Abstract
When individual actors are involved in a policy process, do they assess and revise their policy preferences according to their interests or are they open to other forms of arguments over time? This study examines the effect of policy actors' interests on policy learning. It is based on a survey conducted in 2012 among 376 Belgian actors (from 38 organizations) involved in the European liberalization policy process of two network industries: the rail and electricity sectors. Borrowing from organizational research and behavioral economics, several hypotheses are drawn from a model of the individual shared by various policy approaches, such as the advocacy coalition framework. A “simple gain scores” approach to the measurement of policy learning is introduced. Regression analyses show that policy actors align their policy preferences with the impacts of policies on their own material well-being (personal interests) and the material prosperity of their organization (organizational interests). This tendency is independent of the importance that policy actors give to their interests in their everyday lives. This suggests that policy actors experience a sort of “interest shift” when they assess their policy preferences over time. This shift, however, exerts a limited influence on policy learning. The theoretical and practical implications are discussed.
Keywords
Introduction
Policy processes involve diverse types of policy actors, ranging from politicians and public officials to company and association managers. As a result of various interactions as well as the gradual accumulation of evidence on policy problems over time, those policy actors acquire, translate, and disseminate new information and knowledge (Heikkila and Gerlak, 2013). In turn, they maintain, strengthen, or revise their beliefs and preferences regarding policies. “Policy learning” is a concept that designates this cognitive and social dynamic.
Human learning is a fundamental intermediate factor of change processes. Change requires actors to create or to deal with new information and new experiences. This results in the enduring acquisition or modification of abstract constructs (Vandenbos, 2007). Those alterations, in turn, transform actors' behavioral intentions and their contribution to change (Fishbein and Ajzen, 2010). Hence, policy learning is a causal mechanism linking over time the beliefs held by individual policy actors, the revision of those beliefs, the alterations of collective ideas, and policy change.
The present study focuses on the psychological conditions of individual learning within one policy domain. In particular, it examines whether individual policy actors align, over time, their policy preferences with the effect of policy programs on their own interests. I distinguish two categories of interests: the personal interests of individual policy actors refer to their own material well-being whereas their organizational interests refer to the material prosperity of the organization in which they work.
A better theorization of the psychological factors of policy learning is crucial because it helps to understand its relation with policy change. For example, there are indications that policy actors who acquire more knowledge about a policy problem are more open to considering new, different policy solutions to solving this problem (Leach et al., 2014). Individual policy learning also has other possible corollary outcomes, such as simplifying the emergence of shared interpretations of policy issues or facilitating agreements among policy actors (Brummel et al., 2010; Diduck et al., 2012; Leach et al., 2014). Finally, learning has a “strategic character”, as it can help “policy participants to promote their respective policy interests and values” (Real-Dato, 2009: 129).
To examine the effect of interests in individual policy learning, I rely on the Advocacy Coalition Framework (ACF: Sabatier and Jenkins-Smith, 1993). This approach is most often used to examine the role that coalitions of policy actors struggling with each other play in policy processes. However, the ACF also recognizes that policy change depends on policy learning. To scrutinize the role of cognition in policy processes, the ACF relies on a model of the individual characterized by a bounded rationality and two logics of normative reasoning: a logic of consequences and a logic of appropriateness (Sabatier and Weible, 2007).
There are at least two reasons to rely on the ACF in this study. First, the effect of interests on policy learning and policy change has been a long-standing but unresolved debate, in the ACF literature (e.g., Elliott and Schlaepfer, 2001; Hsu, 2005; Kübler, 2001; Leach and Sabatier, 2005; Nohrstedt, 2005; Sabatier et al., 1987). Second, many policy approaches share with the ACF similar assumptions on individual actors (Dunlop and Radaelli, 2013; Zito and Schout, 2009). However, the ACF assumptions are particularly detailed. Hence, the implications of those assumptions are clearly testable. This makes the ACF a very good choice to examine the effect of policy actors' interests on the revision of their policy preferences over time.
The hypotheses of this study are not only based on the ACF. They also come from research in political psychology on the role of interests in citizens' social and political attitudes (e.g., Lau and Heldman, 2009; Margalit, 2013; Martinussen, 2008; Sears and Funk, 1991), in organizational psychology on bounded rationality (Greve, 2011; Robbins et al., 2013; Simon, 1991), and in behavioral economics on prospect theory (Kahneman and Tversky, 1979; for a review, see Barberis, 2012). This research helps drawing the implications of the ACF model of the individual.
The test of the hypotheses is based on regression analyses of a survey conducted in 2012 among 376 Belgian policy actors who had been involved, during the last two decades, in the European liberalization policy process of two network industries: the rail and electricity sectors. The European Social Survey suggests that Belgian policy actors are representative of many of their European colleagues regarding the importance they give to their interests in their everyday lives. 1 The liberalization process of network industries offers a good balance between policy changes concerning how to provide products/services to citizens and changes in the interests of policy actors that result from the re-organization of the industries. As the liberalization of network industries began in the 1990s, this research is also consistent with Sabatier's (1993) contention that policy processes should be considered “over a decade or more” to capture the real nature of policy learning.
