By Nathan Maddix
Can nudging overcome physics envy?
As is now well-known, tremendous results have been found for nudges – behavioral interventions designed to facilitate choice for welfare-promoting outcomes. In my work over the last 5 years, I have sought not only to design and administer nudges, but also to understand how economists and psychologists alike can overcome what one article has criticized as “physics envy”: Economists want to be the first to discover theories and general laws that specify the exact antecedents and consequents in causal relations, yet the social world unlike the natural sciences often masks how individual priors affect outcomes. In this article, I explore the future of behavioral research, which includes behavioral economics and finance, with five practical realities for nudges in both developing and developed contexts.
What is the future of behavioral research?
The future of behavioral research will require understanding how the very act of introducing the event of an intervention causes individual-level differences in behavior, not only aggregate-level data. It may be possible to harm a subgroup in light of preferences while improving results on average. It may be said then that the Higgs-Bossom, or “God particle,” of behavioral research is knowing how each nudge in the toolkit of choice architecture so far developed, from default plans in retirement savings to commitment devices in ethics contracts, will affect outcomes for within-group differences for each organizational, national, and economic context.
In search of the God particle
(1) Keep prying open the cognitive blackbox at the individual and subgroup level.
In recent years, cognitive science has opened the blackbox of our minds and freed social science from folk psychological assumptions about the soul and identity. Research has outlined the many ways in which cognitive processes lock on to patterns as much as how cognitive errors in memory and gaps in knowledge (so-called informational asymmetries) impact market outcomes. There is still much more work to be done on how identity and self-knowledge affect the power of nudges in thinking about both consumer preferences and the ethics behind nudges. Policies and organizations alike are biased by positive short-term results, but average improvements and rates of change due to behavioral interventions often hide what’s under the hood. It is possible to increase unwanted behaviors for some while data reveals positive results overall. To address these concerns, researchers are beginning to incorporate more robust data on socio-demographic characteristics, as well as personality and background data, to pin down how nudges affect individuals.
(2) Target N in addition to increasing and diversifying N.
It is common knowledge that increasing sample size – “N” – is often necessary for valid and statistically significant research findings. Research labs have difficult increasing diversity, as they have traditionally been limited to student populations. The result is focusing on size rather than quality. On the other hand, economists have benefited from big data in non-lab ‘experiments’ to find patterns and correlations by maximizing N. Behavioral economists have followed suit in bringing down the proverbial largest hammer by relying on culturally normative assumptions about which behaviors we should amplify (like retirement savings). Yet, all researchers need to target subpopulations in addition to large samples. Assumptions for which behaviors are nudge-worthy were established at a period of time when some found them to be choice-worthy over alternatives. It is worth questioning the normative force of existing nudges without refinement. The more complex the nudge, the more likely it will need tweaking in other contexts. Targeting subgroups independently helps one question mainstream norms with clear data on how cultural preferences, attitudes, and economic outcomes may differentially affect low-income groups, for example.
(3) Get out of the lab.
Behavioral policy – the celebrity marriage of behavioral science and public policy – has skyrocketed across the world, as is now well-known, in part due to the power of nudges. Small behavioral changes can and do shift averages. But this wisdom backed by scientific research is normative in essence. As norms vary around the world, so will the efficacy and moral appropriateness of specific nudges. Whether employers default employees into retirement savings or rainy day funds, the outcome will remain the same – increased enrollment – whether or not these outcomes are preferred. In recent research on the Save More Tomorrow program (forthcoming), I find evidence that individuals have difficult managing their finances between previous and future mental accounts for debt and savings, with those with high credit card debt preferring to save for the future when it may be in their best interest to pay down high interest credit card debt first. It is worth asking who is to blame for financial mismanagement when nudges may be so important.
The God particle of behavioral research can be construed as the increasing ability for researchers to incorporate background and demographic data about participants, whether schoolteachers, executives, or employees, into behavioral interventions. Indeed, personality research has been one of the fastest growing overlaps between the fields of psychology and economics, with researchers latching onto established psychological correlates between the individual and society to find new demographic details that differentiate individuals. The social interactions movement is another well-known example as to how the God particle in behavioral research is not inherent to the nudge itself but how local social networks causes differential responses to well-known economic scenarios.
