Benartzi, Beshears, Milkman et al. (2017) on Government Investments in Nudges

Shlomo Benartzi, John Beshears, Katherine L. Milkman, Cass R. Sunstein, Richard H. Thaler, Maya Shankar, Will Tucker-Ray, William J. Congdon, and Steven Galing, “Should Governments Invest More in Nudging?” Psychological Science 28(8): 1041–1055, 2017.

• Government nudge units have thrived since the British showed the way in 2010; these units have set up canny defaults or streamlined paperwork or otherwise looked to make presumably desirable behavior easy for individuals to implement.

• But nudges, with their definitionally-low impact on "standard" (monetary, say) incentives, are not the only way to alter behavior, nor necessarily the cheapest. They do, however, tend to be cheap, so even a small behavioral impact might be well worth the nudge expense. 

• An e-mail nudge aimed at inducing US military personnel to enroll in a retirement plan increased enrollments in one month from a baseline just above 1 percent to somewhere in the 1.6 to 2.1 percent range; this nudge cannot be said to be wildly effective, but it was so inexpensive that the increased retirement savings per dollar invested were quite large (and orders of magnitude larger than with traditional monetary incentives targeted at inducing savings).

• "To be maximally informative, future policy-oriented behavioral science research should measure the impact per dollar spent on behavioral interventions in comparison with more traditional interventions [p. 1042]."

• The authors identify policy outcomes such as the amount of retirement savings and then scour the academic literature (highly-ranked journals) for relevant studies of both nudge and non-nudge policy interventions, for the purpose of calculating impact (on the chosen outcome variable) per dollar spent. In addition to retirement savings, the outcome variables include energy conservation, college enrollment, and influenza vaccination.

• For each outcome measure, the oomph-per-dollar ratio is highest (easily) with a nudge intervention as opposed to a traditional policy lever aimed at, for instance, monetary incentives. (This is not to say that all nudge interventions are comparative winners, just that the most cost-effective policies are nudges.) Automatic enrollment in retirement plans, for instance, is better at increasing savings (per dollar spent) than is a subsidy such as providing a 50% match for individual contributions.

• Nudges seem to be particularly cost effective when the problem that is being addressed implicates shortfalls in the rationality of individual decisionmaking. In these settings, small changes in the choice architecture can have large effects, whereas monetary interventions to improve the benefit-cost calculus that presumably underlies those individual decisions are both expensive and rather ineffective.

• In some circumstances, nudges and traditional policy interventions might be effective in combination.

• The evidence suggests that the current investment in nudge interventions is suboptimal.

Simonson and Kivetz (2018) on Gal and Rucker (2018) on Loss Aversion

Itamar Simonson and Ran Kivetz, "Bringing (Contingent) Loss Aversion Down to Earth — A Comment on Gal & Rucker’s Rejection of 'Losses Loom Larger Than Gains,'" Journal of Consumer Psychology 28(3): 517-522, July 2018.

• The articles outlined in the two previous posts are part of a "Research Dialogue"; Simonson and Kivetz's reply to Gal and Rucker, outlined here, is an element of the same dialogue.

• Gal and Rucker are right in that loss aversion is neither as firmly established nor as universal as is typically thought. Nonetheless, their retention paradigm is not convincing evidence of an endowment effect sans loss aversion, and they underplay some of the strongest evidence in favor of loss aversion: people demonstrate significant aversion to a risky but highly favorable (in expected value terms) bet, when opposed to a riskless gain (or the status quo) that offers much less in expected value. People routinely turn down 50-50 bets that pay $200 if heads and lose $100 if tails. 

• “[T]he question relevant at the present time for our field is not whether loss aversion occurs on average (we think it does), but what factors moderate its presence and magnitude, and relatedly, what are its boundaries [page 518]?” 

• Much of the evidence for the endowment effect (like the unwillingness to trade mugs for candy bars or vice versa) is consistent with plausible, non-loss-aversion explanations, such as the awkwardness in coming to an agreement for trivial trades. But the retention paradigm is not very convincing as new evidence against loss aversion, because of the highly artificial settings that arise in trying to reframe the retention of something you own as an active choice. 

• In many circumstances, loss aversion does seem to be part of what is going on with endowment effects, even if other mechanisms, such as transaction costs, also are at play. Losses do tend to loom larger than gains, but this is a tendency, one contingent on other factors, and not a universal truth. 

• As Gal and Rucker (2018) note, the excessive commitment to loss aversion might crowd out research that can identify other factors that drive decision making.

Higgins and Liberman (2018) Reply to Gal and Rucker (2018) on Loss Aversion

E. Tory Higgins and Nira Liberman, "The Loss of Loss Aversion: Paying Attention to Reference Points," Journal of Consumer Psychology 28(3): 523-532, July 2018.

• The Gal and Rucker loss aversion article outlined in the previous post was part of a "Research Dialogue"; Higgins and Liberman's reply, outlined here, is an element of the same dialogue.

• Higgins and Lieberman agree with Gal and Rucker: the empirical support for loss aversion is not as strong as its reputation would suggest. Loss aversion is not universal. The more general (than loss aversion) notion of prospect theory – that “reference points increase people’s sensitivity to objective changes in value [p. 523]” – is still viable, however.

• Losses and gains in prospect theory are judged relative to some reference point, which often is taken to be the status quo. If the reference point is not the status quo, however, then gains (relative to the status quo) need not be less powerful than losses, even if that loss-aversion-style result would be case were the status quo the relevant reference point. Further, multiple reference points can be at play at any one time.

