Introduction

California’s Assembly Bill 1228 raised the minimum wage for fast-food chain workers — those employed by firms with 60 or more US locations — to $20 per hour effective April 1, 2024. At roughly 77 per cent of the state’s median hourly wage (a Kaitz ratio of 0.77), this is the highest minimum wage affecting fast-food workers in the United States (p.3). Independent restaurants and full-service establishments remained covered by California’s general $16 minimum, creating an unusually demanding test of minimum-wage effects as consumers can substitute toward uncovered competitors and employers face incentives to restructure operations.

Data and method

The study draws on two data sources: the Quarterly Census of Employment and Wages (QCEW) covering 2015Q1 through 2025Q3, and the LEHD Quarterly Workforce Indicators (QWI) through 2024Q4 (p.5). California is compared to 20 states where the binding minimum wage for fast-food employment is the federal floor of $7.25.

Two estimation strategies are applied: conventional difference-in-differences (DD) and synthetic difference-in-differences (SDID). SDID reweights control states and pre-treatment quarters to match California’s pre-treatment trajectory directly, constructing a synthetic control primarily from Idaho (61%), Utah (22%), Pennsylvania (7%), Texas (7%), and Georgia (4%) (p.10).

Results

Wages rose substantially: +7.5 to +7.6 per cent in conventional estimates and +6.4 to +7.5 per cent in synthetic variants, all highly significant (FP p < 0.001). The baseline DD yields an employment effect of −1.4 per cent (OWE = −0.19), while the preferred SDID yields −0.3 per cent (p = 0.50, OWE = −0.04) (p.11).

QWI data yielded more positive employment estimates: −0.6 to +1.7 per cent across 16 specifications, with 12 of 16 non-negative, and an implied OWE ranging from −0.10 to +0.26 (p.11).

Additional evidence from QWI: Firm-size heterogeneity and worker flows

Wage effects follow a sharp and fairly monotone gradient by firm size: +1.1 to +2.4 per cent in the smallest firm-size bin (where AB 1228 does not bind) rising to +10.4 to +11.8 per cent in the 500+ employee bin most clearly covered by the policy (p.13). This gradient provides strong evidence that estimated effects reflect the policy’s coverage structure rather than a generic California-specific post-2024 shock.

The quarterly separation rate fell sharply by 13 to 25 log points across 16 QWI specifications — a roughly 3–6 percentage-point absolute decline against baseline quarterly rates of 25–30 per cent (p.14). The own-wage elasticity of separations ranges from −1.7 to −4.2, roughly 1.5 to 4 times the restaurant-sector benchmark in Dube et al. (2016) (p.14). This is consistent with a monopsonistic quit-reduction channel in which higher wages reduce turnover and mute the employment response.

The fall in separations also helps reconcile divergent QCEW and QWI employment estimates. Because QCEW counts every worker paid during the pay period containing the 12th of the month, falling within-period turnover mechanically shrinks the QCEW headcount even when the true employment stock is unchanged (p.16).

Interpretation and context

Across 32 QCEW and QWI specifications, the OWE ranges from −0.29 to +0.26 — entirely within Dube and Lindner (2024)’s “small” classification (p.14). For context, across 27 post-2010 state minimum-wage increases, the QCEW long-difference median OWE is −0.02 (IQR [−0.34, +1.25]), and across 28 US restaurant-sector studies the median OWE is −0.15 (p.16). Even at the most negative end of the range, total fast-food earnings rose as the wage gain substantially exceeded the employment decline (p.16).

Conclusion

California’s $20 fast-food minimum wage raised wages by roughly 7 per cent with modest to no employment costs. The QWI employment response sits near zero — close to the median across post-2010 state minimum-wage events — despite a Kaitz of roughly 0.77, well above the 0.42–0.62 range spanned by past state events (p.18). The policy’s footprint maps onto its statutory coverage, with limited cross-employer wage spillovers. The paper also notes that synthetic-control–based methods are particularly well suited to single-treated-unit settings where idiosyncratic pre-trends cannot be averaged out across multiple treated units (p.18).