Behavioral Public Policy in Labor Markets

Learning Objectives¶

By the end of Week 9, students should be able to:

1. explain how behavioral frictions change labor-policy design beyond changes in statutory incentives;

2. classify behavioral frictions as correction targets, design levers, implementation constraints, dynamic-learning forces, equilibrium modifiers, and welfare complications;

3. connect tax credits, benefit take-up, UI, training, retirement saving, and disability rules to specific behavioral margins;

4. choose empirical designs for information, simplification, default, learning, take-up, and welfare questions;

5. diagnose when a behavioral policy claim is about behavior, implementation, equilibrium incidence, or welfare.

Opening Orientation¶

Week 9 asks a policy-design question: how should labor-market policy be designed when workers do not always perceive incentives, understand eligibility, complete applications, learn rules quickly, or reveal welfare-relevant preferences through observed choices?

This is not a generic public economics lecture and not a generic “nudges” lecture. The labor object is central. Tax credits operate through earnings and labor supply. Benefits operate through claiming and continued attachment. UI and job-search support operate through search, applications, and reemployment. Training policy operates through post-displacement investment. Retirement and work-linked saving operate through payroll systems and employer menus. Disability and Social Security rules operate through claiming, work incentives, and perceived schedules.

The organizing idea is that behavioral frictions can do several things at once. A salience problem may be a mistake to correct, a lever for policy communication, an implementation barrier, a source of slow learning, a channel mediated by tax preparers or employers, and a welfare problem if observed non-take-up is not fully informed choice.

Bridge¶

Weeks 2 through 4 introduced the worker-side behavioral objects: present bias, nonstandard beliefs, limited attention, salience, complexity, and learning. Week 8 then showed that firms and markets can exploit, accommodate, insure, or segment around those objects. Week 9 turns the full sequence into design questions. If a worker under-saves because enrollment is passive, policy might use auto-enrollment. If an eligible worker underclaims a benefit because the form is complex, policy might simplify the application. If labor-supply incentives are ignored because workers do not know the schedule, policy might change information rather than the subsidy rate.

Weeks 5 and 6 widened the field beyond isolated workers. Policy is experienced through employers, payroll systems, caseworkers, tax preparers, counselors, peers, households, identities, and norms. Behavioral policy is therefore not only worker-facing. It can be amplified, muted, or redirected by organizations and local information environments.

Week 7 supplied the empirical discipline. A policy intervention can shift information, salience, confidence, procedural knowledge, search effort, application completion, or defaults. The right design depends on which object is distorted, which labor margin moves, what variation identifies the effect, and what additional assumptions support welfare claims.

Field Core¶

Standard Labor Policy Versus Behavioral Labor Policy¶

A standard labor-policy model starts with a budget set, an eligibility rule, or an incentive schedule. The core design question is often how generosity, tax rates, benefit phaseouts, or work requirements affect labor supply, search, training, or claiming. That benchmark is indispensable, but it is incomplete when the policy people respond to is not the policy written in the statute.

A compact behavioral-policy environment is:

Here $\tau$ denotes financial incentives or budget-set parameters, $n$ denotes information or salience design, $s$ denotes simplification, administrative burden, or claiming support, and $d$ denotes defaults, commitment, or active-choice architecture.

In a standard model, much of the action runs through $\tau$. In a behavioral labor-policy model, the other components can be just as important. A tax credit can have a high statutory subsidy but low perceived subsidy. A training grant can have positive returns but low take-up because the application arrives during a stressful displacement spell. A retirement plan can have the same match but different saving because the default changes. A disability earnings rule can shape work only if claimants understand the threshold and believe the administration will apply it predictably.

The worker’s choice under a behaviorally mediated policy can be written as:

The action $a_i(p)$ might be hours, earnings, search effort, training enrollment, benefit claiming, retirement saving, or disability-related work. The object $b_i$ summarizes beliefs, attention, sophistication, trust, perceived eligibility, or other behavioral states. The term $K_i(a; p)$ captures procedural or cognitive costs: finding the rule, understanding it, filling out forms, waiting, proving eligibility, coordinating with work schedules, or managing stigma.

This notation makes the policy point clear. A reform can change the objective incentive, the perceived incentive, the cost of acting, the default action, or the pace at which workers learn.

Biases As Objects To Be Corrected¶

The most direct behavioral-policy role is corrective. Workers may misunderstand EITC schedules, underestimate returns to claiming, procrastinate on enrollment, under-save for retirement, misperceive disability work incentives, or fail to invest in training after displacement. In this view, the friction is a wedge between the action workers choose and the action they would choose under better information, lower hassle, or more stable self-control.

