Shapiro–Stiglitz Theory

Shapiro–Stiglitz Theory is a foundational microeconomic and macroeconomic efficiency-wage model developed by Carl Shapiro and Joseph E. Stiglitz in their 1984 American Economic Review paper, "Equilibrium Unemployment as a Worker Discipline Device." The theory explains why real wages fail to fall to clear the labor market even in the absence of minimum wage laws, labor unions, or search frictions, and why involuntary unemployment persists as an equilibrium feature of decentralized competitive markets. By embedding a dynamic principal–agent problem with hidden action (moral hazard) into an aggregate labor market, Shapiro and Stiglitz demonstrate that when employers cannot monitor worker effort costlessly, they must pay an efficiency wage above the market-clearing level—and the economy must maintain a positive pool of unemployed workers—so that the threat of dismissal imposes a credible financial penalty on shirking.

In a textbook neoclassical labor market without informational frictions, labour is treated like any homogeneous physical commodity: worker effort is contractually observable and enforceable, and if labor supply exceeds labor demand at the prevailing wage, unemployed workers bid down the real wage until every worker willing to work at the equilibrium wage is employed. Under such conditions, firing an employee imposes zero net penalty on the worker, because the dismissed individual can instantaneously secure an identical job at the same market-clearing wage w*.

Shapiro and Stiglitz observe that if finding an equivalent job is rapid and costless, and if a worker's on-the-job effort is only imperfectly observable by management, a fundamental incentive problem arises. When employment carries no rent above the outside option, a rational worker who dislikes exerting effort will choose to shirk (withhold effort on the job), knowing that in the worst-case scenario—being caught and dismissed—they will promptly be rehired elsewhere at the identical wage. Furthermore, standard bonding solutions (requiring workers to post large upfront performance bonds that are forfeited upon dismissal) are generally infeasible due to worker wealth constraints and the reverse moral hazard temptation for unscrupulous firms to falsely accuse employees of shirking simply to confiscate their bonds.

To induce positive effort without performance bonds, each firm must pay a wage w high enough that the present discounted value of remaining employed exceeds the present discounted value of being dismissed into the unemployment pool. When all firms in the economy simultaneously raise their wages to motivate their workforces, aggregate labor demand falls below aggregate labor supply, creating equilibrium involuntary unemployment. This resulting unemployment pool serves a social and economic function: by lowering the job-finding rate for dismissed workers, aggregate unemployment turns job loss into a costly penalty that restores workplace discipline across the economy.

Consider a continuous-time economy populated by N identical, infinitely lived, risk-neutral workers and M competitive firms. All agents discount future utility streams at the real interest rate r > 0. An employed worker receives a real flow wage w and chooses whether to exert full work effort e > 0 or to shirk by exerting zero effort (e = 0),yielding instantaneous flow utility:

u(w, e) = w − e

Unemployed workers exert zero effort and receive a flow unemployment benefit (or value of leisure/home production) equal to w₀ ≥ 0.

Employment relationships are subject to two distinct hazard rates:

1. Exogenous Separation Rate (b > 0): All active jobs face an exogenous Poisson rate b of termination due to firm-specific demand shifts, relocation, or plant closures, regardless of worker effort.
2. Monitoring Detection Rate (q > 0): Firms monitor their workforce imperfectly. A worker who exerts full effort (e > 0) is never fired for cause, whereas a worker who shirks (e = 0) is detected and dismissed at an exogenous Poisson arrival rate q > 0.

Unemployed workers find new jobs at an endogenous Poisson job-acquisition rate a > 0, which is taken as given by individual workers and firms but determined in aggregate steady-state equilibrium. Because all firms are identical and cannot observe whether a newly hired applicant was previously laid off exogenously or fired for shirking, all unemployed workers face the same re-employment hazard a.

Let V_E^N denote the expected lifetime utility (asset value) of an employed non-shirker, V_E^S the asset value of an employed shirker, and V_U the asset value of an unemployed worker. The continuous-time Hamilton–Jacobi–Bellman (HJB) asset equations are:

rV_E^N = (w − e) + b(V_U − V_E^N)

rV_E^S = w + (b + q)(V_U − V_E^S)

rV_U = w₀ + a(V_E − V_U)

where V_E = max(V_E^N, V_E^S) represents the equilibrium value of employment. Rearranging the first two asset equations expresses the lifetime values of non-shirking and shirking explicitly in terms of the unemployment value V_U:

V_E^N = [(w − e) + bV_U] / (r + b)

V_E^S = [w + (b + q)V_U] / (r + b + q)

Because a shirking worker produces zero output, every firm designs its compensation contract so that workers choose to exert full effort. A worker chooses not to shirk if and only if the lifetime utility of working honestly is at least as large as the lifetime utility of shirking:

V_E^N ≥ V_E^S

Substituting the two HJB asset expressions into this incentive-compatibility inequality and solving reveals that the net employment rent (V_E^N − V_U) must satisfy:

V_E^N − V_U ≥ e / q

This inequality captures the core economic intuition of the model: the expected utility loss from being fired (V_E^N − V_U) multiplied by the probability per unit time of being caught shirking (q) must at least equal the flow disutility of effort (e). If either the detection rate q approaches zero or the employment rent (V_E^N − V_U) vanishes, shirking cannot be deterred.

