Matteo M. Galizzi, Glenn W. Harrison and Marisa Miraldo
Abstract
The use of behavioral insights and experimental methods has recently gained momentum among health policy-makers. There is a tendency, however, to reduce behavioral insights applications in health to “nudges,” and to reduce experiments in health to “randomized controlled trials” (RCTs). We argue that there is much more to behavioral insights and experimental methods in health economics than just nudges and RCTs. First, there is a broad and rich array of complementary experimental methods spanning the lab to the field, and all of them could prove useful in health economics. Second, there are a host of challenges in health economics, policy, and management where the application of behavioral insights and experimental methods is timely and highly promising. We illustrate this point by describing applications of experimental methods and behavioral insights to one specific topic of fundamental relevance for health research and policy: the experimental elicitation and econometric estimation of risk and time preferences. We start by reviewing the main methods of measuring risk and time preferences in health. We then focus on the “behavioral econometrics” approach to jointly elicit and estimate risk and time preferences, and we illustrate its state-of-the-art applications to health.
1.1. Introduction
In the last few years, insights from behavioral economics have gained momentum among health policy-makers. Governments in developed countries have constituted “behavioral insights teams,” within their civil services, including the so-called “Nudge Unit” in the UK Cabinet Office, the behavioral science teams within the UK Department of Health, the NHS, and Public Health England, and analogous initiatives in other governments and international institutions (Dolan & Galizzi, 2014a; Sunstein, 2011).
At the same time, experimental methods have attracted attention among researchers and policy-makers. Most researchers and decision-makers in health economics, policy, and management, however, still implicitly associate experiments with developing countries, where the priorities and organization of health policies and systems are centered around access to care and the outspread of infectious diseases. Almost invariably, these experiments are also viewed as synonymous with “randomized controlled trials” (RCTs).
There is a tendency to reduce the remit of “behavioral” applications to health to “nudges,” to subtle changes in the decision environment or “choice architecture,” that trigger changes in health behavior at an automatic or unconscious level (Galizzi, 2014). We argue that there is much more to behavioral insights and experimental methods in health economics than just nudges and RCTs. We argue this for two reasons.
First, from a methodological perspective the current emphasis on RCTs is misleading because it overlooks the richness of a broad array of experimental methods spanning the lab to the field. The discussion about experimental methods in health economics should be centered around the advantages, disadvantages, and potential scope of the different types of experiments proposed by Harrison and List (2004), each of which could prove useful in health economics. The different types of experiments, including RCTs, should be seen as complementary rather than substitutes. More generally experiments should be seen as complementary to theory and econometrics.
Second, in health economics, policy, and management experimental methods and behavioral insights go beyond nudges and are applied to contexts other than developing countries. Many challenging applications are in high- and middle-income countries, which are redesigning their health systems to address the pressing challenges brought about by aging populations and increased prevalence of noncommunicable diseases.
Why does this focus on RCTs and nudges matter when we are concerned about developments in health econometrics? Advocates of nudges and RCTs often argue that in assessing interventions one should “let the data speak,” that is design an experiment that lets the data speak directly to those wanting policy evaluations. Simple “average treatment effects,” ideally, become just a matter of arithmetic. This is viewed as a positive thing as it allows policy evaluations that do not depend on econometric assumptions, beyond the basics of randomization. The same agnosticism extends to the use of formal theory. In what follows, however, we will argue that theory, behavioral experiments, and econometrics are complementary methods, rather than substitutes.
We illustrate this point by describing the applications of experimental methods and behavioral insights to one specific topic of fundamental relevance for health research and policy: the elicitation and estimation of risk and time preferences.
Section 1.2 discusses the links between RCTs and the broader range of types of experiment. Sections 1.3 and 1.4 illustrate the use of behavioral experiments in health, focusing on risk and time preferences. Section 1.3 contains reviews of the main methods to measure risk and time preferences in health, while Section 1.4 focuses on the “behavioral econometrics” approach to jointly elicit and estimate risk and time preferences, and illustrates applications to health.1 Finally, Section 1.5 concludes.
1.2. Randomized controlled trials and experiments in health economics
Many policy-makers tend to automatically associate behavioral experiments in health with RCTs, probably because one of the most influential reports by the UK’s Behavioural Insights Team, when it was still within the Cabinet Office, stressed the need to conduct RCTs to develop public policy (Haynes, Service, Goldacre, & Torgerson, 2012).
This emphasis on the use of RCTs as the exclusive or fundamental experimental method in health economics, policy, and management calls for two conceptual clarifications.
First, health economists are particularly well aware that the use of RCTs in science is not a new development. Modern evidence-based medicine and pharmacology are all based on RCTs, starting from the pioneering work on scurvy by James Lind in 1747, to the first published RCT in medicine by Austin Bradford Hill and colleagues in 1948. Far from novel, the idea of using randomized controlled experiments has been advocated for decades even for policy applications (Burtless, 1995; Ferber & Hirsch, 1978; Manning et al., 1987; Rubin, 1974).
Second, while the term RCT is now widely in vogue in policy circles, it is often used in a quite peculiar way. In the health economics applications in developing countries, the term RCT is typically used to denote large-scale experiments conducted with entire organizations (e.g., hospitals, villages) without necessarily allowing the stakeholders in those organizations to explicitly express their views or their consent to the proposed manipulations. This is a major conceptual difference with respect to RCTs in medicine or pharmacology, where subjects are always explicitly asked to give informed consent prior to take part into RCTs, and allowed to drop out with important ethical, political, and logistical implications. The term “RCT” is therefore conceptually inappropriate and practically misleading in a health economics and policy context, since it conveys the false im...