Association for Public Policy Analysis & Management

In the 1990s and early 2000s, many U.S. metropolitan jurisdictions (e.g. cities, counties) enacted living wage (LW) ordinances to help lift working families out of poverty. Adopted policies mandated wage floors that were substantially (>50%) higher than prevailing minimum wages, and typically included provisions for wage updating (e.g. with inflation), health insurance coverage, and worker protection. Despite covering only a small percentage of low-wage workers (typically those employed by contractors and/or businesses receiving economic assistance), extant evidence suggests that these LW policies resulted in sizable declines in poverty and modest disemployment. Whether these economic effects translated into health improvements remains largely unknown. Analyzing the health effects of these policies can uniquely inform the potential of economic policy for improving the health and well-being of working families.

We analyzed data from four rounds (1996-2004) of the Community Tracking Study (CTS) on a representative sample of adults (18-64) from 46 metropolitan sites across the U.S. (n=126,863). LW policies were enacted in 13 sites, varying in effective dates, mandated wage floors, worker coverage, enforcement, insurance offering, and geographic region. CTS rounds provided baseline and follow-up data spanning 1-3 years before and after LW enactment. We exploited the potentially exogenous timing and location of LW policies in a difference-in-difference-in-differences (DDD) design, contrasting mean outcomes across time (pre/post-LW), sites (with LW vs. without), and within-site potentially treated/control groups defined by income/poverty level (treated: 150%FPL; control: 150-250%FPL). This follows prior evidence showing workers living in families just around the federal poverty line to be most affected by LW policies. In DDD models controlling for site/state- and individual-level covariates, year and site fixed effects, and site-specific trends, we analyzed LW policy effects on poor self-rated health status, feeling depressed, frequently smoking, and having unmet healthcare needs.

We observed significantly greater changes in the outcomes in the potentially treated poor/near-poor group relative to the lower-middle-income controls. Our difference-in-differences balancing tests also showed all covariates to be balanced on average across LW/non-LW sites. Overall, our DDD models estimated large but imprecise effects: a decline in poor health by 5.2 percentage points (pp) (standard error 3.7), an increase in smoking by 3.5 pp (SE 2.3), and little changes in depressive symptoms and unmet health care needs. The DDD models also estimated, consistently with prior economic evidence, a 7.9% (SE 5.0) increase in average real wages and little change in employment. These effects were stronger in sites with broader worker coverage and expansive implementation, where depressive symptoms also fell by about 4.4 pp (SE 2.8). Though imprecise, the changes we observed in poor health, smoking, and feeling depressed were nontrivial relative to their baseline levels (-20%, +10%, -37% respectively), and were consistent across different model specifications. However, these effects tapered off at 2+ years following policy introduction. Pending the final analyses and further robustness checks currently in progress, these preliminary findings provide support for the role urban economic policy could play as a modifiable determinant of population health and health equity.