This study contributes to the policy science in various ways. First of all, much policy learning research (e.g., Jacobs and Barnett, 2000; Meseguer, 2004; Pemberton, 2003) looks at the entire mechanism linking human cognition and policy learning to policy change. In contrast, this study clarifies the effect of human cognition on policy learning itself. In addition, this study introduces an innovative, “simple gain scores” approach to the measurement of policy learning.
Second, many studies account for policy processes by looking at interests together with beliefs or ideas (e.g., Dehnhardt, 2014). Following calls for research focusing on the exact influence of interests (e.g., Hoberg, 1996; Nohrstedt, 2005), I try to reject the null hypothesis that interests do not exert any influence on policy learning. In addition, most researchers oppose interests to beliefs and ideas. In this study, I recognize the ideational character of interests by using self-reported measures of interests. This is far more reliable than proxy measures of interests deduced by the researcher on the basis of policy actors' characteristics or affiliations (Darke and Chaiken, 2005).
Third, the role of actors' interests in policy processes has been a perennial issue in the policy science (Béland, 2009; Genieys and Smyrl, 2008). In addition, the ACF shares its model of the individual with various other approaches to the policy process, as well as many public administration and public management studies. Hence, the findings of this study talk to diverse researchers in public policy and administration, within and beyond the ACF community.
This article should be understood as part of the efforts to develop a behavioral approach to the study of public policy (Shafir, 2013). The findings of this article will shed light on the psychological conditions of individual-level policy learning. In terms of significance, this means that this article does not make any assumption about the ultimate influence of policy actors' interests on group-level policy learning or on collective action. The actual impact of individual learning on collective processes depends on social interactions, organizational practices, and institutional contexts in which it takes place (e.g., Witting and Moyson, 2015). In the same perspective, there are not only psychological conditions to individual-level policy learning, but also social, organizational, and institutional factors (Dunlop and Radaelli, 2017). However, research efforts to clarify the behavioral foundations of public policy are crucial because decision-making processes based on bad intuitive psychology are less likely to succeed (Shafir, 2013).
This article follows a classical structure in which the theoretical expectations are presented before the research design, the measures, and the analysis. Finally, the findings are discussed.
The role of interests in long-term policy learning
Policy learning and policy change in the advocacy coalition framework
The ACF (Sabatier, 1987; Sabatier and Jenkins-Smith, 1993, 1999; Sabatier and Weible, 2007) conceptualizes the policy process as a political struggle among (coalitions of) policy actors involved in a given policy subsystem. A policy subsystem is a set of “actors from various public and private organizations who are actively concerned with a policy problem or issue such as air pollution control, and who regularly seek to influence public policy in that domain” (Sabatier and Jenkins-Smith, 1999: 119).
The ACF assumes that each policy actor holds a belief system composed of three strata. At the first stratum, “deep core” beliefs are very broad in scope (e.g., “I believe that justice is an important value”). At the second stratum, “policy core” beliefs are specific to one subsystem (e.g., “I believe that this policy option increases the degree of justice among groups of the population”) and determine actors' preferences regarding key policies (e.g., “I believe that this policy option is better than others”). At the third stratum, “secondary” beliefs concern particular issues and decisions within the subsystem (e.g., “I believe that this administrative decision facilitates the implementation of my preferred policy option”).
One important objective of the ACF is to explain policy change, defined as “fluctuations in the dominant belief systems (i.e., those incorporated into public policy)” (Sabatier, 1987: 682). For example, policy change can result from major “shifts in the core attributes of the subsystem”, called “shocks”, such as a legal shock or a shock in the distribution of natural resources (Weible et al., 2009: 124). Policy actors can also use their resources and coordinate their political activity within “advocacy coalitions” to impose an understanding of policy problems and their preferred definition of policy solutions to other advocacy coalitions (Sabatier and Jenkins-Smith, 1993).
However, policy change does not only depend on shocks and power relations but also on the beliefs and preferences held by policy actors. The revision of those preferences is a causal mechanism called “policy learning”, which determines policy decisions. The ACF defines policy learning as “relatively enduring alterations of thought or behavioural intentions that result from experience and which are concerned with the attainment or revision of the precepts of the belief system of individuals or of collectivities” (Sabatier, 1993: 42).
The belief system of a policy actor (or any individual) is composed of predictive beliefs and evaluative beliefs (Fishbein and Ajzen, 2010). In the context of policy processes, predictive beliefs concern the expected outcomes of policy programs (e.g., ‘the liberalization policy will allow new companies to operate trains on the Belgian railways’). In common language as well as in psychological and educational research, the term ‘learning’ most often refers to a process that leads an individual to revise his or her predictive beliefs as a result of knowledge acquisition or social interactions. In the ACF, however, “policy learning” is a broader concept: it equally denotes the cognitive and social processes that lead to changes in the evaluative beliefs of policy actors. Evaluative beliefs concern the desirability of the outcomes of policy processes (e.g., “it would be desirable that new companies are allowed to operate trains”) and, ultimately, the desirability of policies themselves (e.g., “the liberalization policy would be desirable”). The latter are called “policy preferences”.
