Low-wage-sector employment barely moves after a state raises its minimum wage — a ~14% average hike is followed by a statistically insignificant 0–3% employment change over the next five years

Summary

Across the 24 U.S. states that enacted a discrete minimum-wage increase of at least $0.75/hour (averaging +$1.10, or +14%) between 2014 and 2020, employment in the two most minimum-wage-exposed sectors — food service (NAICS 722) and retail trade (NAICS 44-45) — did not fall by a statistically detectable amount relative to 14 states that kept their minimum wage pinned to the $7.25 federal floor the entire period. A two-way fixed-effects panel across all 51 jurisdictions, 2010-2024, puts the food-service employment elasticity with respect to the state minimum wage at -0.03 (SE 0.058, p=0.56) and the retail elasticity at -0.06 (SE 0.049, p=0.24) — both statistically indistinguishable from zero. An event-study design (never-raised states as controls) shows employment drifting from roughly flat in the hike year to about -2.7% five years out and -4.9% six-plus years out relative to the year before the hike — small, trending negative, but not statistically significant at any horizon (all p>0.10), and pre-trends are flat (p=0.32), so there is no visible pre-existing divergence between the treated and control states. Wages in the same sector rose sharply and reliably with the minimum wage (elasticity +0.15, p<0.001), confirming the mechanism worked as intended. This matches the weight of the published literature: Cengiz & Dube (QJE 2019) found essentially no aggregate job loss over five years following 138 state minimum-wage increases 1979-2016, and Dube & Zipperer's 2024 synthesis of 72 published studies puts the median own-wage employment elasticity at -0.13 (about 13% of potential wage gains offset by job loss) — this analysis's point estimates fall within, and mostly below, that range.

What the data show

Design. BLS QCEW (econ.state_wages) supplies annual private-sector employment and average weekly wage by state and 3-digit NAICS industry, 2008-2025 (measured complete, no interior gaps). DOL's state minimum-wage history (fiscal.state_minimum_wage_history) supplies the state minimum wage by year, 2000-2026 declared. Two designs were run:

• Continuous panel: log(employment) regressed on log(state minimum wage) with state and year fixed effects (absorbing both national trends and fixed state characteristics), clustered by state, 2010-2024, 44 states with usable single-value minimum-wage data.
• Event study: states were split into 24 "treated" states (first hike of $0.75+ during 2014-2020) and 14 "never-raised" control states whose minimum wage stayed flat at the federal floor ($7.25, or Georgia/Wyoming's nominal $5.15) the entire 2013-2024 span, with year and state fixed effects and coefficients measured relative to the year before each state's own hike.

Both designs point the same direction: a small, not-statistically-significant negative employment response in the sectors where minimum-wage workers are concentrated, alongside a clear, statistically significant rise in measured average pay in those same sectors.

An elasticity of -0.03 to -0.06 means: for every 10% the state minimum wage rises, low-wage-sector employment moves roughly 0.3%-0.6% lower — a magnitude too small to distinguish from zero with 44-51 states of data, and smaller than the -0.13 median elasticity Dube & Zipperer (2024) find pooling 72 published studies.

Caveats and what this analysis adds vs. what it reproduces

Reproduces: the qualitative finding of the Cengiz-Dube QJE 2019 "bunching" design (no detectable aggregate job loss following 138 state minimum-wage changes 1979-2016) and sits within the Dube-Zipperer (NBER WP32925, Sept 2024) meta-analytic range (median OWE -0.13 across 72 published studies; "estimates published since 2010 tend to be closer to zero" — this analysis's window is entirely post-2010, which is consistent with finding smaller-than-median magnitudes).

Extends: this analysis measures whole-sector QCEW employment counts (a proxy for "low-wage jobs" via the sectors where minimum-wage workers concentrate) rather than Cengiz-Dube's wage-bin "missing jobs" design built from individual wage microdata, which this corpus does not carry. It also runs both a continuous-treatment panel and a discrete event-study with an explicit pre-trend test — most public discussion of this question cites only a single point estimate.

Improves on a raw before/after comparison: a naive OLS of log(employment) on log(minimum wage) with no state/year fixed effects returns a strongly positive coefficient (+0.45, p=0.01) purely because larger, higher-cost-of-living states both pay higher minimum wages and have larger nominal low-wage-sector payrolls — the uncontrolled-confound warning this connector raised on that regression is correct, and the two-way fixed-effects design above (which nets out state size and national trends) is what should be read as the answer, not that raw correlation.

Real-world magnitude check: the average state hike studied here (+14%, +$1.10/hr) is smaller than California's single-sector 2024 fast-food minimum wage, which rose 25% ($16 to $20/hr) for chains with 60+ locations. Hamdi & Sovich (2025, fetched directly) find that even at that much larger dose, fast-food firms hired somewhat fewer new workers, but the slowdown did not outpace a simultaneous drop in turnover, so measured employment rose slightly — and they find no employment or wage spillovers into other low-wage sectors. That is directionally consistent with this analysis's near-zero estimates, at roughly double the wage shock.

Limitations disclosed, not resolved: (1) only 37-44 states enter each regression (small-n, wide confidence intervals — the honest reading of every non-significant coefficient above is "too imprecise to rule out a modest effect in either direction," not "proven zero"). (2) The event study's cluster-robust SEs are flagged by the tool itself as unreliable below ~40 clusters, biasing toward overstated precision — the true uncertainty band is probably wider than shown. (3) Adoption is staggered across 6 distinct years, which can bias two-way fixed-effects estimates (Goodman-Bacon 2021); restricting controls to states that never raised their own minimum wage (rather than using already-treated states as controls, the specific failure mode Goodman-Bacon identifies) mitigates this but the tool's generic warning still applies. (4) Whole-sector QCEW employment blends minimum-wage workers with higher-paid managers/staff in the same NAICS code, diluting any true effect toward zero relative to Cengiz-Dube's individual-wage-bin design.