This paper reviews the past fifteen years of implementation research, focusing on the strengths and weaknesses of top-down and bottom-up approaches. It argues that the 4–6 year time-frame used in most implementation research misses many critical features of public policy-making. The paper outlines a conceptual framework for examining policy change over a 10–20 year period that combines the best features of top-down and bottom-up approaches with insights from other literatures. Early American studies were pessimistic about the ability of governments to implement their programs. The second generation of studies were more analytical and comparative, seeking to explain variation in implementation success across programs and governmental units. However, they maintained the same top-down perspective, starting with a policy decision and examining its objectives over time. In the late 1970s and early 1980s, a different approach emerged, focusing on the strategies pursued by various actors in pursuit of their objectives. These studies showed that local actors often deflect centrally-mandated programs toward their own ends. The paper examines the top-down and bottom-up approaches in detail, including an analysis of their strengths and weaknesses. It then suggests a synthetic framework that integrates most of the strengths of the respective approaches and applies it to a longer time frame than in most implementation studies. The top-down approach starts with a policy decision and asks questions about the extent to which the actions of implementing officials and target groups are consistent with the policy decision, the extent to which the objectives are attained, the factors affecting policy outputs and impacts, and how the policy is reformulated over time. The Sabatier and Mazmanian framework identified six conditions for effective implementation: clear and consistent objectives, adequate causal theory, legally structured implementation, committed and skillful officials, support of interest groups, and changes in socio-economic conditions. The bottom-up approach starts with an analysis of the actors involved in service delivery and asks about their goals, strategies, activities, and contacts. It uses these contacts to identify the actors involved in the planning, financing, and execution of relevant programs. This approach is more suitable for policy areas involving a multitude of public and private actors. The paper argues that the top-down approach is useful in cases where there is a dominant public program or where the analyst is solely interested in the effectiveness of a program. The bottom-up approach is more appropriate in policy areas involving a multitude of public and private actors. The paper suggests a synthesis of the two approaches, combining the best features of both. It argues that the top-down approach has a comparative advantage in situations where there is a dominant piece of legislation or where research funds are very limited. The bottom-up approach is more appropriate in situations where there is no dominant piece of legislation but rather large numbers of actors without power dependency. The paper concludes that a synthesis of the two approaches is needed to understand policy change over periods of a decade or more. It suggests an advocacy coalition framework that combines the best features of both approaches, focusing onThis paper reviews the past fifteen years of implementation research, focusing on the strengths and weaknesses of top-down and bottom-up approaches. It argues that the 4–6 year time-frame used in most implementation research misses many critical features of public policy-making. The paper outlines a conceptual framework for examining policy change over a 10–20 year period that combines the best features of top-down and bottom-up approaches with insights from other literatures. Early American studies were pessimistic about the ability of governments to implement their programs. The second generation of studies were more analytical and comparative, seeking to explain variation in implementation success across programs and governmental units. However, they maintained the same top-down perspective, starting with a policy decision and examining its objectives over time. In the late 1970s and early 1980s, a different approach emerged, focusing on the strategies pursued by various actors in pursuit of their objectives. These studies showed that local actors often deflect centrally-mandated programs toward their own ends. The paper examines the top-down and bottom-up approaches in detail, including an analysis of their strengths and weaknesses. It then suggests a synthetic framework that integrates most of the strengths of the respective approaches and applies it to a longer time frame than in most implementation studies. The top-down approach starts with a policy decision and asks questions about the extent to which the actions of implementing officials and target groups are consistent with the policy decision, the extent to which the objectives are attained, the factors affecting policy outputs and impacts, and how the policy is reformulated over time. The Sabatier and Mazmanian framework identified six conditions for effective implementation: clear and consistent objectives, adequate causal theory, legally structured implementation, committed and skillful officials, support of interest groups, and changes in socio-economic conditions. The bottom-up approach starts with an analysis of the actors involved in service delivery and asks about their goals, strategies, activities, and contacts. It uses these contacts to identify the actors involved in the planning, financing, and execution of relevant programs. This approach is more suitable for policy areas involving a multitude of public and private actors. The paper argues that the top-down approach is useful in cases where there is a dominant public program or where the analyst is solely interested in the effectiveness of a program. The bottom-up approach is more appropriate in policy areas involving a multitude of public and private actors. The paper suggests a synthesis of the two approaches, combining the best features of both. It argues that the top-down approach has a comparative advantage in situations where there is a dominant piece of legislation or where research funds are very limited. The bottom-up approach is more appropriate in situations where there is no dominant piece of legislation but rather large numbers of actors without power dependency. The paper concludes that a synthesis of the two approaches is needed to understand policy change over periods of a decade or more. It suggests an advocacy coalition framework that combines the best features of both approaches, focusing on