Some ACF research has looked at the conditions of cross-coalition policy learning or the role of policy learning in policy change (Weible et al., 2009). However, the psychology of individual learning has been rather overlooked. Here, I focus on one result of individual learning: the stability or evolution of policy actors' preferences over time. This is the dependent variable of this study. In this respect, the study adopts a centered perspective on policy actors' evaluative beliefs. But policy preferences have been recognized as “normative beliefs that project an image of how the policy subsystem ought to be, provide the vision that guides coalition strategic behavior, and helps unite allies and divide opponents” (Sabatier and Weible, 2007: 195). As policy preferences play a fundamental role in collective action, it is crucial to understand the psychological foundations of their evolution over time.
The nature of interests
Following a consistent body of research in political psychology on the role of interests in citizens' social and political attitudes (e.g., Lau and Heldman, 2009; Margalit, 2013; Martinussen, 2008), the interests of a given entity—for example, an individual or an organization—may be defined in reference to the material well-being or prosperity of this entity. Restricting interests to material aspects allows for more clarity regarding other concepts such as values (Sears and Funk, 1991).
It has been suggested that interests should be analyzed in opposition to beliefs in order to understand their respective role in policy processes (Hoberg, 1996: 143; Nohrstedt, 2005: 1045–1046). In contrast, I follow Sabatier and Jenkins-Smith's (1993: 28) original argument that interests are subjective beliefs among others: “while belief system models can (…) incorporate self-interest and organizational interests, they also allow actors to establish goals in quite different ways (e.g., as a result of socialization)”. Similarly, Laird (1999) argued that “actors use ideas in order to understand what their interests are, what those interests mean to them, and how to achieve them” (cited by Dudley, 2007: 411).
Yet, if interests are beliefs, how are they to be distinguished from other beliefs? Two different concepts may be used: “policy interests” and the “generalized importance of interests”. On the one hand, policy interests are beliefs about the impacts of a policy on someone's material well-being (e.g., “I believe that the implementation of this policy has improved my salary”). On the other hand, the generalized “importance” or “salience” of interests indicates how much they motivate someone's attitudes and preferences. Policy interests are policy core beliefs—they are specific to the policies of a given policy subsystem—whereas the generalized importance of interests results from a set of deep core beliefs—they are general in scope.
Finally, policy interests and the generalized importance of interests can concern personal interests or organizational interests. Personal policy interests of individual policy actors are their beliefs about the impacts of a policy on their own material well-being (e.g., “I believe that this policy will increase my personal working conditions”). Organizational policy interests are their beliefs about the impacts of a policy on the material prosperity of the organization which they are affiliated to (e.g., “I believe that this policy will improve the financial situation of my organization”). Similarly, I differentiate the generalized importance given by policy actors to their personal and organizational interests.
Most studies looking at the influence of interests on social and political attitudes focus on personal interests (e.g., Lau and Heldman, 2009). Sometimes, typically in studies assimilating parties' electoral performance to interests (e.g., Gilardi, 2010), personal and organizational interests are measured together. Indeed, electoral successes serve the interests of individuals as well as the interests of their party. However, as the theoretical and practical implications resulting from the effect of personal and organizational interests might not be similar, I define them differently and measured them with different questions, in the survey of this research.
Theoretical expectations on the relation between interests and policy learning
The ACF model of the individual is based on two important assumptions that lead to contradictory expectations regarding the role of interests in policy learning. The first assumption is that the belief system of policy actors is internally coherent (Sabatier and Jenkins-Smith, 1993). In particular, policy actors develop over time a specialized knowledge which allows them to formulate policy preferences that are consistent with their beliefs. This suggests that they are naturally able to opt for policies that fit their interests (e.g., “I believe that my personal interests are important and I believe that this policy improves my material well-being, so I believe that it is a desirable policy”).
However, the second important assumption of the ACF is recognizing that individuals' rationality is “bounded” (Sabatier and Weible, 2007). This assumption results from a “behavioralist turn” (Zito and Schout, 2009) adopted by the ACF in parallel with many other approaches to the policy process (Dunlop and Radaelli, 2013), and borrowed from organizational research (Simon, 1991; see also Greve, 2011; Robbins et al., 2013) as well as behavioral economics (prospect theory: Kahneman and Tversky, 1979; see also Kahneman, 2011; for a review, see Barberis, 2012). Bounded rationality suggests that policy actors' ability to formulate policy preferences that are consistent with their beliefs is limited. Indeed, the information available about policies can be of poor quality or low quantity. In addition, the inherent ability of individuals to process this information is limited, especially in contexts of uncertainty or urgency (Birkland, 2006; Moynihan, 2008; Svenson and Maule, 1993).
