Andrew Caplin told us about a new experiment that adds to the debate about “nudges.”

We have initiated experiments to study this tradeoff experimentally in a setting where imperfect perception seems highly likely and choice quality is easy to measure. In each round, subjects are presented with three options, each of which is composed of 20 numbers. The value of each option is the sum of the 20 numbers, and subjects are incentivized to select the object with the highest value. In the baseline treatment (“33%, 33%, 33%”), subjects were informed that all three options were equally likely to be the highest valued option, but in two other treatments, they were nudged towards the first option. In one of the nudge treatments (“40%, 30%, 30%”), subject were informed that the first option was 40% likely to be the highest valued option (the other two were both 30% likely). In the other nudge treament (“45%, 27.5%, 27.5%”), subjects were told that the first option was 45% likely to be the highest valued option (the other two were both 27.5% likely). Subjects completed 12 rounds of each treatment, which were presented in a random order.

The subjects got the best option only 54% of the time revealing that effort was required to add up all 20 numbers three times to find the largest sum.  The nudges gave them hints but notice that the hints also lower the return to search effort.  So in theory there will be both income and substitution effects.  And in the experiment you see evidence of both.  Their choices reveal that they utilized the hints: they more often chose the highlighted alternative. But, the interesting finding is that their chances of getting the best alternative did not increase.  In essence, the hint perfectly crowded out their own search effort.

You could take a pessimistic view based on this:  nudges don’t improve outcomes, they just make people lazier.  But in fact the experiment suggests a nuanced interpretation of nudges.  Even if we don’t see any evidence that, say published calorie counts improve the quality of decisions, that doesn’t imply that they have no welfare effects.  Information is a fungible resource.  If you give people information, they can save the effort of gathering it themselves.  Given that information is a public good, these are potentially large welfare gains that would be hard to measure directly.