Cross checking vs nudges

The concept of an economic nudge is a small change to default information presentation or defaults will lead to big changes in pro-social behavior.  One of the keystone policies that relies on a nudge is automatic enrollment with an opt-out for 401K/403B retirement plans.  The logic was that people typically don’t save enough for retirement and the wonders of compound interest means that small amounts invested early is worth a lot by the time people retirement.  People had to make an active choice to sign up for a 401K and select the salary diversion to the 401K.  Doing nothing meant no money was put aside.

The nudge was to flip the cost of action.  Doing nothing now meant money was put aside, often with an employer match.

The political project behind nudging is that there could be an activist government that was barely noticed as the choice architecture could be technocratically tweaked.  It seemed like it was effective.  Peer reviewed studies routinely showed large changes. Some of my research is at least nudge adjacent as I’m working on choice architecture and administrative burden of zero premium plans.  I have a new manuscript that just went under review this week that looks at how people make choices under different information vectors.

However, there is a new systemic review of the impact of nudging across multiple policy domains.  Maier et al does something very nifty.  They attempted to estimate the effects of publication bias — studies that show no results are far less likely to get published — on the overall estimates of the effects of nudging:

compares the unadjusted results to the publication bias–adjusted results.* Since publication bias–corrected three-level selection models are computationally intractable, we analyzed the data in two ways: 1) ignoring the three-level structure (column 2) and 2) using only the most precise estimate from studies with multiple results (column 3). Strikingly, there is an absence of evidence for an overall effect and evidence against an effect in the “information” and “assistance” intervention categories, whereas the evidence is undecided for “structure” interventions. When using only the most precise estimates, we further find evidence against an effect in most of the domains, apart from “other,” “food,” and “prosocial” (the evidence is indecisive) and weak evidence for the overall effect.^† However, all intervention categories and domains apart from “finance” show evidence for heterogeneity, which implies that some nudges might be effective, [MY EMPHASIS]

There is some evidence that there might be something here in the right situation, right intervention but nudging is neither a policy nor policy panacea.