Survey experiments for public management research

Introduction Over the past decades, public management and administration research has generated a vast body of knowledge using survey research. Some even argue that surveys have become the modus operandi for much contemporary public management scholarship (Groeneveld et al. 2015). Indeed, a large share of published empirical evidence about public management and administration comes from surveys about such topics as work motivations (Steijn 2008; van Loon 2016), experiences with red tape (Brewer and Walker 2010; Kaufmann and Feeney 2012), strategic management practices (Andrews, Boyne and Walker 2006) or citizen satisfaction with and trust of government (Van de Walle, Roosbroek and Bouckaert 2008; Van Ryzin 2007). While the sheer number of studies using data from surveys of public managers or citizens has increased, so has the methodological advancement of survey research more generally. Most recently, public management scholars have started using surveys collected in multiple countries (Bullock, Stritch and Rainey 2015; Jilke 2015) and at different points in time (Kjeldsen and Jacobsen 2013; Vogel and Kroll 2016), and have complemented these undertakings with elaborations on survey measurement (Jilke, Meuleman and Van de Walle 2015; Meier and O'Toole 2013). The aim of these efforts is to increase the empirical credibility of survey research. Yet critiques have increasingly pointed to the limitations of conventional survey research to draw firm causal conclusions (e.g., Perry 2012). Problems of selection bias, spurious correlations, omitted variables or reverse causation constitute serious challenges (see Chapter 4 for more discussion of these issues). And indeed, the potentially endogenous nature of a great deal of public administration and management survey research has clear drawbacks. As Chapter 1 in this volume notes, this has led public management researchers to turn increasingly toward experimental studies, including so-called survey experiments. Survey experiments involve randomisation of question wording, text, images or other information in a survey instrument that is administered to a sample, often representing a larger population. Survey experiments can be used to study a large variety of research questions, both substantive and methodological. When conducted on large, representative samples, they combine the internal validity provided by randomised experiments with the external validity provided by survey sampling.