When information, time or ability to process information is limited, prospect theory (Kahneman and Tversky, 1979) suggests that individuals tend to rely on “shortcuts” or heuristic-based modes of reasoning. For example, due to a so-called “devil shift”, policy actors tend to exaggerate the political risk represented by other actors having different policy preferences: they see those actors as more evil and powerful than they are in reality (Leach and Sabatier, 2005; Sabatier et al., 1987). The devil shift results from cognitive biases such as risk aversion or loss aversion, meaning that individuals dislike risks and losses more than they like certitudes and gains (Kahneman and Tversky, 1979). Similarly, as a result of the “certainty effect” (Kahneman, 2011; McGraw et al., 2010), policy actors generally prefer to rely on their pre-existing normative and perceptual beliefs rather than admitting that they are wrong (Leach et al., 2014; Sabatier and Jenkins-Smith, 1999). All in all, bounded rationality sheds skepticism on policy actors' propensity to changing their policy preferences over time and to revising them in a way that fits their policy beliefs. Hence, the null hypothesis of this research states that policy actors do not significantly tend to align their policy preferences with their interests over time.
However, despite the lack of extensive empirical evidence, there are several reasons to think that policy actors are rational enough to maintain consistency between their policy preferences and their policy interests over time. First, some ACF research suggests that interests are a tangible driver of policy processes, with an impact on the formation and behavior of advocacy coalitions (e.g., Elliott and Schlaepfer, 2001; Hsu, 2005; Kübler, 2001) as well as policy change (Hoberg, 1996; Nohrstedt 2005). However, in those studies, the concept of “interest” covers very diverse empirical realities. In addition, the ideational character of interests is not always really well recognized in their definition and measurement.
Second, the role of interests in the formation of citizens' social and political attitudes was demonstrated by a consistent body of research in political psychology (e.g, Lau and Heldman, 2009; Margalit, 2013; Martinussen, 2008; Sears and Funk, 1991). However, those studies are inconclusive about the strength of the interests–attitudes connection.
Third, as explained above, policy actors' ability to opt for policy decisions that are consistent with their interests is limited in contexts characterized by uncertainty or urgency (Birkland, 2006; Moynihan, 2008). However, the members of a policy subsystem are, for the most part, experienced policy “elites”. For this reason, “there are strong grounds for assuming that most actors will have relatively complex and internally consistent belief systems” (Sabatier, 1993: 181). In addition, the present study focuses on long-term policy learning. In the long run, policy actors should have more opportunities to structure and organize their policy knowledge. This should allow them to align their policy preferences with their interests, which leads me to the first alternative hypothesis of this research: in the long run, policy actors tend to align their policy preferences with their policy interests.
Furthermore, the ACF borrows from organizational research (March and Olsen, 2008) that policy actors’ attitudes are not only driven by interests but also by a logic of appropriateness (Sabatier and Weible, 2007). This means that the course of policy actors’ preferences is determined, not only by interests, but also by various other internalized values, norms, and rules. In other words, policy actors can differ from each other with respect to the generalized importance they give to their interests. If policy actors are internally coherent, the more they give importance to their interests in their everyday lives, the more they should align their policy preferences with their policy interests. This leads me to the second alternative hypothesis: the relation between the revision of policy actors’ preferences and their policy interests should be stronger when policy actors give more importance to their interests in their everyday lives. In statistical terms, this means that the importance that policy actors give to their interests would ‘moderate’ the relation between policy interests and policy preferences over time.
Alternative hypotheses.

Analytical framework and alternative hypotheses of the study.
Research design
To examine the role played by interests in policy learning, a web survey was submitted to the Belgian policy actors involved in the implementation of the European liberalization policy process within two network industries: the rail and electricity sectors. Network industries “are characterized by the delivery of products or services to final customers via a ‘network infrastructure’ linking upstream supply with downstream customers” (European Commission, 1999). Network industries are typical in sectors like telecommunications, energy, transport, or postal services.
Since the 1980s, many network industries have been subject to a liberalization policy process (Genoud, 2004; Geradin, 2006). The implementation of this European process in national policy subsystems may be considered a major policy change. Gradually, the network activities have been unbundled. Previously, a state-owned company (or “incumbent”) had a monopoly on the management and commercial exploitation of the network. Nowadays, a public “infrastructure manager” is responsible for the maintenance and security of the infrastructure, whereas the incumbent competes with other private companies (or “new entrants”) for use of the infrastructure. In addition, various independent regulatory agencies have been created at European and national levels.
This study focuses on two national subsystems of policy actors: the Belgian rail and electricity policy subsystems. In the railways, the European liberalization process began in 1991, with European directive 91/440/EEC. The implementation of this process, in Belgium, began with the Royal decree of 5 February 1997 (for more details, see Dehousse and Gadisseur, 2002; Moyson and Aubin, 2011). A similar process of liberalization of the European electricity sector was launched with directive 96/92/EC. The implementation of this process, in Belgium, began with the Federal Law of 29 April 1999 (for more details, see Declercq, 2000; Declercq and Vincent, 2000a, 2000b; Glachant and Perez, 2011).
