Read the full paper here: static1.squarespace.com/static/5db66...

Lastly, we develop and estimate a job search model to quantify these frictions. We find that the letters reduce fixed learning costs by $584 (18%) while work-search requirements cost claimants ~$200/week but raise job offer arrival rates by 3%.

Surprisingly, inducing more workers to claim UI did not prolong unemployment. If anything, it modestly improved re-employment.

We argue this is bc work-search requirements screen out those unwilling to verify their search while accelerating job finding among those who comply.

Surprisingly, inducing more workers to claim UI did not prolong unemployment. If anything, it modestly improved re-employment.

We argue this is bc work-search requirements screen out those unwilling to verify their search while accelerating job finding among those who comply.

The de-stigmatization message further increased take‑up, but primarily for high‑wage workers.

In a separate survey, we show that this message successfully reframes UI as an earned benefit rather than a means-tested assistance program.

The letters also increased rejection rates, though. We find that complier applicants have much higher rejection rates than Always Takers (68% vs 36%), largely because they do not meet work‑search requirements or fail to respond, rather than because of underlying eligibility.

We find that the informational letters increased the application rate by 1.4pp (83%⬆️relative to control) and benefit receipt by 0.4pp (42%⬆️), with the largest effects concentrated among low-wage workers.

All workers in the treatment group received a letter with generic information about the program, who qualifies, and how to apply.

We also test two additional treatment arms that add language to reduce the stigma of claiming benefits or to set expectations about job‑finding.

We construct our experimental sample by identifying "potential job losses" in near real time. These workers were stably employed in prior quarters, experienced a sudden drop in hours in the most recent quarter, and then received our letter ~5 weeks after that quarter ended.

To do so, we implement a large-scale field experiment among 50,000 recently unemployed workers in WA State by mailing letters to potential claimants with information about UI benefits, and then tracking the impact on take-up and job search.

Only about half of eligible US workers claim UI, leaving thousands on the table at a time when they need it most. This paper explores what keeps them from claiming, who responds when frictions are reduced, and how higher take-up may affect job search

Main Results:

1️⃣ Informational letters increased applications and benefit receipt, esp for low-wage workers

2️⃣ Higher rejection rates indicate reduced learning costs rather than improved beliefs about eligibility

3️⃣ Positive (but modest) effects of the letters on re-employment

I want to quickly note that this is Brendan's job market paper, and one of our three co-authored papers that explore the causes and consequences of incomplete UI take-up.

🚨🚨🚨 NEW WORKING PAPER 🚨🚨🚨
"Barriers to Benefits: Unemployment Insurance Take-Up and Labor Market Effects (with Brendan Moore)

➡️ We ran a large-scale field experiment among 50k likely eligible workers to study the role of information frictions in the incomplete take-up of UI

The Benefits of Unemployment Insurance For Marginally Attached Workers Expanding unemployment insurance access improves job quality and earnings with minimal delay in re-employment, especially when paired with employment services.

A link to the full paper here
equitablegrowth.org/working-pape...

A thread with more details here
bsky.app/profile/case...

Making the case for expanding UI eligibility to more workers, based on our recent working paper with @equitablegrowth.bsky.social

See below for the full paper, my website with more research, and my email for comments/questions!

📝: casey-mcquillan.github.io/files/McQuil...
💻: casey-mcquillan.github.io
📧: caseycm@princeton.edu

One takeaway is that raising benefit levels is more expensive than you'd think, but the broader lesson is that incomplete and endogenous take-up is a first-order concern for the optimal design of social insurance.

Also, recall our earlier paper estimated the MVPF for expanding UI access to low-income workers, which suggested this would be a much, much more cost-effective policy.

bsky.app/profile/case...

Existing estimates of the MVPF don’t account for this take-up response, and so they overstate the cost-effectiveness of raising benefits.

If this is the measure we use to decide where to put tax dollars, this difference meaningfully shifts policy priorities.

Combining theory and empirics, we quantify the policy implications. Accounting for endogenous take-up:
- Lowers the optimal benefit level by 27% ($633➡️$451)
- Reduces the marginal value of public funds (MVPF) by 29 percent (0.90➡️0.66)

Optimal policy depends on total benefit payments, whether from higher take-up or longer claims. We already knew that the duration elasticity enters the optimal policy condition, but the insight from this paper is that the take-up elasticity belongs there as well

We develop a model with worker-specific hassle costs, which explains both incomplete take-up and why take-up rises with benefits. We then derive optimal policy in three cases: (1) no behavioral responses, (2) endogenous search only; and (3) endogenous search and take-up

As a benchmark, we consider what happens if we only look at the duration response: a 10% increase in weekly benefit extends avg claim duration by 1.6%. This means the take-up response drives more than two-thirds of the effect on benefit payments, tripling the fiscal externality.

Our results suggest that a 10% increase in weekly benefit would increase take-up by 4.7%, which drives a 6.2% increase in the number of benefit payments

To identify the causal effect, we use a regression kink design (RKD) that exploits non-linearities in the benefits schedule. The intuition is that a kink in benefits should create a kink in the outcome, and we can estimate the effect by comparing the size of these kinks.

We construct a sample of likely-eligible workers using employer-employee matched data, and then match this sample with admin records of claims and benefit payments to determine whether if a worker received benefits and the duration of their claim.

After a job separation, a worker needs to file an initial claim, and then a ``waiting week’’ claim that will not be paid out, and then another claim the following week that can result in payment. We track workers from their job separation through these steps.

We typically think about UI in terms of an insurance-incentive tradeoff: more generous benefits better insure workers against job loss, but also reduce the incentive to find re-employment.

But this overlooks that more generous UI may mean more workers take up in the first place!