Labor-supply incentives are a natural example. In EITC settings, workers may not understand where the phase-in, plateau, and phaseout apply to them. Chetty et al., 2013 shows why knowledge differences across neighborhoods matter for responses to the EITC. The policy lesson is not that the statutory schedule is irrelevant. It is that the effective schedule depends on knowledge, attention, and local information.

Information about Social Security claiming rules has a similar logic. Liebman & Luttmer, 2015 studies whether people behave differently when they better understand Social Security. The labor-policy object is not only a benefit formula. It is perceived lifetime returns to claiming, work, and retirement timing.

Learning the tax and benefit schedule can also change labor supply. Kostøl & Myhre, 2021 is useful because it treats schedule knowledge as a first-order object rather than as a footnote to elasticities. In disability or transfer settings, weak response to incentives may reflect low schedule knowledge, not necessarily low preference sensitivity.

The corrective view should still be disciplined. Not every non-take-up or low response is a mistake. Some eligible workers may rationally avoid a program because the expected gains are small, the stigma is high, the compliance risk is real, or the interaction with work is costly. A behavioral correction claim needs evidence on the friction and a benchmark for what the worker would choose absent that distortion.

Biases As Tools / Levers For Policy Design¶

Behavioral frictions are not only problems to remove. Policy can sometimes harness predictable decision patterns to improve outcomes. Defaults, active choice, reminders, commitment devices, framing, and simplification can move labor-market outcomes even when financial incentives are unchanged.

Retirement saving is the cleanest anchor. Madrian & Shea, 2001 shows that default enrollment can strongly affect 401(k) participation. The behavioral mechanism can include inertia, procrastination, endorsement effects, limited attention, and choice overload. The labor-policy relevance is direct: saving is work-linked, employer mediated, and often implemented through payroll systems.

Default design also exposes the welfare problem. Bernheim et al., 2015 studies default options in 401(k) plans and asks how default-induced choices map into welfare. A large default effect does not automatically imply a large welfare gain. Some workers are helped by inertia being pointed toward saving; others may be pushed into a contribution rate or asset allocation that is not ideal for them.

The broader public-policy logic is developed in Chetty, 2015 and Bernheim & Taubinsky, 2018. Behavioral tools can complement prices when people underreact to incentives, but they can also substitute for prices in narrow settings where the key margin is attention, enrollment, or follow-through. In labor markets, this distinction matters because the same intervention can change labor supply, benefit receipt, saving, and administrative exposure.

The design question is not “nudge or tax?” It is: which component of $p = (\tau,n,s,d)$ is the binding margin for the target population, and what evidence shows that moving it improves welfare rather than only increasing compliance with the planner’s preferred action?

Implementation, Take-Up, Salience, And Administrative Burden¶

Many labor policies fail or succeed at the implementation margin. Take-up is not a clerical afterthought. It is a choice under information, trust, time pressure, stigma, and procedural cost.

A compact take-up object is:

Here $D_i(p)$ is participation or claiming, $\tilde V_i(p;b_i)$ is the perceived value of participation, and $C_i(p)$ is the perceived and actual cost of taking up the policy. Policy can change take-up through generosity, but also through salience, simplification, trust, reminders, timing, assistance, and procedural support.

Bhargava & Manoli, 2015 is the primary empirical anchor because the design turns psychological frictions and incomplete take-up into measurable policy objects. The setting is benefit claiming, but the logic travels to labor policy: EITC filing, UI applications, training enrollment, recertification, disability claiming, and post-displacement services all depend on whether people can recognize eligibility and complete the process.

Administrative burden can also change the composition of participants. If high-return or liquidity-constrained workers face the largest paperwork burden, simplification can raise efficiency and equity. If burden screens out low-gain applicants, simplification has different implications. This is why take-up designs should estimate not only average effects but also heterogeneity by baseline knowledge, expected gain, work constraints, language, trust, prior program experience, and local assistance.

The empirical methods often start simple: randomized letters, reminders, deadlines, pre-populated forms, assistance offers, and redesigned notices. Outcomes include application completion, claim timing, benefit receipt, earnings, employment, search behavior, and persistence. But interpretation requires care. A treatment may bundle information, salience, hassle reduction, deadline attention, and perceived legitimacy.