The web survey was administered between April and November 2012 to 1256 people holding top to middle positions within 51 public and private organizations involved in the liberalization process. Given their position, those people had been regularly involved in the implementation process of the European liberalization policy: they form two policy subsystems. 2 The response rate of the survey was 32.88% (413 policy actors out of 38 organizations). This is very good, considering that web surveys conducted with email invitations and one reminder typically reach a response rate of 25% (Kaplowitz et al., 2004). 3 The survey invitation was personally addressed to each of the invitees (“Dear Mr./Ms. ‘Name of the invitee’,”) and sent to their individual, professional email address. Hence, it can fairly be assumed that the survey was filled by the invitees themselves. Together with the results of the survey, the anonymous ID of survey participants was provided, along with their language (French or Dutch), their gender, as well as their hierarchical level within their organization of affiliation. On those dimensions, the distributions of the sample and the population are similar, which suggests the sample is well representative.
Measures
Dependent variable
The evolution of respondents' policy preferences was measured with the “simple gain scores” method. Respondents were requested to report their initial policy preferences with the following question: “At the beginning of the liberalization process or—if later—when you were involved in the Belgian rail/electricity sector for the first time, what was your position about the following major principles of this policy?” Four Likert-type items ranging from “Very unfavorable” [−2] to “Very favorable” [+2] were submitted in each sector. In each sector, the two following items were used: “the application of regulation by independent regulatory bodies” (item 1); and “the unbundling of operations on and management of infrastructure” (item 2). In the rail sector, two additional items were submitted: “the introduction of competition in the railway transport of freight (item 3 rail) / international railway transport of passengers (item 4 rail)”. In the electricity sector, the additional items were: “the introduction of competition in the generation and supply of high-voltage electricity (item 3 electricity)/low-voltage electricity (item 4 electricity)”.
Then, the respondents were invited to report their current policy preferences with the following question: “And today, what is your position about the following principles?” The items used for reporting the initial preferences were submitted again. To get an idea of how respondents' preferences evolved over time, initial preferences' values were retrenched from current preferences' values. This provided a new list of items or “gain scores”—four in the railways and four in the electricity sector. Those scores measure the evolution of respondents' policy preferences toward the liberalization between the beginning of this policy process and 2012 (i.e., the year of the survey). For example, a minimum score of [−4] indicates a respondent who had a very positive opinion [+2] about a liberalization principle at the beginning of the policy process and turned out to be very negative [−2] about this principle in 2012. Inversely, a respondent got a maximum score of [+4] when his extremely negative opinion [−2] had become extremely positive [+2] in 2012.
Factor analyses were conducted on the list of four gain scores, in each sector separately. The exploratory factor analysis suggested to keep all scores in each sector. The confirmatory factor analysis validated this structure in the rail sector (χ2 = 0.21, p = 0.90; root mean square error of approximation (RMSEA) = 0.00; standardized root mean square residual (SRMR) = 0.014; comparative fit index (CFI) = 1.00) as well as in the electricity sector (χ2 = 0.69, p = 0.71; RMSEA = 0.00; SRMR = 0.009; CFI = 1.00). 4 The scores of the two factors were normalized to get one scale common to the two sectors. This scale ranges from [−6.99] to [+6.99]. This is the dependent variable of the present study. The items of the dependent variable, as well as their statistics, are reported in Appendix 1.
A cross-sectional survey comparing retrospective and contemporary data is not the best possible tool for assessing the evolution of policy actors' preferences over time. A panel survey, for example, could be even more appropriate. Still, this study offers a valid and reliable method that significantly improves existing approaches to the measurement of policy learning for two reasons.
First, the simple gain scores method overcomes the two possible types of systematic error in the measurement of policy learning. The first possible type of systematic error concerns the measurement of preferences change. In particular, respondents could be tempted to provide socially desirable answers—e.g. respondents willing to show that they are stable and reliable people or, on the contrary, that they are able to change their minds. The second possible type of systematic error concerns the measurement of policy preferences themselves. In particular, the survey concerned professional issues and was distributed in a professional context. Hence, there are good reasons to suspect respondents' willingness to appear more or less favorable to the liberalization process when he or she worked among colleagues militating for/against this policy, for example.
The existing studies which measured policy learning are relatively scarce and relied, most often, on one set of items on preferences change (“did you change your opinion on…”: e.g., Montpetit, 2007). Such an approach does not control for any type of measurement error mentioned above. In the simple gain scores method, in contrast, two sets of items—one about past preferences and one about current preferences—are used and compared by the researcher. On the one hand, this drastically decreases the ability of respondents to strategize regarding the social desirability of the reported changes in their preferences. On the other hand, the simple gain scores approach does not remove systematic error in the measurement of preferences themselves. However, simple gain scores modeling of preferences change protects regression results from the possible effects of such a measurement error—it provides unbiased results (see below).
Second, this study addresses recollection issues. Indeed, it can be difficult to remember past preferences (Janson, 1990). However, a confident attitude toward memory is a reasonable indicator of its accuracy (Roediger, 2012). In turn, conviction is a reliable indicator of attitude confidence/certainty (Holland et al., 2003). Hence, respondents were also asked to report their degree of conviction in their policy preferences on a five-point Likert scale. The respondents who reported to be “completely unconvinced” [−2] or “rather unconvinced” [−1] of their past or current preferences were removed from the sample (32 respondents were removed).
