Positive vs. normative

Normative: how people should behave according to a model

“A rational agent facing a lower price should demand more of a good.”

Positive: how people actually behave in the data

“When we reduced the price of air purifiers, we observed higher application rates.”

Economics uses both — theory (normative) generates predictions that we test with data (positive).

The rational choice model

A mathematical model of human behavior based on two core assumptions:

1. Agents evaluate costs and benefits of any decision
2. Agents choose optimally subject to their constraints

Implication: if health benefits exceed cost, a rational agent should adopt an air purifier. Research suggests those benefits far exceed the sticker price — so why do so few people apply?

iClicker

Given the health benefits exceed the costs, what would the rational choice model predict will happen when we offer people a fully subsidized (i.e., free) air purifier?

1. Nearly everyone accepts the air purifier

2. Almost no one accepts the air purifier

3. Somewhere in between

4. Inconclusive — we need more information

Behavioral economics

The rational choice model is powerful but relies on strong assumptions. When do they fail?

• Fail to act even when benefits clearly outweigh costs (inertia, procrastination)
• Are deterred by small hassles more than rational cost-benefit would predict (“sludge”)

Behavioral economics studies why these failures occur and what we can do about them.

Background: air quality and health

• Fine particulate matter (PM2.5) is the largest environmental risk factor for health and mortality in the United States (Liang et al., 2021, PNAS)
• People spend an estimated 90% of their day indoors
• Air purifiers can reduce indoor PM2.5 exposure — health benefits likely far exceed their cost; people are leaving health benefits on the table
• Motivating question: What is the constraint preventing adoption?

Research questions

1. How do subsidies affect application rates?
   • Does a bigger subsidy lead to more applications, and by how much?
2. How does “sludge” compare to monetary costs?
   • Do small non-monetary frictions deter as many applicants as large price increases?
3. Do these costs change who applies?
   • Does the subsidy level or type of friction shift the income or demographic composition of applicants?
   • Do effects on application rates vary by income and pollution exposure?

Consistent with rational choice

• Higher subsidies → more applications (downward-sloping demand)

Inconsistent with rational choice

Health benefits from air purifiers far exceed their cost. Under rational choice, people should apply even at 0% subsidy and tolerate minor paperwork to get a free purifier.

• Yet application rates fall sharply with small price increases and minor administrative requirements
• This scale of deterrence is hard to reconcile with a rational cost-benefit calculation

Experimental design

~90,000 California households received a mailing, randomly assigned to one of 9 treatment arms:

Stratified randomization by income × household size × PM2.5 quartile.

Application rates by treatment arm

Demand curve: subsidy arms only

ITT regression: application submitted

Application funnel: proof of residence arms

Did informing applicants about the document requirement screen out those who would not complete?

iClicker

Do you think treatment arms with higher subsidies resulted in more applicants with lower income?

1. Yes

2. No

3. Inconclusive

Who applies? Demographics of applicants

Who applies? Income distribution (KS tests)

HTE: by income group — specification

Does the effect of each treatment arm differ for high-income vs. low-income households?

\[y_i = \alpha + \sum_{j \neq 3} \beta_j T_{ij} + \delta \cdot \text{HiInc}_i + \sum_{j \neq 3} \gamma_j (T_{ij} \times \text{HiInc}_i) + \mu_s + \varepsilon_i\]

• \(T_{ij}\) = 1 if arm \(j\) (ref: T3); \(\text{HiInc}_i\) = 1 if mailer income > $75k; \(\mu_s\) = strata FE
• Table shows \(\hat{\gamma}_j\) (interaction terms only)

HTE: by income group

HTE: by PM2.5 exposure — specification

Does the effect of each treatment arm differ for households with high vs. low pollution exposure?

\[y_i = \alpha + \sum_{j \neq 3} \beta_j T_{ij} + \delta \cdot \text{HiPM}_{i} + \sum_{j \neq 3} \gamma_j (T_{ij} \times \text{HiPM}_{i}) + \mu_s + \varepsilon_i\]

• \(T_{ij}\) = 1 if arm \(j\) (ref: T3); \(\text{HiPM}_i\) = above-median PM2.5 (Census tract, CES 5.0); \(\mu_s\) = strata FE
• Table shows \(\hat{\gamma}_j\) (interaction terms only)

HTE: by PM2.5 exposure

What fits the rational choice model

• Higher subsidies → more applications (H1 confirmed): demand slopes downward
  • 0% subsidy arm has dramatically lower take-up than 100% arm

The core prediction — monetary price matters — holds up.

What is harder to explain (1/2)

Non-monetary frictions are surprisingly large relative to monetary costs

• Uploading one document deters as many applicants as a large price increase
• Inconsistent with rational choice given the large estimated health benefits — the cost of compliance is trivial relative to the health gains being foregone

What is harder to explain (2/2)

Did informing applicants about the document requirement screen out those who wouldn’t complete it?

• T5 (informed of requirement) and T6 (informed + helped) both affect the composition of who starts and who completes
• If start rates fall but completion rates rise, people may be rationally self-selecting in response to the information
• See the funnel slide for the data

Takeaways

1. Subsidies work — removing monetary barriers raises take-up

2. Sludge deters applicants on par with steep monetary costs — a simple document requirement reduces take-up by as much as a large price increase

3. Policy implication: reducing application complexity may be as effective as raising subsidies — and cheaper

Questions?

Thank you!

Matt Brooks | msbrooks@ucdavis.edu | mspitzerbrooks.github.io