• Reference points tend to be outcomes which attract our attention. As a result, we are more sensitive to changes around those points than from changes elsewhere. But this increased sensitivity need not be asymmetric, need not involve loss aversion: sensitivity to either gains or losses or both can increase around references points.

• A second suggestion is that a relevant reference point when judging an outcome is what might have happened instead, the chief counterfactual; gains or losses relative to that alternative will take on intensified value. To just make a train is more enjoyable than making it easily, and to just miss it is more painful than to be much too late. Again, this approach does not suggest the sort of asymmetry that loss aversion requires.

• Reference points such as goals – 10,000 steps per day – might suggest loss aversion: step 10,000 is worth a lot more than step 10,001 – but, Higgins and Lieberman argue, goals as reference points need not involve loss aversion. Many market-based goals have built-in incentives that are more sensitive above the goal – for instance, an increased percentage of royalties from book sales – than below the goal.

• In long-term pursuits, dual reference points can be at play: the starting position might be most salient early in the process, but the ultimate goal takes on more prominence as the pursuit unfolds. Recall that Gal and Rucker suggest that what is taken to be evidence of loss aversion in the literature often can be explained by an inaction bias, where no loss aversion is at play. For long-term pursuits, the “action” alternative is the one for which this dual reference point view seems most apt, and the additional reference point (the goal) can be the source of a greater sensitivity in valuation from changes in the action alternative than in the inaction alternative. 

• Some people (the “promotion-focused”) might concentrate on progress, and others (the “prevention-focused”) might concentrate on avoiding losses. Even if the status quo is the same for both individuals, they compare it with different alternative reference points. For the promotion-focused, the status quo is a loss relative to the desired progress; for the prevention-focused, the status quo is a gain relative to the feared worsening. 

• If a prevention-focused person found herself below the status quo, she might choose risky strategies if they are her only hope of restoring the status quo. Promotion-focused people, alternatively, starting from below the status quo, will not feel all that motivated to regain the status quo (both are losses, given the reference points at work), but will be more motivated to go from the status quo to a better point. This story, for which there is empirical support, is not consistent with standard prospect-theory-style loss aversion. That is, Gal and Rucker are right, in that the psychological evaluation of negative events (losses) are not always greater than the evaluation of equivalent gains, and people are not always more motivated by the threat of losses than by the prospect of gains. 

Gal and Rucker (2018) on the Loss of Loss Aversion

David Gal and Derek D. Rucker, “The Loss of Loss Aversion: Will It Loom Larger Than Its Gain?” Journal of Consumer Psychology 28(3): 497-516, July 2018.

• Social scientists seem to all but universally believe in loss aversion, the notion that losses “loom larger” psychologically than do similarly-sized gains.

• Gal and Rucker claim that the actual evidence does not support any general tendency for losses to loom larger than gains: everything depends upon the context.

• There’s a bit of circularity in the promotion of loss aversion: some phenomenon (like the equity premium puzzle or the endowment effect) is “explained” by loss aversion, and then the existence of the phenomenon (the equity premium, the endowment effect) is taken to be evidence that loss aversion is pervasive. 

• The status quo bias might reflect a preference towards inaction – and such a preference can exist in the absence of loss aversion, due to the lack of a motive for action, or economizing on processing costs, or the tendency to regret errors of commission more than errors of omission. 

• When asked to trade their original good for an essentially identical one, loss aversion is not implicated – but people still show a large status quo bias. Action v. inaction confounds the loss-gain story. 

• Is the endowment effect just a case of a status quo bias, and therefore does not require loss aversion? 

• The “retention paradigm” recasts endowment effect experiments as willingness-to-pay (WTP) to obtain an item v. WTP to retain an item – so now the “inaction” choice is not to have the good in both cases. (That is, the confounding of loss aversion with inaction is sidestepped.) 

• If loss aversion is active, then in the retention paradigm, the WTP to retain will be higher than the WTP to obtain. But in the experiments, there was no premium to retain a good or service. For “mundane” goods – mugs, notebooks – obtaining tended to have a higher WTP than retaining. 

• One has to be creative to come up with reasonable “retain” scenarios! Fixing a broken phone, perhaps? 

• Analogous experiments ask if you would like to receive $0 for a good you own, or exchange it for another good. The second condition swaps the owned and alternative good. Mug v. $5 shows no endowment effect – even though the standard exchange paradigm (that is, not the "retention" paradigm) with these goods shows a significant endowment effect. 

• For the standard “loss aversion ratio” test, not accepting the bet is the status quo. This test, too, can be recast: Would you rather receive $0 with a probability of 1 or take a 50-50 bet with the possibility of winning or losing $15? People seem to have a slight preference for the risky alternative. 

• When stakes are higher, preferences shift toward the sure thing – but this could reflect risk aversion, not loss aversion. (And, loss aversion is generally taken to be independent of the stakes.) 

• When you ask people directly about the psychological impact of winning or losing something – you ask them how they feel about these events – losing doesn't dominate in terms of the magnitude of feelings. How do you feel about losing a mug versus winning a mug? 

• How do you feel about losing $3 versus winning $3? What about $100? At the low stakes, people seem to care more about the gain. 

• Loss “frames” are not generally more motivating than “gain” frames (despite some evidence to the contrary in certain domains). 

• Why is loss aversion so popular given its questionable evidentiary base? Perhaps in part due to a status-quo bias among researchers(!), or confirmation bias within Kuhnian “normal science.” 

• Loss aversion holds intuitive appeal: we all feel some losses acutely. And the name “loss aversion” itself is persuasive.