Buszkiewicz et al. Respond to “Methods to Estimate Minimum Wage Health Effects”

We thank Leigh (1) for his thoughtful comments on our study evaluating the relationship between state minimum wage increases and health in working-age adults using 2008–2015 restricted-use National Health Interview Survey data (2). We largely concur with his points about common limitations shared by this study and others in this field, including cross-sectional data, short observation periods, and imperfect treatment and control group definitions. We also agree that rigorous methods are needed to identify the likely treated group with respect to minimum wage policies. As we acknowledge, using education as an identification strategy likely misclassifies some unexposed individuals as exposed. Leigh suggests instead identifying workers with low wages as the treatment group (3). Future studies must grapple with the costs and benefits to using a characteristic that is affected by the exposure to identify the treatment group. Individual wages are on the pathway between minimum wage and health, so restriction based on this variable could potentially violate assumptions of strict exogeneity and bias estimates (4). For this reason, even when a data source offers a good measure of hourly wage, it might not be the best approach to identifying the treatment group in a minimum wage study using the difference-in-differences approach.

Leigh also critiqued a variety of sample inclusions in our study. We agree that the inclusion of the self-employed might attenuate estimates. Beyond that, we would caution against too narrowly defining those likely affected by minimum wages. While minimum wage workers are the primary target of this policy, there are theoretical effects on 2 other groups: 1) the unemployed who are looking for work and might be affected by changes in the availability of jobs or hours; and 2) those who make more than the minimum wage and might be affected by wage compression or efforts to avoid it (5). Our inclusion of the unemployed was not done to capture whether people became unemployed because of the minimum wage but to evaluate potential consequences of higher minimum wages on the demand for less-educated workers. The increased cost of workers could theoretically change an employer’s hiring decisions. In Seattle, Washington (6), researchers found that the benefits of higher minimum wages accrued primarily to workers who were steadily employed prior to the law passing, while less stably employed workers saw no such benefit. Consistent with this approach, studies have shown separate health consequences of minimum wage in the unemployed (7–9). Seattle researchers (10) also found a shift in the wage distribution up to $20/hour, suggesting that defining the treatment group at the exact value of the minimum wage could lead to misclassification. Finally, while we agree that restricting to full-year, full-time workers who are most likely to receive consistent wages might produce “more salubrious effects,” we argue this is not the reality of most low-wage employment. To understand the average effects of the minimum wage, it is essential to consider that many workers earning low wages have unstable employment and erratic schedules (11, 12).

We echo Leigh’s call for future work to focus on longitudinal analyses, longer-term consequences of minimum wage on health, and improved identification strategies. In addition, while it was beyond our scope to examine the pathways between minimum wage and health, this is an essential question for future study. We think it is plausible that those pathways include not only employment, earnings, and income but also stress, job satisfaction, and subjective financial well-being. We further advocate for a greater focus on heterogeneous treatment consequences of the minimum wage on health (13–15) as well as the implementation of advanced epidemiologic methodologies to address time-varying confounding and effect measure modification (8, 16, 17).