(4) Identify, organize, and create the data you need for subpopulations.
Finding what I am calling the God particle of behavioral research will require integrating socio-demographic data about individuals as much as or more than the behavioral experiments, theories, and interventions themselves. Data science is a growing movement not only due to the profitability of buying and selling data, but also due in part to the accessibility of large and open source data sets. To say the least, effective behavioral research requires data expertise. One can in some ways think of database management as large subject pools. If you can run one intervention or regression on all the subjects in that database, then you can identify how that variable correlates with important aspects of those subjects. Of course, economists often run such experiments using data on revealed preferences, but behavioral researchers will do well to incorporate economic and socio-demographic data into experimentation going forward.
Where there is no precedent, and where psychologists have the upper hand, create a psychological scale that correlates with established economic indicators, like the Entrepreneur Finance Lab has done in developing nations to measure credit-worthiness for those without credit scores. In large measure, it is usually a good idea to pilot within one lab setting, run a lab experiment in multiple labs when possible, and then apply the finding in field settings where applicable, using careful judgment and discretion as to how the external group differs from the captive lab population. The art in the science is how the researcher marries lab and fieldwork. Indeed, many are bridging psychology and economics within lab contexts, and in this sense particle physics could even be achieved in a lab setting. By building the most diverse subject pool with representatives of all ages, socio-demographic, racial, and cultural backgrounds, behavioral interventions can be tested to understand how they affect diverse and specific populations. Even in those cases like defaults and reminders in which nudges have created well-known behavioral changes, it is not yet clear how the magnitude of these behavioral nudges might differ across contexts, regions, and nations.
(5) Don’t be fooled: Shouldn’t we expect variance in consumer preferences?
Researchers need to think more about how behavioral interventions affect every subpopulation that intervention affects rather than administering nudges point blank. One needs to identify the variance in behavioral interventions, not only the rates at which they are successful on average. For whom is each behavioral intervention most successful? Are there any groups that elicit reactance or rebound effects to behavioral interventions based on principles, circumstance, or how the nudge was presented? Perhaps this is a reason to revise the way a message is framed or an incentive design is structured. It may mean ‘reverting back’ to qualitative methods to gather intuitions about what may be causing these behavioral differences in some groups while the intervention appears so successful for others.
Think beyond RCTs
Traditional behavioral economics experiments slice groups by condition, and many assume that the magic in randomized controlled trials (RCTs) is the RCT itself. While this is somewhat true, practitioners can put almost any pair (a, b) into an RCT and likely come up with some differences without knowing what causes them to be different, or if these are even optimal compared to other treatment alternatives (c, …, z). As is common in market research, these findings may reveal nothing more than a greater sign (a > b) without research design integrated into current psychological theories and findings.
With more differentiation about individual differences, what was once a cure-all nudge, such as reminders, may reveal a more refined nudge would create the optimal change in behavior. What types of reminders most optimally work for which groups of people? The future of behavioral economics will move toward more precise measurements of demographic groups, and we will be able to account for variance that may exist in systematic cognitive biases, especially considering cultural influences. Such research will likely also point the way in showing how to overcome these biases by identifying for which groups behavioral interventions are not necessary and evoke no change.
The exciting aspect of the search for the God particle in behavioral research is that, like the international collaborations that led to the Higgs-Bosson, discovering how nudges affect each context requires international teams around the world. There will not be one researcher who discovers the God particle, and there is neither a theory of everything nor singularity to be found through nudges. Nudges are hot because they work so well, and each person can refine and implement nudges within one’s own domain. It is a small incentive mechanism, a governmental order without an official government.
Economists commonly choose to study behavioral ‘events,’ such as income shocks, pay periods, public policies, and institutions, where naturally occurring differences are compared in ‘natural experiments.’ We no longer need to be tied to the feeling that we are physicists, or even that it would be good for society if we were. In behavioral research, applied researchers are more akin to physicians, working with populations (and subpopulations) and diagnosing which interventions have worked and could be improved upon. Like careful physicians, part of the inscription in the Hippocratic Oath is a by-line that reads “for whom?” rather than “to whom it may concern.”