Labor Supply Elasticity

The definition — three elasticities, one decision

A worker chooses how much to work. A higher wage shifts the trade-off. The question is: by how much does the worker actually change hours? "Elasticity" answers in percentage terms — what percent change in hours per percent change in wage. The point of labor-supply elasticity is that there is no single right answer. Three conceptual experiments correspond to three different elasticities, and they disagree systematically.

• Marshallian (uncompensated). Wage rises permanently; worker keeps the extra income. Hours change reflects both substitution (work pays more) and income (worker is richer, may want more leisure). The total response. Symbol: ε^M.
• Hicks (compensated). Wage rises but income is simultaneously adjusted to leave utility constant. The income effect is removed by construction. Only the substitution channel remains. ε^H.
• Frisch (λ-constant). Wage rises temporarily within a life-cycle plan; the marginal utility of wealth λ is held fixed. Relevant for short-run intertemporal substitution between periods. Symbol: ε^F.

For normal leisure, the algebraic ordering is ε^F ≥ ε^H ≥ ε^M. The Frisch is the biggest because it isolates pure intertemporal substitution. The Marshallian is the smallest because the income effect on leisure typically partially offsets the substitution. The Hicks is between.

The Slutsky equation makes the relationship precise:

∂L/∂w  =  ∂L^h/∂w  −  (T − L) · ∂L/∂Y
       (Marshallian) = (Hicks) − (income leak)

where T − L is hours worked. Substitution effect minus income effect (the working hours times the income elasticity of leisure) equals the total response. If the income effect is large enough, the total response can be negative — backward-bending labor supply.

What the empirics show

Modern labor economics has invested heavily in this parameter. Quasi-experimental designs — using EITC expansions, tax reforms, lottery winnings, marriage transitions, and Hartz reforms — have produced estimates that are converging across methods. The headline numbers, after thirty years of work:

Group / margin	Frisch / Hicks (intensive)	Extensive margin	Source examples
Prime-age men	0.1 - 0.3	0.2 - 0.7	MaCurdy 1981, Blundell-MaCurdy 1999, Chetty 2012
Married women	0.5 - 1.0	0.7 - 1.5	Mroz 1987, Eissa-Liebman 1996, Heim 2007
Single mothers	—	0.5 - 1.0	Meyer-Rosenbaum 2001, Eissa-Hoynes 2004
Older workers (60+)	0.3 - 0.7	0.6 - 1.5	Coile-Gruber 2007, French 2005
Self-employed / pros	0.3 - 1.0	—	Feldstein 1995, Saez-Slemrod-Giertz 2012
Macro consensus (Chetty)	≈ 0.5 (intensive)	≈ 0.25 (extensive)	Chetty et al. 2011, 2013

The single most influential synthesis is Chetty, Guren, Manoli, and Weber (2011, 2013), which combines dozens of micro and macro estimates with a structural framework and proposes a "macro labor-supply elasticity" of roughly 0.5 on the intensive margin and 0.25 on the extensive margin. Combined Hicks elasticity: around 0.75. This is the number now embedded in most central-bank policy models and many optimal-tax computations.

A worked numerical example — tax cut and hours response

Suppose a worker earns w = $30/hour and works L = 40 hours/week, for $1200 in pre-tax weekly earnings. The marginal tax rate is 35%, so the after-tax wage is $30 · 0.65 = $19.50/hour. Now the government cuts the marginal tax rate to 25%, raising the after-tax wage to $22.50/hour — a 15.4% wage increase.

Apply the Hicks elasticity of 0.3 (intensive margin, prime-age male calibration). The compensated hours response is 0.3 · 15.4% = 4.6%. So hours rise by 4.6% · 40 = 1.85 hours, to 41.85.

Apply the Marshallian elasticity of 0.05 instead. Hours response is 0.05 · 15.4% = 0.77%, or 0.31 hours. The total response is much smaller because the income effect — being richer, taking more leisure — claws back most of the substitution gain.

Apply Chetty's macro intensive-margin elasticity of 0.5. Hours rise by 0.5 · 15.4% = 7.7%, or 3.1 hours, to 43.1. The difference matters: $30 · 22 weeks · 3.1 hours = roughly $2050 extra earned income per year per worker. Aggregated to a labor force of 160 million, the macro difference is hundreds of billions of dollars in taxable income response — directly relevant to revenue forecasting and optimal-tax design.

Why the parameter matters — optimal taxation

The Saez (2001) and Diamond-Saez (2011) optimal top tax-rate formula:

τ*  =  1 / (1 + a · e)

where e is the elasticity of taxable income (effectively the compensated labor-supply elasticity for high earners) and a is the Pareto parameter of the income distribution (around 1.5 for the US top). With e = 0.25 (low-elasticity calibration), τ* ≈ 1/(1 + 1.5·0.25) = 73%. With e = 0.5, τ* ≈ 57%. With e = 1.0, τ* ≈ 40%. The optimal top rate is highly sensitive to the elasticity.

