For a policy analyst the policy problem is the starting point for the policy analysis process. During this process the policy analyst structures the policy problem and makes a choice for an appropriate set of methods or techniques to analyze the problem (Goeller 1984). The methods of the policy analyst are often referred to as the toolbox or toolkit of the policy analyst. In this metaphor, the toolbox of the policy analyst contains the different methods and techniques that are applied to the diversity of problems for which a policy analysis is needed. Numerous contemporary policy problems related to public policy, organizational strategy, and organizational change can be characterized as multi-actor problems. Such multi-actor policy problems are characterized by multiple actors with their own perspectives and interests regarding the problem. As a consequence of this multi-actor perspective, the toolbox has been extended with actor analysis methods. That is, actor analysis methods are being applied by policy analysts to analyze multi-actor policy problems. A literature review, however, shows that till now little policy analytic research has paid attention to the suitability of actor analysis methods to this type of problems. The selection of a specific actor analysis method for a given problem is complicated by the fact that the different methods till now have not been compared to each other for the same multi-actor problems. Most applications reported in scientific literature deal with the application of a single actor analysis method for a problem situation. As a result, we often do not know what the consequences for the analysis of a specific problem would have been if the analyst would have applied another actor analysis method. This research takes up the challenge of investigating the performance of actor analysis methods for multi-actor problems. It aims to develop a framework that can help guide the selection of actor analysis methods for multi-actor policy problems. Although actor analysis methods are starting to receive attention in the policy analysis literature no clear and agreed upon demarcation of actor analysis methods exists. We therefore find it important to address the question of what constitutes an actor analysis method and what not. We argue that the primary difference between a non-actor analysis method and an actor analysis method is that the method itself focuses on the analysis of characteristics of multiple actors. After the demarcation of the field we argue that actor analysis methods can be divided into two groups: actor analysis methods that model multi-actor decision-making, and structural actor analysis methods. Both of these groups are further specified. The actor analysis methods that model multi-actor decision-making can be subdivided into two groups: bottom up and top-down methods. The structural actor analysis methods can be subdivided into structural methods that focus on the actors or on the relationships between the actors. Because the policy problem is the basis on which the policy analyst selects the methods to analyze the problem, we develop a framework that relates multi-actor problems to actor analysis methods. However, little previous literature is available and therefore we could not rely on the work of others who attempted to develop frameworks that aid in matching multi-actor problems to specific actor analysis methods. In order to construct our framework we turned to published applications of actor analysis methods. From the categorization of actor analysis methods we selected five actor analysis methods for further study. In order to structure our analysis we viewed the application of an actor analysis method as a modeling exercise and used a conceptual model of the modeling process to study the individual applications of actor analysis methods. The analysis resulted in three types of multi-actor problems for which different actor analysis methods have been applied: conflict, negotiation, and deadlock. These results constitute our framework which shows the types of multi-actor problems and the actor analysis methods that have been applied to these problems. The types of multi-actor problems were inferred from the descriptions of the problems provided by the individual researchers. This because the researchers themselves do not explain why that particular method was selected for that particular problem. That is, the contemplation of why a certain method would be most suitable for a specific problem is not clearly expressed in literature and is the result of a choice made by an individual researcher. The framework shows that not all actor analysis methods have been applied to all different types of multi-actor problems (conflict, negotiation, and deadlock). The fact that for a certain category of problems in the available literature a method has not been applied to one of the problems is an interesting observation. This as there does not seem to be a fundamental limitation to the applicability of any of the actor analysis methods to the specific types of multi-actor problems. Not only does literature not clearly specify such limitations, the properties of the methods themselves do not seem to delimit the use of the method. Another question is, which actor analysis method would be most suitable for a given problem. One might conjecture that some specific properties of methods make that method more suitable for a certain type of problem situation. The notion of a more suitable method assumes that experience is collected which contrasts the results of the various methods when applied to a specific type of problem. The empirical evidence that would provide these properties is however lacking. We therefore conducted two case studies to gather this type of empirical evidence on the methods. These case studies represent two specific real world policy problems to which we have applied multiple actor analysis methods. As a measurement of method performance we chose to use the predictive ability of the methods as a proxy for method performance, i.e., we assed how well the methods predicted real world events. Our first case study studied the premise that different methods applied to the same problem will yield different results. One can interpret the framework in such a way that all methods that have been applied to the same problem are equally suitable for that problem and would perform equally well. We thus wanted to apply all methods that according to the literature have been applied to a certain problem to that specific problem. We selected a deadlock problem. For deadlock problems, three actor analysis methods have been used in literature: conflict analysis, transactional analysis, and Q-methodology. The first case study concerned the possible flooding of part of a polder in The Netherlands called the Groot Mijdrecht polder. For this case Q-methodology identified large disagreement about the policy problem as