but endeavor to achieve the final outcome. The same is needed for the projects in the prioritized countries. Lastly, the evaluation of the policy transfer project should be conducted with expanded criteria to assess the policy reflection or adoption beyond the successful delivery of policy advice.

This study, however, has limitations that can be further addressed in future studies. First of all, the estimation result is exposed to the possible omitted variable bias due to the factors that are failed to be addressed in the model. We have tried to address country-level political and economic variables, as well as project-specific factors that are related to both the “policy” being transferred and project management. Nevertheless, the outcome of policy transfer is known to be affected by other factors as well. For example, quality of policy advice (e.g. Dolowitz & Marsh, 2000; Hwang & Song, 2019), policy actors and governance system (e.g. Cornell, 2013; Dolowitz & Marsh, 2000; Gibson et al., 2015; Hwang & Song, 2019; Knack, 2001; Sam et al., 2017), economic factors (e.g. Gibson et al., 2015; Sam et al., 2017), institutional constituency (e.g. Bhamidipati et al., 2019), the relationship between the policy transfer provider and the recipient (e.g. Bazbauers, 2017; Katz & Wandersman, 2016), social characteristics (e.g., Kwon & Kim, 2015; Nutley et al., 2012; Tavits, 2003) and cultural factors (e.g., Ellison, 2017; Lenschow et al., 2005; J. Zhang et al., 2021) are the main factors that have been identified and studied in previous studies. Most of the studies above are qualitatively carried out for a single case, so some of the factors are not feasible to be quantitatively transformed or applied to the analysis of multiple cases, such as this study. Moreover, in the process of developing a quantitative variable for “policy factors” and “project factors”, there is a possibility of exposure to evaluator’s bias. We employed the heuristics that are established based on previous literature to assess the policy and project factors in order to minimize the evaluator’s bias. However, the assessment was based on the project documents (including midterm and final evaluation reports) that are also exposed to evaluator’s bias.

Acknowledging these limitations, the results of this study call for further investigation into the contextual factors related to policy transfer. Similar factor analysis with diverse categorization of policies (i.e. promotive or regulative policy, policy by sector, etc.) or categorization of countries (i.e. by economic-level or governance-level) may lead to additional meaningful insights.