The Laffer curve peak — the revenue-maximising tax rate — depends on the same parameter. With e = 0.25, the Laffer peak is around 80%; with e = 0.5, around 67%; with e = 1.0, around 50%. The "Laffer rate" debate hinges almost entirely on which elasticity you believe. This is why thirty years of careful empirical work has been worthwhile despite the parameter sounding like an abstract curiosity.

The Hansen-Rogerson puzzle

Real business cycle macro models, developed by Kydland-Prescott (1982), Hansen (1985), and many successors, need a large Frisch elasticity — typically 2 to 4 — to match the observed volatility of employment over the business cycle. Micro estimates of the intensive-margin Frisch are 0.3 at best. This is the Hansen-Rogerson puzzle: a 5-10x gap between macro and micro.

Two reconciliations dominate the literature. First, indivisible labor (Hansen 1985, Rogerson 1988): in macro data, employment fluctuates mostly through the extensive margin — people moving in and out of work — not through hours per worker. Aggregate up extensive-margin and intensive-margin responses and you get something larger than the intensive-margin micro estimate. Second, Chetty et al. (2011, 2013): combining hours, participation, and life-cycle margins, plus adjustment costs, plus heterogeneity, the macro elasticity needed is closer to 0.75-1.0, not 2-4. RBC models with Frisch elasticities above 1 are now somewhat out of favour for that reason — Smets-Wouters' estimate is roughly 0.5.

The backward-bending curve and the Industrial Revolution

If income effects are large enough, the labor-supply curve bends backwards: higher wages reduce hours. Historical evidence is striking. In nineteenth-century Britain, hours per worker fell from around 65 per week to 55 as real wages roughly doubled. Across the twentieth century, hours per worker in advanced economies fell by another 25-40%, even as wages rose six- to ten-fold. The income effect plainly dominated over century-scale changes.

Cross-country at any given time: Europeans now work about 30% fewer hours per year than Americans at higher hourly wages — a backward-bending pattern at the country level. Prescott (2004) argued this gap was due to higher European marginal tax rates rather than income effects per se, but the substitution-vs-income-effect debate continues.

Within individual life cycles, peak earners often reduce hours at the top of the wage curve, then work less again in semi-retirement. The pattern is consistent with a Marshallian elasticity that is small and possibly negative at high wages.

When each elasticity is the right one

Question	Use which elasticity	Why
How will hours respond to a permanent tax cut?	Marshallian ε^M	Permanent shock, income effect is real
What's the welfare cost of a tax?	Hicks ε^H	Income effect is offset by transfer, deadweight loss runs through substitution only
How will hours respond to a transitory wage spike?	Frisch ε^F	Wealth effect is small for transitory shocks; intertemporal substitution dominates
Optimal income tax rate	Hicks ε^H (compensated, taxable-income version)	Saez formula uses compensated elasticity
DSGE macro model labor margin	Frisch ε^F	RBC and NK models linearise around steady state — pure intertemporal substitution
Life-cycle labor supply	Frisch ε^F	Holds λ constant across periods

Identification — the empirical challenge

Estimating labor-supply elasticity is hard because wages and hours are jointly determined. People who work more hours often command higher wages, and vice versa, so a naive regression of hours on wages confounds supply and demand. The literature has converged on three identification strategies.

• Tax-reform variation. Marginal tax rate changes provide arguably exogenous after-tax wage variation. Feldstein (1995) used the 1986 Tax Reform Act; Goolsbee (1999, 2000) used several reforms; Saez-Slemrod-Giertz (2012) summarise.
• EITC and welfare-reform experiments. Expansions of the Earned Income Tax Credit gave plausibly exogenous after-tax wage increases to specific demographic groups. Eissa-Liebman (1996), Meyer-Rosenbaum (2001) extracted large extensive-margin elasticities for single mothers.
• Lottery and inheritance shocks. Pure wealth shocks, unconfounded by wages. Imbens, Rubin, Sacerdote (2001) showed lottery winnings reduce labor supply, identifying the income effect cleanly. Cesarini et al. (2017) replicate in Swedish data.
• Bunching at kinks. Saez (2010) developed a method using bunching of taxpayers at marginal-rate kinks to identify elasticities directly. The approach has been widely applied; Chetty et al. (2011) use it for Denmark.

Limits and common pitfalls

• Margin matters. Intensive (hours per worker) and extensive (working or not) responses can be very different. Reporting one without the other can be misleading. Chetty's framework decomposes both.
• Heterogeneity is large. Average elasticities mask big differences across age, gender, occupation, family structure. Policy aimed at specific groups should use group-specific elasticities.
• Optimisation frictions. Workers don't continuously adjust hours. Job contracts often impose 40-hour weeks. Short-run elasticities can be artificially small for that reason — Chetty (2012) emphasises adjustment costs.
• Anticipation and persistence. A wage change anticipated to be temporary generates a different response than one perceived as permanent. Empirical identification needs to be matched to the experiment.
• Income measurement. The Slutsky decomposition needs separate identification of wage and non-wage income effects, which is hard without exogenous wealth variation.
• Macro vs micro mapping. Going from a cross-section regression to a representative-agent macro elasticity requires aggregation assumptions that are easy to violate.