Autor Manning and Smith 2016 — The Contribution of the Minimum Wage to US Wage Inequality over Three Decades

Summary

This paper reassesses Lee (1999)'s influential ordinary least squares (OLS) finding that the minimum wage explains virtually all of the growth in lower-tail US wage inequality during the 1980s. Using an instrumental variables (IV) strategy that instruments the effective minimum (log minimum minus log state median) with the statutory minimum and its interactions with average state median wages, the authors show Lee's OLS estimate is severely upward-biased due to two sources: transitory shock bias (state median appears on both sides of the regression) and omitted variable bias (high-wage states have more latent inequality but lower effective minima). The corrected two-stage least squares (2SLS) estimates imply the minimum wage explains $30$ – $55\%$ of female lower-tail inequality growth in the 1980s — substantial but far less than Lee's $>100\%$ . The paper also shows that apparent minimum wage spillovers (effects above the binding percentile) are consistent with being entirely artifacts of Current Population Survey (CPS) wage measurement error ( $\approx 20\%$ misreporting), requiring no appeal to true economic spillovers.

Key Claims

Lee (1999)'s OLS estimate of the minimum wage's contribution to lower-tail inequality is severely upward-biased due to two distinct identification problems: transitory shock bias and omitted variable bias from state-level latent inequality differences.
Instrumenting the effective minimum with statutory minimum, its square, and statutory minimum $\times$ average state median yields very strong first stages ( $F$ -stats $251$ – $684$ ) and purges both biases.
The minimum wage explains approximately $11.3$ – $15.1$ log points of the $24.6$ log-point rise in female 50/10 inequality during the 1980s ( $40$ – $55\%$ ); male contribution is negligible ( $\approx 1.2$ – $1.4$ log points of $2.5$ ).
Over the full 1979–2012 period, the minimum wage explains $14.8$ – $18.6$ log points of female lower-tail inequality growth; effects are more similar across sexes in later decades.
Apparent spillovers of the minimum wage to percentiles above the binding level are statistically consistent with being entirely driven by CPS measurement error ( $\gamma \approx 0.80$ , $\approx 20\%$ misreporting); the paper cannot reject the null that true economic spillovers are zero.
Lee's sample (1979–1989) had insufficient state-level variation in statutory minima — only one state above the federal floor from 1979–1985 — making IV infeasible; post-1991 variation (up to $31$ states above federal level by 2008) drives the IV identification.

Concepts Introduced or Extended

Entities Mentioned

Quotes

"Lee's estimates would imply that absent the minimum wage, the rise in female lower-tail inequality between 1979 and 1988 would have been only $2.9$ to $4.3$ log points rather than the $24.6$ log points actually observed — a result that seems implausible to us."

"While the minimum wage clearly matters for lower-tail inequality, it does not appear to be the dominant force that Lee's estimates imply."

"We cannot reject the null hypothesis that all apparent spillovers are measurement artifacts."

My Take

This paper is a careful and credible corrective to Lee (1999). The two-bias decomposition (transitory shock + omitted variable) is pedagogically clean and the IV strategy is well-motivated. The measurement error treatment of spillovers is innovative — using maximum likelihood (ML) estimation to distinguish a latent-spike model from true spillover models — though the identifying assumptions (constant misreporting rate $\gamma$ ) are maintained rather than tested. The finding that $\approx 40$ – $55\%$ of female lower-tail inequality growth in the 1980s is attributable to the minimum wage is still a large number and policy-relevant. The negligible male contribution remains unexplained mechanically and warrants further research.