Dynamic Learning, Endogenous Information Acquisition, And Policy Persistence¶

Behavioral labor-policy effects are often dynamic. Workers learn rules from repeated filing, neighbors, employers, caseworkers, peers, and personal experience. They may acquire information only when the stakes become high enough, when a deadline approaches, or when an intermediary translates the rule into an actionable choice.

A generic learning object is:

The next behavioral state depends on the current state $b_{it}$, information acquisition $m_{it}$, new experience or outside information $x_{i,t+1}$, and the policy environment $p_t$. A worker may learn the EITC schedule through repeated tax filing, learn UI search rules through caseworker contact, learn training returns after a layoff, or learn disability earnings rules after approaching a threshold.

Dynamic learning matters for both estimates and design. Short-run estimates may understate long-run effects if people learn slowly. They may overstate durable effects if a reminder creates only temporary salience. Responses can diffuse through neighborhoods or workplaces. A reform that changes the interface may affect both current behavior and future beliefs.

This is one reason Chetty et al., 2013 and Kostøl & Myhre, 2021 are central for behavioral labor policy. They push students to ask whether observed labor-supply response is a preference elasticity, a knowledge elasticity, a learning process, or a composite of all three.

A useful teaching decomposition is:

The equation is not meant to be literal in every application. It is a reminder that policy effects may run through incentives, information, simplification, and defaults at the same time. The empirical task is to isolate or at least diagnose the channels.

Firms, Intermediaries, And Equilibrium Responses¶

Behavioral labor policy is not only about workers. Employers design retirement menus, payroll communication, scheduling practices, training supports, and leave interfaces. Tax preparers translate tax-credit rules. Job centers and counselors translate UI, search, and training policies. Caseworkers and disability administrators shape perceived risk, compliance costs, and trust.

Intermediaries can amplify policy. A tax preparer may explain the EITC schedule in a way that makes earnings incentives salient. An employer may set a retirement default that turns inertia into saving. A counselor may reduce training enrollment costs by helping a displaced worker navigate forms.

Intermediaries can also mute or redirect policy. Firms may adjust hours, scheduling, or benefits in response to tax credits or mandates. Employers may choose retirement menus that interact with worker inertia. Caseworkers may ration attention. Local networks may spread partial or inaccurate information. If a policy is evaluated only as a worker-side treatment, these institutional responses may be missed.

This carries forward the Week 8 equilibrium lesson. Policy treats intermediaries as part of delivery only after asking how firms and markets strategically adapt to predictable behavioral frictions.

Welfare, Targeting, And Normative Ambiguity¶

Behavioral labor policy makes welfare harder because observed choices may be distorted, but the analyst does not automatically know which choices are mistakes. A worker who does not claim a benefit may be inattentive, confused, liquidity constrained, distrustful, stigma averse, or simply facing low net gains. A worker who stays at a retirement default may be helped by the default or passively locked into a poor fit.

A compact welfare object is:

Here $a_i^\star(p)$ is a benchmark or welfare-relevant action, and $\Psi_i$ maps the chosen action, the benchmark action, and the policy environment into welfare. The hard question is what $a_i^\star(p)$ should be. It might be the action under full information, the action under lower hassle, the action chosen by sophisticated workers, the action implied by a structural model, or the action a planner treats as welfare-relevant.

This is where normative ambiguity enters. Revealed preference is not enough if choices are distorted. But paternalism is not innocuous if the analyst misclassifies real preferences as mistakes. Behavioral welfare analysis therefore needs transparent assumptions: which preference is respected, which distortion is corrected, which workers are targeted, and what distributional weights matter.

Targeting adds another layer. Should policy target workers with the largest mistakes, the largest potential gains, the lowest administrative capacity, the highest earnings response, or the greatest equity concern? A reminder may be cheap but weak. A simplification reform may be broad but administratively expensive. A default may be powerful but normatively contested. A price change may be transparent but blunt if the constraint is knowledge or hassle.

Frontier Directions In Behavioral Labor Policy¶

The frontier is not “more nudges.” It is better integration of mechanism, implementation, equilibrium, and welfare.

First, researchers need sharper policy-channel designs. Haaland et al., 2023 emphasizes that information experiments should define the information object clearly. In labor policy, that means distinguishing belief updating from salience, procedural knowledge, eligibility confidence, perceived stigma, and learning speed.

Second, dynamic learning remains underbuilt. Many policies are repeated environments: filing taxes, claiming UI, interacting with disability rules, enrolling in training, and choosing retirement contributions. Designs that measure persistence, decay, diffusion, and repeated exposure are needed.

