Definition
Minimum wage spillovers refer to the empirical finding that minimum wage increases appear to raise wages not only at or below the statutory floor but also at higher percentiles of the wage distribution — percentiles where the minimum wage does not nominally bind. For example, if the minimum wage is at the 10th percentile but wages at the 15th or 20th percentile also rise following an increase, this suggests either a true economic spillover (employers adjust wages above the floor to maintain internal pay structures) or a statistical artifact of wage measurement error.
Key Ideas
- Two competing explanations for observed spillovers:
- True economic spillovers: Firms maintain internal wage hierarchies (relative fairness norms, efficiency wages); a floor increase compresses the bottom, prompting firms to raise wages at higher rungs to preserve differentials.
- Measurement error artifact: Current Population Survey (CPS) wages are reported in whole-dollar amounts, causing heaping and misclassification. If a true wage spike forms at the minimum, workers just above the statutory floor may be recorded as earning the minimum, and workers at the minimum may be recorded as earning slightly above. This makes the measured wage distribution respond to the minimum at percentiles above where it truly binds.
- Autor, Manning, and Smith (2016) measurement error model: The paper specifies that measured wages equal true wages with probability γ and equal true wages multiplied by a noise factor with probability 1−γ (roughly: ≈20% of wages are misreported, γ≈0.80). Under this model, the authors estimate the latent true wage distribution using maximum likelihood (ML) and show that a simple spike at the minimum in the true distribution (no spillovers) is sufficient to generate the observed apparent spillovers in the measured CPS data.
- Cannot reject zero spillovers: After accounting for measurement error, the paper cannot reject the null hypothesis that all apparent spillovers are measurement artifacts — the point estimate of the true spike at the minimum lies within the confidence interval of the measured mean effect at each percentile in almost all years.
How It Works
The ML approach estimates the distribution of true wages conditional on the observed heaped distribution, exploiting the structure of CPS rounding. If the true distribution has a spike (mass point) exactly at the statutory minimum but no mass at higher values, measurement error with γ≈0.80 will spread this spike across nearby percentiles in the measured data. The key test is whether the estimated true spike's mass is large enough to explain all measured effects above the nominal binding point. If so, no true spillovers are needed.
Why It Matters
- If spillovers are real, minimum wage increases have larger distributional effects than the mechanical binding-percentile effect, strengthening the equity case for higher minimums.
- If spillovers are entirely artifacts of measurement error, analysts who rely on observed wage effects at non-binding percentiles to estimate total minimum wage effects will systematically overstate them.
- The result highlights a broader methodological issue: wage heaping in survey data (CPS and others) can generate spurious distributional effects that are hard to distinguish from true effects without an explicit measurement error model.
Open Questions
- Does the zero-spillover finding hold for very large minimum wage increases (e.g., $15 minimum in high-cost cities), where the floor binds at higher percentiles and measurement error may play a smaller relative role?
- Are there other mechanisms (firm-level pay compression, union wage patterns) that could generate true spillovers detectable after measurement error correction?
- How does the misreporting rate γ vary across industries, firm sizes, or survey waves?
Related
Sources