The Supplemental Nutrition Assistance Program (SNAP) is the only universal component of the U.S. safety net. In 2019, the U.S. Government spent over 60 billion dollars to provide nutrition benefits to more than 36 million U.S. citizens. Despite being the largest means-tested Federal program, the literature relating SNAP to labor supply is extremely sparse (particularly in the modern era) and there is hardly any evidence about the intensive margin. SNAP program formulas are designed to reduce labor supply distortions with benefits that gradually fade out as earned income increases, but whether these intensive-margin distortions occur in practice is an empirical question that has important policy implications for the SNAP program and for work requirements in the safety net more broadly.
Studying SNAP is particularly challenging because income eligibility is often determined at the monthly level, which is not collected in most surveys and because other factors (e.g., dependent care and excess shelter cost disregards)—which are rarely measured and are likely subject to greater measurement error than income—impact eligibility. To overcome this obstacle, this study utilizes new administrative records containing every SNAP recipient for two States: Colorado from 2012-2013 and Oregon from 2008-2016. Importantly, for these States, the researchers can observe gross and net income—after removing SNAP earnings disregards and other adjustments—allowing them to precisely place recipients along program formulas.
This project assesses how the modern program rules of SNAP influence labor supply decisions. The project tests whether the point where benefits are initially taxed (i.e., where net income equals zero) creates labor supply distortions along the intensive margin. SNAP benefits are a function of net income, i.e., the gross income of an assistance unit less a complicated set of income disregards. A household's gross income includes all earnings from work, as well as some other sources such as disability payments and child support. Net income is then calculated by subtracting the following deductions: The standard deduction, earnings deduction of 20 percent of any earnings, as well as deductions for dependent care, child support, medical, and shelter costs. SNAP benefits are initially taxed away by 30 cents for every additional dollar of income at the point where net income equals zero. This kink point is the focus of the study.
The study tests for intensive margin labor supply distortions by exploring whether there is an excess density (mass) of SNAP cases at the zero net income kink in program benefits, using well-known bunching estimators utilized in other contexts. Intuitively, the research design approximates the shape of the counterfactual bunching region by interpolating the shape of the surrounding areas with a seventh-degree polynomial.
In the aggregate, neither State shows evidence of bunching at zero net income. However, SNAP units with fewer recipients, particularly single-person units, may have fewer labor market frictions and may exhibit different bunching behavior as a result. The study explores this possibility and finds that, in Colorado, there appears to be an excess mass around the kink point for single-person cases.
The Colorado data uniquely feature information on self-employment income. This is important because bunching behavior in other settings is typically strongest amongst self-employed individuals. Cases with one recipient and some self-employment income exhibit bunching at the zero net income kink that is not visible in the cases without self-employment income. The researchers estimate a statistically significant excess density of 0.002, or 3,470 cases in the $400 dollar region surrounding the kink. However, even though they can detect bunching for this group, the effect size is economically quite small. For context, the excess mass represents less than 0.3 percent of cases in the Colorado sample.
Finally, because benefit reduction rates vary by shelter deduction use, the study tests for bunching heterogeneity by shelter-deduction levels. The distributions show little heterogeneity in bunching along this dimension.
Across both States, this study finds little evidence of labor supply distortions where benefits are initially taxed. The one exception is that the researchers observe an excess mass for single-unit cases with self-employment income in Colorado. However, the magnitude of the bunching is small in relation to total state caseloads. Thus, the study concludes that labor supply distortions from the increase in the benefit tax rate are not a first-order concern.