Doing Experiments in Public Management Research: A Practical Guide

Abstract

Observational data are routinely used in both quantitative and qualitative public management research. They can be used to study a broad range of research questions. However, it is often challenging to draw causal conclusions from studies of observational data. This is due to selection bias, omitted variables and endogeneity, which are problems that may be difficult to avoid when collecting observational data. There are various techniques that can be employed ex-post to remedy these problems. But these solutions may not always be available, and they are often challenging in terms of complexity. In contrast, the core idea of experimental methods is collect good data that do not need ex-post correction in order to be used for causal analysis. This is why experimental methods are sometimes referred to as a design-based approach to causal research. The emphasis is on building a strong research design. The quality of the data means that the ensuing analysis of the collected data can often be done in a simple and transparent way. The evidence may lie in a simple comparison of means between control and experiment groups. Experiments come in different types which each have distinct advantages and disadvantages. They are discussed in more detail by Blom-Hansen, Morton and Serritzlew (2015). In this paper we focus on the challenges of practical experimental research.