In addition, this study focuses on policy actors which have been involved in the European liberalization process for a long time. As this process has been a major and long-term policy change in network industries, there are good reasons to think that policy actors have good memories of their past preferences regarding this change. Indeed, research in cognitive psychology suggests that the importance of an event or process, as well as the number of opportunities to hear and discuss it, increases the accuracy of memories on past opinions toward it (Kvavilashvili et al., 2003; Neisser et al., 1996).
Independent variables
Personal policy interests were operationalized with three aspects of the policy actors' own material well-being which were likely to have been influenced by the liberalization process: their salary (like in Martinussen, 2008), but also their material working conditions and promotion possibilities. Whether these aspects had actually been influenced by the liberalization process over time was assessed by the respondents themselves, with Likert-scale items, ranging from “Strongly disagree” [−2] to “Strongly agree” [+2]. The aggregation of these items results in the ordinal variable “Personal policy interests” ranging from [−6: “the liberalization policy has influenced my personal interests very negatively”] to [+6: “the liberalization policy has influenced my personal interests very positively”].
Similarly, organizational policy interests were operationalized with two aspects of the material prosperity of policy actors' organization which were likely to have been influenced by the liberalization process: its finances as well as its success. The aggregation of Likert-scale items results in the ordinal variable “organizational policy interests” ranging from [−4: “the liberalization policy has influenced the material prosperity of my organization very negatively”] to [+4: “the liberalization policy has influenced the material prosperity of my organization very positively”].
The generalized importance of interests may be expressed by subjective beliefs (Clore and Schnall, 2005). For example, “I believe that earning a good salary is very important” is a good indicator that personal interests are important for an individual. The importance given to salary, but also professional opportunities and material working conditions, was assessed by each respondent thanks to three Likert-type items. The aggregation of these items results in the ordinal variable “importance of personal interests”, ranging from [−6: “I don't pay any importance to my personal interests”] to [+6: “I pay a great deal of importance to my personal interests”]. The importance given to the finances and success of the organization were assessed by each respondent thanks to two Likert-type items. The aggregation of these items results in the ordinal variable “importance of organizational interests” ranging from [−4: “I don't pay any importance to my organizational interests”] to [+4: “I pay a great deal of importance to my organizational interests”].
All independent variables have a cronbach alpha equal to, or higher than 0.70 (see Appendix 1). This study also accounts for the sector in which the respondent worked (rail sector = 0; electricity sector = 1) as well as several covariates: gender (male = 0; female = 1), age (from less than 20 year old = 1; to more than 70 year old = 12; by intervals of 5 years) and educational level (1 = secondary education or less; 2 = undergraduate; 3 = graduate or more). Battaglio and Legge (2009; Battaglio, 2009) also suggested that “satisfaction with democracy” and “political curiosity” (called “political interest”, in their own study) are two political predispositions regarding liberalization processes. These variables were measured with the two following Likert-type items: “I find that the Belgian democracy works well these days” and “I am interested in politics”.
Analysis
Summary statistics.
Out of the 413 survey respondents, 32 were removed because they were not confident enough about their past policy preferences (see above) and five others were removed because they did not provide any answer to one or more questions used for constructing the independent variables. The missing values of gender, age, and educational level were replaced by their mean (consistent with Allison, 2002). Hence, the final sample is composed of 376 respondents: 185 in the rail sector and 191 in the electricity sector. They come from 38 different organizations: 12 in the rail sector and 26 in the electricity sector.
To analyze the role played by interests in policy learning, this study relies on simple gain scores modeling, in which the difference between respondents' final and initial preferences is regressed on the independent variables. In “residual gain scores” modeling, in contrast, the initial attitudinal level is introduced as a regressor of the final attitudinal level, next to the other independent variables (Allison, 1990). The respective advantages of the two approaches have been subject to much debate, recent studies suggesting that they most often lead to similar results (Johnson, 2005). There has been wider consensus, however, that simple gain scores modeling is more appropriate when some error in the measurement of the initial attitudinal level is suspected (Johnson, 2005). Indeed, in simple gain scores modeling, measurement error is entirely within the dependent variable whereas, in residual gain scores modeling, the error in the measurement of the initial attitudinal level appears in the independent variables. In the present study, error in the measurement of initial policy preferences is suspected (see above). Hence, the simple gain scores approach offers a better model specification and unbiased results.
Correlation matrix of the dependent and independent variables.
+p < 0.10; *p < 0.05; **p < 0.001.
Regression models. a
Standard coefficients are computed with the OLS regression method. Standard errors are clustered robust. The 38 clusters are based on the organizational affiliation of respondents. BIC: Bayesian information criterion; SE: standard error.
+p < 0.1; *p < 0.05; **p < 0.01; ***p < 0.001.
