Chetty and Saez 2009 — Teaching the Tax Code Earnings Responses to an Experiment with EITC Recipients

Summary

Chetty and Saez randomize Earned Income Tax Credit (EITC) schedule information to $43{,}002$ tax filers at H&R Block offices in Chicago (2007 tax season) and find no significant average treatment effect on earnings, but strong heterogeneity driven by whether the tax professional complied with the intervention. Complying professionals shift clients' earnings toward the first-kink EITC maximum (raising EITC amounts by $\$58$, $p < 0.01$), while non-complying professionals' clients shift earnings in the opposite direction ($+\$247$ into the phase-out range, $p < 0.05$). The results identify information frictions as a meaningful but individually-mediated barrier to intensive-margin labor supply optimization under the EITC.

Key Claims

• No average effect: Full-sample intent-to-treat (ITT) $= \$24$ EITC increase ($p < 0.1$, not significant); zero mean effect on earnings is consistent with prior non-experimental literature showing EITC has no intensive-margin effect.
• Complier effect: Clients of complying tax professionals earned $\$55$ more (wage-based) and received $\$58$ more in EITC ($p < 0.01$). Earnings concentrated near the first kink ($\$7{,}000$–$\$15{,}400$), consistent with optimization toward the EITC maximum.
• Non-complier effect: Clients of non-complying professionals earned $\$247$ more in year 2 ($p < 0.05$), shifted into the phase-out region — possibly reflecting strategic income inflation or misunderstanding of the phase-out.
• Diff-in-diff: Complier vs. non-complier gap $= +\$90$ EITC, $+\$420$ in earnings ($p < 0.01$); confirms that heterogeneity is driven by tax professional response, not baseline client differences.
• Calibration — compliers: A $\$58$ increase in EITC amounts is equivalent to a $33\%$ expansion of the EITC program itself; reachable via a $\approx 0.25$ hours/week labor supply increase.
• Calibration — non-compliers: A $\$247$ earnings increase attributable to a marginal tax cut would require reducing the implicit marginal tax rate by $5.4$ percentage points (pp) (using intensive elasticity $= 0.25$).
• Information scarcity: Survey evidence: $<5\%$ of EITC recipients know the nonlinear shape of the EITC schedule; recipients know their current refund but not counterfactual refunds at nearby income levels.
• Tax professional heterogeneity: Out of $1{,}461$ tax professionals in the study, the median professional had $30$ treatment clients; complier/non-complier status varies substantially. Tax professionals are a plausible conduit for information because they are the primary point of contact for low-income filers.
• Complier identification: Professionals classified via a leave-one-out procedure: complier if the fraction of their other clients with "middle income" ($\$7{,}000$–$\$15{,}400$) is higher in the treatment group than control. Identification is valid under the null of zero treatment effects (confirmed: no correlation with pre-treatment outcomes).

Concepts Introduced or Extended

• Earned Income Tax Credit
• Tax Salience
• Information Frictions
• Labor Supply Elasticity
• Optimal Social Insurance

Entities Mentioned

• Raj Chetty
• Emmanuel Saez

Quotes

"We find no significant average effect of the information treatment on earnings, but strong heterogeneous effects: tax professionals who complied with the experiment significantly increased the EITC amounts and earnings of their clients while non-complying professionals had the opposite effect."

"These findings suggest that the lack of intensive margin labor supply responses to the EITC documented in prior work could partly be driven by information frictions rather than solely by optimization or structural barriers."

My Take

The design elegantly exploits the H&R Block institutional context — tax professionals as a randomized information delivery channel — to get around the difficulty of reaching low-income filers directly. The complier/non-complier split is the paper's most important contribution: it explains why the average effect is zero while still identifying a mechanism. The main limitation is external validity: H&R Block clients in one metro area; tax professionals who actively deviate from the protocol (non-compliers) may be systematically different in ways that matter for policy. The calibration equivalences ($33\%$ EITC expansion, $5.4$ pp tax cut) are useful for policy comparison but depend on an assumed elasticity of $0.25$, which the authors acknowledge is uncertain.