Critical perspectives on the practice of evidence-based behavioral public policy: in this symposium, we critically examine different aspects of evidence-based behavioral public policy. The three papers provide theoretical, methodological, and practical analyses of the popular evidence-based behavioral policy in both poor and rich countries, drawing on actual practice of experimental, behavioral and development economists, as well as cognitive psychologists and policy makers. The papers also offer several concrete proposals to improve scientific practice and its applications.
Nudge and boost are two proposals to improve social outcomes by influencing individual decisions. We argue that these approaches offer a limited basis for behaviourally informed policy because of their focus on individual behavior. This is problematic, because (1) with some well-known exceptions, individual rationality is neither necessary nor sufficient for aggregate rationality; and (2) social outcomes typically involve complex interactions between agents ignored by nudge, boost and their respective theoretical foundations in the psychological research on individual decision-making..
The behavioral sciences used to gather evidence for behavioral policies need to pay more attention to meso-level social coordination and to the effectiveness of possible interventions aimed at influencing it. Nudge and boost must be seen as complementary instruments to what we call design interventions targeted directly at institutions and social structures.
Why has the debate between nudgers and boosters been framed so as to leave out the possibilities of intervening on the rules of interaction (market and norm design)? Moreover, the debate completely ignores a vast body of empirical studies on the effectiveness of different techniques of behavior change relating to a variety of social and health issues. We argue that these oversights are due to problems in transferring knowledge over disciplinary boundaries.
It is not surprising that both nudge and boost have more or less ignored system-level effects, since both are rooted in experimental psychology of individual decision making, and the debate between them has been framed as being fundamentally about the scope of decision theory. Prospect theory underlying many of the nudge policies is an attempt to tie economics back to psychology according to the picture of a direct bottom-up interface between psychology and economics, the individual and the systemic. This compatibility of prospect theory with economics has probably been one of the principle reasons why nudge policies have been so easily accepted: there is an existing conceptual slot into which deviations and biases, and their implied nudges, can be inserted. However, this convenient interface also leads naturally to the assumption that correcting deviations in rationality at the individual level leads automatically to improvements at the system-level.
In contrast, advocates of boost attack the core ideas of decision theory in claiming that it represents a wrongheaded picture of (individual) rationality and consequently leads to ineffective (and morally suspect) policy recommendations. Fast and frugal heuristics aim to provide a more psychologically plausible picture of processes of rational decision-making in a complex world. However, theory of fast-and-frugal heuristics does not offer any systematic theory for reasoning about the system-level effects of boosts. This lack of a clear interface between fast and frugal heuristics and existing economic models has likely been a major factor hindering the adoption of the boost approach among social scientists and policy makers. Furthermore, the missing shared theoretical ground between research on behavior change in social psychology (see. e.g Michie et al.: Behavior change wheel) and the decision-theoretic focus of nudgers and boosters explains the divided state of behaviorally informed policy research. We argue that both parties to the debate would benefit by treating decision theory more as a language in which to reason about social behavior, rather than as a substantial explanatory theory of individual decision making that one should either defend or attack on empirical or normative grounds.
In recent years development economics has undergone an “empirical turn” (Angrist and Pischke, 2010), namely the extensive use of randomized field experiments (RFE) to produce evidence. Abhijit Banerjee and Esther Duflo at the Jameel Abdul Latif Poverty Action Lab (J-PAL) are the two leaders of this movement. They characterize J-PAL’s approach as a “new development economics” (Banerjee, 2005) based on the unique use of RFEs in order to (1) produce evidence on the effectiveness of development programs and then (2) use these evidence to guide policy makers.
Although the “new development economics” claims to be “theory-free”, its practice is largely informed and guided by the framework of behavioral economics: the J-PAL’s researchers focus on the behaviors of the poor, assuming that the poor, like the rich, suffer from various cognitive biases that hinder rational decision making, thereby keeping them trapped in poverty. RFEs are thus designed mainly to assess the power of different “nudging” devices to counteract these cognitive biases. We show how this approach operates in practice by examining a paradigmatic RFE study of this type, Pascaline Dupas’ experiments on measures to increase the use of bednets to fight malaria (one of the main causes of death in developing countries). We then argue that the continuous failure of the series of her experiments to find any effective nudge to change the behavior of the poor is partly but importantly due to the individualistic perspective on decision making, which the study inherits from behavioral economics. That is, the practice of allegedly “theory-free” RFEs in fact suffers empirically from the implicit theoretical perspective that takes as the main explanatory/causal factor individual decision making in isolation from interactive, social and institutional contexts.