Findings
The effect of covariates on the dependent variable is examined before turning to the results related to the independent variables. Model 0 contains all covariates. This model is significant but has a small adjusted R-squared (0.06). As a result of higher “reception skills” or society's greater “conformity expectations”, women are expected to show higher compliance—e.g. to policy decisions—than men (Petty and Wegener, 1998). The results confirm that female policy actors' preferences regarding the European liberalization have evolved more favorably than their male counterparts. In contrast, there is no evidence about the effect of age or educational level on policy learning.
I also suspected that the policy actors with a higher political curiosity could have become more favorable to the liberalization process over time due to the political nature of this process. Battaglio (2009) also suggested that such a curiosity increases citizens' ability to understand complex policies (rather than a tendency to reject those policies in the absence of such an ability). The results do not confirm this expectation.
Battaglio and Legge (2009) tested two contradictory hypotheses regarding satisfaction with democracy: “while it might be guessed that (1) satisfied citizens are not prone to change, (2) those who are supportive of the political system may have more confidence and familiarity to know its shortcomings and take steps that they view as reform, such as privatization” (p. 702). Their results show a positive relation between satisfaction with democracy and support of privatization. In contrast, my results show a negative effect of satisfaction with democracy on the evolution of policy preferences toward the liberalization policy over time. This finding suggests that the policy actors who feel satisfied with democracy have more confidence and familiarity with the political system to revise their opinion toward a policy negatively. Interestingly, this effect becomes even more significant when the independent variables are introduced in models 1–5.
A real but limited influence of policy interests on policy learning
Models 1 and 2 examine the effect of personal policy interests and organizational policy interests on policy learning. The likelihood ratio tests as well as the decrease of the Bayesian information criterion (BIC) index suggest that the two models significantly improve our understanding of the dependent variable, when compared to model 0. The standard coefficients of the variables demonstrate their substantial effect on the dependent variable. Hence, hypotheses 1.1 and 1.2 are validated: policy actors do align their policy preferences with their (personal and organizational) policy interests over time.
For the first time, this study provides support to the ACF's initial contention (Sabatier, 1993) that interests drive policy actors' attitudes toward policies. This evidence is based on self-reported measures of interests, which reinforces the view that interests are ideational constructs (Dudley, 2007; Laird, 1999). At the same time, it suggests that policy actors' belief system is internally consistent (Sabatier and Jenkins-Smith, 1993).
Interestingly, the results also give an accurate idea to what extent interests do influence policy actors' cognition. Model 3, which combines personal and organizational interests, accounts for 11% of the variation of the dependent variable. This means that interests definitely influence the course of policy actors' thoughts regarding policies. At the same time, this shows that 89% of the variation is explained by other factors. Those factors could of course be other forms of interests which have not been operationalized in this study, but also other types of non-materialistic beliefs and arguments (e.g., “I have become more favorable to this policy because it has improved the public service”). Hence, those results allow rejection of the null hypothesis that interests do not influence policy processes (Hoberg, 1996; Nohrstedt, 2005). At the same time, they quantitatively confirm that much variation in policy actors' preferences over time results from various kinds of policy beliefs and not only from purely egocentric, material interests (Kübler, 2001; Lau and Heldman, 2009; Martinussen, 2008; Sabatier and Jenkins-Smith, 1993).
Policy actors experience an “interest shift”
In this section, I examine whether the importance that respondents give to their interests in their everyday lives has a moderation effect on the relation between their policy interests and the evolution of their policy preferences. In regression models, such effect is assessed with the examination of the interaction term between the moderating variable and the main variable (Hayes, 2013). Model 4 shows that the interaction term between the importance of personal interests and personal policy interests is insignificant. Similarly, Model 5 suggests that the moderation effect of the importance of organizational interests on the relation between organizational policy interests and the evolution of policy preferences is insignificant. Hence, hypotheses 2.1 and 2.2 cannot be validated. In other words, this study suggests that the importance that a policy actor gives to his interests in his everyday life does not influence the alignment of his policy preferences with his policy interests over time.
As mentioned above, prospect theory (Barberis, 2012; Kahneman and Tversky, 1979) has already been used in the ACF to explain the “devil shift”. According to this shift, opposed policy actors see each other as more “devil” than they actually are. As individuals are naturally loss averse, policy actors exaggerate the risks represented by their political opponents (Leach and Sabatier, 2005; Sabatier et al., 1987).
Consistent with prospect theory too, the results of the present study suggest the existence of another shift, namely: an “interest shift”. Policy actors determine which policy options serve their interests best and maintain or strengthen their preferences for those options in the long run. However, the intensity of this interests–learning relation is independent from the generalized importance of interests. Rather, policy actors are averse to opt for policies which involve personal loss, even if they do not pay much importance to their interests in their everyday lives. This finding also confirms that policy actors' rationality is bounded rather than perfect (Sabatier and Weible, 2007; Simon, 1991).