Third, digital administration changes the behavioral environment. Online portals can simplify claiming, but they can also create password, documentation, language, disability-access, and trust barriers. Digital design is not neutral implementation.

Fourth, intermediary response needs to be built into research designs. Policy effects may run through employers, tax preparers, counselors, caseworkers, and software vendors. A worker-only treatment effect may miss the channel that actually scales.

Fifth, welfare and targeting need more transparent frameworks. Bernheim & Taubinsky, 2018 and Allcott et al., 2022 are useful because they force the analyst to state when behavioral tools increase welfare and what assumptions support that conclusion. In labor policy, the welfare question must also include work, earnings, program access, human capital, firm response, and distribution.

Research Lab¶

The Week 9 lab is organized around Reproduce -> Diagnose -> Transfer.

Reproduce. The primary anchor is Bhargava & Manoli, 2015. Students use deterministic synthetic data to reproduce a compact benefit take-up and reminder design. The bounded path estimates treatment effects on eligibility understanding, application completion, claim timing, take-up, and earnings-linked program interaction. The goal is not an official replication. It is to practice treating implementation, salience, and procedural support as labor-policy margins.

Diagnose. Students classify the intervention using five questions: what friction is targeted, whether the policy corrects a mistake or harnesses a regularity, whether the main margin is incentives or take-up, what welfare benchmark is being used, and where intermediaries or equilibrium response may matter.

Transfer. The secondary anchor is Chetty et al., 2013. Students transfer the same diagnostic logic from take-up to EITC knowledge and local labor-supply response. The challenge branch uses Bernheim et al., 2015 for defaults and welfare or Kostøl & Myhre, 2021 for dynamic learning about tax-benefit schedules.

The bounded student path runs locally without confidential data. It trains students to move from policy tool to behavioral friction to econometric design to welfare benchmark.

Methods Box¶

Reading Ladder And References¶

Framing. Start with Chetty, 2015 for a pragmatic behavioral public-policy perspective and Bernheim & Taubinsky, 2018 for behavioral public economics and welfare questions.

Labor-supply incentives and knowledge. Use Chetty et al., 2013 for EITC knowledge and local labor-supply response, then Kostøl & Myhre, 2021 for learning the tax and benefit schedule.

Take-up and implementation. Read Bhargava & Manoli, 2015 as the primary take-up and psychological-frictions anchor. Pair it with Liebman & Luttmer, 2015 on Social Security information and claiming-related behavior, and Barr & Turner, 2018 on information, UI recipients, and postsecondary enrollment.

Defaults and work-linked saving. Use Madrian & Shea, 2001 for default enrollment and Bernheim et al., 2015 for welfare analysis of default options.

Methods and frontier design. Use Haaland et al., 2023 for information-provision design and Allcott et al., 2022 for welfare questions around nudges.

Exercises And Discussion Prompts¶

1. In equation (1), choose one labor policy and classify which component of $p = (\tau,n,s,d)$ is most likely to be binding.

2. In equation (2), give one example where $K_i(a;p)$ is mostly paperwork and one where it is mostly cognitive or trust-related.

3. In equation (3), explain how reminders, simplification, and assistance can raise $D_i(p)$ without changing statutory generosity.

4. In equation (4), describe one setting where short-run estimates may understate long-run policy effects because workers learn slowly.

5. Pick a default policy in retirement saving. What evidence would you need to decide whether the default corrects a mistake or imposes the planner’s preference?

6. For an EITC, UI, training, or disability setting, name one likely intermediary and explain how that actor could change policy incidence.

Reproducibility And Code Lab Note¶

The Week 9 lab uses deterministic synthetic data. It is a bounded teaching exercise, not an official replication of Bhargava & Manoli, 2015, Chetty et al., 2013, Bernheim et al., 2015, or Kostøl & Myhre, 2021. The local path is labs/09-behavioral-public-policy-in-labor-markets/. Students should run the smoke test, inspect the produced tables and figures, and write a short diagnosis memo using the Reproduce -> Diagnose -> Transfer structure.

Slide Companion Note¶

The Week 9 slide deck should not duplicate the chapter. It should define the central question, show the six-role taxonomy, contrast standard and behavioral policy design, isolate take-up and learning margins, and flag intermediary and equilibrium response.

Bridge Forward¶

Use this closing bridge to carry the module’s labor object, mechanism, and evidence into the next course step or research-design exercise.