formulated by the involved authorities. The main result of the conflict analysis was the conclusion that a new process with an open problem formulation, as requested by the inhabitants, was infeasible. Transactional analysis suggests that a possible solution can be found in a new scenario where respect for the landscape is combined with the need to store water. To date no final decision regarding the possible flooding polder has been made. The case is still evolving, and what the final outcome will be regarding the polder is not yet certain. As a result, the current outcome of the case cannot be used as proxy for the performance of the method. Although the final outcome of the problem situation is not yet known, we could assess the intermediate situation. While we do not yet have the benefit of complete hindsight, at present certain options do not seem to be as likely as they were when we did our analysis. Q-methodology gave a better prediction on the short term than both conflict analysis and transactional analysis. On the longer term transactional analysis came closest to predicting real world events. It is important to underscore that the methods did yield different results and that the selection of methods therefore is far from arbitrary. The conclusion that different methods yield different results is independent of what will be the best performing method. If in future we will have the final outcome, we can then evaluate which of the methods had the highest predictive value. In our second case study we studied the premise that if a method had not been applied to a problem type yet, it would perform less than a method that had been applied to that problem type. One could argue that when a method has not been applied to a specific type of problem this represents a limitation of that method. On the other hand, we noted that the properties of the methods themselves do not seem to fundamentally prohibit use of a method for any of the problems. We choose to apply a transactional analysis and conflict analysis to a conflict situation. Transactional analysis, according to our literature review, had not yet been applied to a conflict situation before. The second case study concerned the potential unbundling of the network companies from the large integrated power companies in the Netherlands. The conflict analysis showed that the government had the power to pursue its preferred outcome: the unbundling of the network companies. This corresponds to what happened in reality. The result of the transactional analysis was more nuanced. That is, it suggested an intermediate solution and no complete unbundling. The assumption of the transactional method, that people trade control over issues, was not the most appropriate for the studied conflict situation as in the end no trade was made. We concluded that for this case the conflict analysis was indeed the more suitable method. We also concluded that although transactional analysis had never been used for conflict problems, this method could be used. The framework developed in this thesis is a first attempt at defining a structure that may help in the selection of actor analysis methods for multi-actor problems. The two cases are a first attempt of applying the methods in the light of the structure of the framework. As the first case is still evolving a final outcome is not known for this case and therefore a best performing method can not be established at this point in time. As a result, we can not relate certain properties of the methods to the differences in the performance of the methods. For the second case this can be done. In this case the best performing method, conflict analysis, was the method that according to the framework was the most suitable method. From a scientific viewpoint, the generalization of the results of a single case to general properties of the methods regarding the problem situations cannot be done. We thus conclude that although we are able to make a framework with the potential to guide the selection of methods, in order to reach this potential it must be substantiated by a larger number of cases all pointing to the same strengths and weaknesses of the methods. Only then can generalization in the form of a more detailed discussion of methods and problems be entertained in a scientifically sound fashion. In this thesis we proposed to use predictive value as the proxy for method performance. The strengths and weaknesses of the methods are not limited to the predictive value of the methods. Using the predictive ability of the methods to judge performance is a very narrow definition of the suitability of a method. Final judgment on whether a method is suitable can also depend on other issues like the goal of the analysis, the required results and practical considerations. This thesis thus represents an initial effort in work on a framework that relates methods (and properties of those methods) to types of problems. The major limitation of this thesis is that we can only provide initial, case-based support for our framework. As a result, future comparative research will have to clarify whether this framework will indeed aid in structuring the choices a policy analyst has to make in selecting actor analysis methods when faced with a multi-actor problem. When contemplating future research, the first issue that needs to be addressed is a further development of our understanding of the relationship between the method selected and the result obtained when dealing with a specific problem. Research dedicated to comparing methods needs to be pursued in order to better understand the characteristics of those methods. As experience with multiple methods applied to the same problem increases, our ability to identify specific characteristics of those methods in relation to the problems will be growing. Ideally, it would result in diagnostic rules that would guide the policy analyst in selecting methods and, maybe even more importantly, interpreting the results of those methods. One might argue that this type of comparative research is too time-consuming or too expensive. We would argue that if a scientific field is to further develop, a well-grounded understanding of the potential and limitations of the methods used is an absolute requirement. In the absence of that understanding, the decision of the individual policy analyst to apply a method for a specific problem remains a subjective decision – that is, not defended in scientific discourse. Although others might propose different strategies to study the relative merits of methods in relation to problems addressed, we feel that the topic of comparing methods in relation to similar problems must be further pursued. As policy analysts we have to move beyond the casuistry – the use of case histories – that describes single problems and single methods and work on formulating diagnostic rules by comparing the performance of multiple methods to given problems.