The failure of Dupas’ experiments and the unfulfilled promise of the “new development economics” more generally suggest that the evidence-based development economics movement may gain from shifting the focus from isolated individual decision makers to aggregate choices in social and institutional contexts. There are different ways of implementing this perspective shift, such as interdisciplinary collaborations with non-experimental social scientists such as anthropologists and area studies researchers to better understand the contexts in which the poor make decisions. Here we propose an alternative, i.e. to reconsider experimental practice in behavioral economics upon which the new development economics is built. Our proposal is motivated by the new practice called “experimetrics” (Bardsley and Moffatt 2007) or “behavioral econometrics” (Andersen et al. 2010), which adopts econometric techniques to explicitly model heterogeneity of data generating processes in the population. This opens up a way to understand how interactions of people with different beliefs and preferences result in aggregate results, and how the same individuals change behavior from one context to another. These are key knowledge to effective policy interventions, which however RFEs have failed to provide so far. We propose that the proponents of RFEs drop the rhetorical emphasis on its “evidence-based” nature and shift the individualistic perspective on poverty, which, upon careful examination of different strands of experimental practice, we argue, turns out to be not only unnecessary but unsatisfactory as a guiding tool for policy.
Recent theoretical efforts in modelling boundedly rational agents are having considerable impact on policy making, leading to the design of interventions that take into account the cognitive limitations of decision makers. Yet considerations of bounded rationality have not led to a uniform kind of public policy proposal. In this paper, we distinguish two kinds of policy proposals, nudges and boosts, which arise from theories of bounded rationality, and provide a framework that allows assessing, in a given context, which policy type is more likely to achieve a specific goal.
Nudges are inspired by the heuristics and biases approach championed by Kahneman, Tversky and others, whereas boosts are promoted by the ecological rationality paradigm of Gigerenzer and colleagues. According to the heuristics and biases program, while heuristics may lead to good decisions and behaviours in some circumstances, often they must be regarded as irrational from the normative viewpoint of probability theory. Nudges constitute interventions on the decision environment that circumvent the limiting effect of cognitive biases by facilitating desirable, and deterring undesirable, behaviours. For example, it is assumed that many people would wish to save more for their retirement if they seriously considered their needs at an older age. However, due to an alleged mixture of presence-bias, inertia and visceral influences, they often fail to do so. Consequently, nudge proponents have suggested changing the retirement savings default for new employment contracts, which assume that employees make a high monthly savings contribution, unless they actively choose against it.
By contrast, the ecological rationality program argues that the rationality of heuristics depends on their fit with the decision environment. For instance, if a heuristic is biased to process only some but ignore other cues (frugality), it is regarded as ecologically rational if the few processed cues are valid predictors of a desirable criterion while unprocessed cues do not increase the heuristic’s predictive power. Consequently, boosts aim at aligning the decision environment and the kinds of heuristics people use by changing the decision environment and/or by extending people’s heuristic toolbox. For instance, fast and frugal decision trees may be designed to match the cue structure of an environment and evaluated against alternative algorithms. If they show improved performance over current practice, they should be taught to decision makers.
In this paper we leave aside theoretical questions regarding whether the two kinds of intervention follow directly and exclusively from the competing theoretical approaches, as well as normative considerations about their legitimacy and desirability. Instead, we treat them as complementary policy tools and assess the conditions under which they are more likely to achieve a specific goal in a given context. We argue that the effectiveness of nudges and boosts hangs on several dimensions and thereby identify the kind of evidence needed to establish it. Our approach provides a common framework in which to evaluate the applied value of two popular concepts of bounded rationality and point to ways in which some of the disputes about nudges and boosts may be settled empirically.