Not sure anyone's actually on here anymore, but just in case, here's a

NEW PAPER ALERT

"The Behavioral Foundations of Default Effects: Theory and Evidence from Medicare Part D"

w/ @zarekcb @AdelinaYWang @borisvabson

https://www.nber.org/papers/w28331 

Thread

Defaults have been shown to be powerful in all kinds of settings: 401k, organ donation, health insurance, etc.

Less is known about *why* people follow defaults.

This is not just an academic curiosity. Different models of default effects have vastly different implications for both the value of defaults and the design of defaults.

For example, while some models suggest defaults should match consumer preferences, other models suggest defaults should be set to be as far away from consumer preferences as possible!

Intuition here is that mis-matched defaults can "shock" consumers into making active choices.

In this paper, we try to make some progress toward determining which of these classes of models is closer to the truth when it comes to health insurance markets for low-income households and thus what optimal defaults should look like in those markets.

To do so, we leverage several natural experiments in the Medicare Part D low-income subsidy program. We start by using those experiments to show that, yes, consumers overwhelmingly follow defaults in these settings.

Result 1: Less than 20% of LIS beneficiaries opt out of their *randomly-assigned* default plan prior to enrollment. By the end of 5 years, fewer than 50% of benes have opted out.

*** At any given time, a full 2/3 of LIS benes are in a plan they did not choose to enroll in ***

This holds for sick and healthy benes, benes with lots of drugs and benes with few drugs, benes who are assigned to plans that are particularly bad matches for them (based on how the formulary matches their drug regimen) versus particularly good matches, etc.

Result 2: When the default is exogenously switched from staying in one's prior plan to a forced switch to a new *randomly-chosen* plan, almost everyone switches and follows the new default.

On left, default = stay in plan
On right, default = switch to randomly selected plan

Clearly, people follow defaults in this market. Given this, how should we design these defaults?

To answer this, we develop a general model of discrete choice with costly attention that nests various models of active choice from the behavioral lit.

Specifically, the framework nests models of rational inattention, boundedly rational attention, and exogenous/random attention.

We then use the model to ask how a regulator should set the default to maximize welfare. The key comparative static is below.

We show that the question of whether to set a paternalistic default that matches consumer preferences or a "shocking" default mis-matched to consumer preferences depends on a critical parameter: The elasticity of active choice to the value of the default

Intuition is straightforward: If setting a worse default does not trigger anyone to make active choices, then there is no downside to setting a good default that matches preferences.

We then leverage the *randomly-assigned* defaults in the LIS program to estimate this parameter.

First, we show that enrolling in a plan whose formulary is a poor match for your drugs (blue) lowers your drug consumption by much more than a plan that is a good match (red).

Then we show that assigning someone a poor-fitting default plan induces almost exactly the same amount of active choice as assigning someone a well-fitting default plan.

In other words, the elasticity is approximately zero.

If worse defaults don't induce active choice, then the optimal default is clearly a paternalistic one well-matched to consumer preferences.

In the paper, we do more work to differentiate between different models of active choice, with most evidence pointing to a model of random/exogenous attention.

We find little evidence that you can "shock" someone to make an active choice with a default that is "bad enough," an idea that is often suggested in the behavioral literature. Instead, people seem to randomly decide when to pay attention and choose.

All of this

(1) suggests that defaults should be used to steer people to options that are best for them not to shock them to make choices and

(2) raises important concerns about the assumption that demand-driven competition will produce efficient outcomes in these markets.