Effect of Expanding the Earned Income Tax Credit to Americans Without Dependent Children on Psychological Distress

Abstract

Anti-poverty policies have the potential to improve mental health. We conducted a randomized trial to investigate whether a fourfold increase in the Earned Income Tax Credit for low-income Americans without dependent children would reduce psychological distress relative to the current federal credit (Paycheck Plus, New York City site). Between 2013 and 2014, 5,968 participants were recruited; 2,997 were randomly assigned to the treatment group and 2,971 were assigned to the control group. Survey data were collected 32 months post-randomization (N=4,749). Eligibility for the program increased employment by 1.9 percentage points and after-bonus earnings by 6% ($635 per year) on average over the three years. Treatment was associated with a marginally statistically-significant decline in psychological distress relative to the control group (-0.30 points; 95% CI, -0.63 to 0.03; p=0.076). Women in the treated group experienced a half-a-point reduction in psychological distress (-0.55; 95% CI, -0.97 to -0.13; p=0.032) and noncustodial parents reported a 1.36 point reduction (95% CI, -2.24 to -0.49; p = 0.011) in psychological distress. An expansion of a large anti-poverty program to individuals without dependent children reduced psychological distress for women and noncustodial parents – the groups who benefitted the most in terms of increased after-bonus earnings.

Invited Commentary: Conducting and Emulating Trials to Study Effects of Social Interventions

“…The degree of nonadherence also means the ITT effect is unlikely to generalize to the other sites in the study, which all had perfect adherence. This sentiment also applies to the study by Courtin et al ( 1 ), in which the paperwork involved in receiving the tax credit within the study differed from the way the Earned Income Tax Credit is implemented outside the trial setting. In this particular case, the per-protocol effect may arguably be more transportable to other populations or contexts, depending on whether the degree and type of nonadherence would be comparable across other settings.…”

“…With this in mind, we outline some key challenges and opportunities to consider when conducting and emulating such trials, with special consideration of how to make the most of all such feasible studies. Our comments are grounded in Courtin et al’s accompanying study ( 1 ) of the effect of randomization to a 4-fold increase in the Earned Income Tax Credit for low-income Americans without dependent children on psychological distress over 3 years of follow-up. Though our primary points apply more generally beyond social interventions, these core issues are worth underscoring because of their relevance and prevalence in answering social epidemiologic questions.…”

“…Consider the intention-to-treat (ITT) effect: the effect of being randomized to the social policy (e.g., randomized to be eligible for the expanded tax credit program). In settings with incomplete follow-up (such as the 69% who had outcome information at 3 years in the Courtin et al study ( 1 )), even estimating the ITT effect requires strong assumptions about the reasons why and how people are lost to follow-up. Selection bias due to loss to follow-up can be mitigated if we design our trials with this in mind by collecting baseline and longitudinal data on measures suspected to be related to remaining in the study.…”

“…Estimating the ITT or per-protocol effect in a trial can often raise immediate questions about several adjacent treatment strategies. For example, Courtin et al estimated the effect of randomization to eligibility for an expanded tax credit of a specified amount received annually ( 1 ), but their results raise several subsequent questions (some of which were also raised by the authors themselves): What is the effect of different amounts of tax credits at different intervals; of providing the tax credit alongside guidance or instruction for possible use; of offering the tax credit but not requiring onerous paperwork or working with a particular organization in order to obtain it? These “adjacent” questions can be viewed as novel but related interventions, new joint interventions (possibly evoking the eventual implementation elements ( 13 )), and/or refining multiple versions of treatment into a sufficiently well-defined (or improved) intervention ( 14 ).…”

“…Broadly speaking, interference means that the intervention in one individual may affect other individuals’ adherence and outcomes. For example, for noncustodial parents (one of the subgroups studied by Courtin et al ( 1 )), the effect of receiving a tax credit may be affected by whether or not the custodial parents receive a tax credit. Interference implies there could be many questions of interest, such as direct effects of the intervention, spillover effects, and overall effects ( 17 , 18 ).…”