Organizational policy interests are more decisive than personal policy interests
The comparison of Models 1, 2, and 3 suggests that organizational policy interests are a more decisive factor of policy learning than personal policy interests. The effect of personal policy interests (Model 1) becomes insignificant when organizational policy interests are introduced in the model (Model 3). The reverse is not true: the effect of organizational policy interests (Model 2) remains significant when personal policy interests are introduced (Model 3). Similar conclusions may be drawn from the analysis of the BIC index: the introduction of organizational policy interests (Model 2) reduces the BIC more than the introduction of personal policy interests (Model 1). Furthermore, the introduction of personal policy interests next to organizational interests (Model 3) slightly increases the BIC index, compared to a model in which only organizational interests are present (Model 2). Hence, respondents' personal policy interests are statistically less useful to account for the evolution of their policy preferences.
This result could be interpreted in terms of mediation—policy actors perceiving the effect of the liberalization policy on their personal interests “through” the effect of this policy on their organization. A variable B has a mediation effect on the relation between an independent variable A and a dependent variable C only if all the following effects are significant: A on C, A on B, and B on C. If this condition is met, the effect of A on C in a full regression model (with B) is subtracted to the same effect in a partial model (without B) to know the net mediating effect of B (Hayes, 2013). This means that the net mediating effect of organizational policy interests (B) on the relation between personal policy interests (A) and the evolution of respondents' policy preferences (C) equals 0.21 − 0.11 = 0.10. In other words, half of the effect of personal policy interests is mediated by the effect of organizational policy interests. This confirms the key role of organizational interests (rather than purely personal interests) in shaping policy actors' preferences over time.
Conclusion
The ACF (Sabatier and Jenkins-Smith, 1993) is a theory which considers the role of policy learning in policy change processes. Policy learning is a cognitive and social dynamic in which new information and knowledge resulting from various experiences and interactions can elicit enduring alterations of policy actors' preferences regarding policies. Those alterations have been the dependent variable of this study.
This study has demonstrated that policy actors tend to align their policy preferences with their interests over time. This tendency, however, is limited. In addition, it is independent from the importance that policy actors give to their interests in their everyday lives, which suggests that policy actors are subject to a sort of “interest shift”, when they assess policy options. Finally, when policy actors revise their policy preferences, the perceived impact of policies on their own material well-being (personal policy interests) is less decisive than the perceived impact of policies on the material prosperity of the organization which they are affiliated to (organizational policy interests).
As suggested by previous studies (Hoberg, 1996; Nohrstedt, 2005), these results shed new light on the exact influence of policy actors' interests on policy learning and policy change. The hypotheses of the research were based on implications of the ACF model of the individual, drawn from organizational research (Simon, 1991) and behavioral economics (prospect theory: Kahneman and Tversky, 1979). This model recognizes the bounded rationality of policy actors, the ideational character of interests, as well as a double logic of normative reasoning—a logic of consequences and one of appropriateness. As the results fit those assumptions, they provide new support to the validity of this model. This is not only relevant for the community of ACF researchers but also for scientists using the various policy approaches based on a similar model of the individual (Dunlop and Radaelli, 2013; Zito and Schout, 2009), as well as the researchers looking at the role of interests in social and political attitudes (see Sears and Funk, 1991; as well as subsequent studies).
This study opens several avenues for future research. First, future studies could look at the social practices and institutional settings that increase or decrease the influence of interests on policy learning and policy change (i.e., the contexts in which the “interest shift” is stronger or weaker). Second, the internal consistency among answers to a cross-sectional survey can be artificially higher than in a panel study. If possible, longitudinal data should be privileged in future studies on policy learning. Third, I used self-reported measures for the importance policy actors pay to their interests in their everyday lives. This can draw socially desirable answers. This potential problem could be overcome by using vignette questions, in which respondents have to report how they would behave in given contexts. Fourth, this study used a classical ACF approach, in which the focus is on one output of policy learning: the evolution of policy actors' preferences (Sabatier and Jenkins-Smith, 1993). However, individual learning is also a process of knowledge acquisition and translation (Heikkila and Gerlak, 2013). As knowledge acquisition and translation also contribute to the emergence of shared understandings of policy problems and solutions (Leach et al., 2014), future research could also look at the effects of psychological factors, such as interests, on this process.
To conclude, this study suggests that policy actors assess and revise their policy preferences over time using filtering lenses such as the perceived consequences of policy programs on their material interests. However, the influence of this filter is rather limited and policy actors seem open to many other forms of arguments. This finding suggests to policy practitioners that policy actors could be effectively convinced of the strengths and weaknesses of a policy program with arguments based on various values or societal norms, and not only on their interests. This confirms one of the fundamental intuitions of the ACF.
Footnotes
Acknowledgements
I am grateful to David Aubin, Eric Montpetit and Christopher Weible, as well as the two anonymous reviewers of Public Policy and Administration, for their precious comments on earlier drafts of this article. I also thank Steven Van de Walle for his trust. Finally, I acknowledge Anne-Catherine Denies for her invaluable support.
Declaration of conflicting interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: the Belgian Fund for Scientific Research (FNRS); and the Dutch Organization for Scientific Research (NWO) [Vidi grant no. 452-11-011].