In an incentive-compatible equilibrium where workers do not shirk (V_E = V_E^N), subtracting rV_U from rV_E^N yields the equilibrium employment rent in terms of wages and transition rates:

V_E^N − V_U = (w − e − w₀) / (r + b + a)

Combining this expression with the incentive constraint V_E^N − V_U ≥ e / q produces the worker-level No-Shirking Condition (NSC):

w ≥ w^NSC = w₀ + e + (e / q)(r + b + a)

To express the No-Shirking Condition in terms of aggregate employment L and the unemployment rate u = (N − L) / N, note that in a steady state where no workers shirk, the flow of workers losing jobs due to exogenous separations (bL) must equal the flow of unemployed workers finding jobs (a(N − L)):

bL = a(N − L) ⇒ a = bL / (N − L) = b(1 − u) / u

Substituting a = bL / (N − L) into the wage inequality yields the aggregate No-Shirking Condition in (L, w) space:

w ≥ w₀ + e + (e / q)[r + bN / (N − L)] = w₀ + e + (e / q)[r + b / u]

To close the model, suppose each of the M identical firms operates a concave production function f(L_i) when its employees exert full effort, giving an aggregate production function F(L) = Mf(L / M) with F′(L) > 0 and F″(L) < 0. Profit-maximizing firms hire workers up to the point where the marginal product of effective labor equals the efficiency wage:

w = F′(L)

This downward-sloping aggregate labor demand curve intersects the upward-sloping aggregate No-Shirking Condition at the unique Nash equilibrium (L*, w*):

F′(L*) = w₀ + e + (e / q)[r + bN / (N − L*)]

Several foundational properties emerge directly from this equilibrium condition:

1. Impossibility of Full Employment (u* > 0): As aggregate employment L approaches the total labor force N (so that the unemployment rate u → 0), the steady-state job-finding rate a = bL / (N − L) approaches infinity. An unemployed worker would find a new job instantaneously, eliminating any penalty from being fired and driving the required no-shirking wage w^NSC to infinity (a vertical asymptote at L = N). Because the marginal product of labor F′(N) is finite, the downward-sloping labor demand curve F′(L) must intersect the NSC to the left of N (L* < N).
2. Involuntary Nature of Unemployment: At the equilibrium wage w*, employed workers enjoy a positive utility surplus over unemployed workers (V_E^N − V_U = e / q > 0). Unemployed workers are identical to employed workers and prefer to work at w* (or even at a slightly lower wage w − δ), yet no firm will hire them at a lower wage because any worker hired below w^NSC would find it optimal to shirk. Thus, unemployment is genuinely involuntary.
3. Inefficiency of the Decentralized Equilibrium: The laissez-faire equilibrium is not constrained Pareto efficient. First, each firm fails to internalize the pecuniary externality that hiring an additional worker raises the aggregate job-finding rate a, which shifts the NSC upward for all other firms. Second, because w* > e, the marginal social product of labor F′(L*) exceeds the marginal social opportunity cost of work e, meaning aggregate employment is inefficiently low. Wage subsidies financed by lump-sum taxes or profit taxes can improve social welfare.

The Shapiro–Stiglitz framework provides a microfounded resolution to several classical puzzles in macroeconomics and labor economics:

• Cyclical Wage Rigidity and Employment Fluctuations: When an adverse aggregate productivity shock shifts the labor demand curve F′(L) inward, real wages cannot fall freely to preserve full employment because doing so would violate the No-Shirking Condition. Instead, adjustments occur substantially through quantities (layoffs and higher unemployment u*), amplifying business-cycle employment fluctuations.
• Dual Labor Markets and Inter-Industry Wage Differentials: Extending the model to two sectors—a primary sector where output quality is sensitive to effort and monitoring is difficult (low q), and a secondary sector where output is easily metered via piece rates (high q → ∞)—explains why observationally identical workers earn persistent wage premia in capital-intensive primary industries while queuing for primary-sector jobs.
• Personnel Economics and Efficiency-Wage Testing: Empirical studies of corporate compensation, such as Alan Krueger's (1991) comparison of company-owned versus franchised fast-food restaurants and Daniel Raff and Lawrence Summers's (1987) analysis of Henry Ford's 1914 five-dollar-day wage doubling, confirm the model's core predictions: where direct owner-manager supervision is weaker (low q), firms pay significantly higher wages, reducing turnover and absenteeism while raising productivity.

Why can unemployed workers not bid down the wage to get a job in the Shapiro–Stiglitz model?

Even if an unemployed worker offers to work for a wage slightly below the prevailing efficiency wage w*, a rational employer will refuse the offer. The employer knows that once hired at a wage below the No-Shirking Condition w^NSC, the value of keeping that job is too low relative to the unemployment pool, making it optimal for the worker to shirk rather than exert effort e.

Why does the No-Shirking Condition have a vertical asymptote at full employment?

At full employment (L = N, or u = 0), there is zero wait time in the unemployment pool (a → ∞). Because a worker fired for shirking would be rehired by another firm instantaneously with zero income loss, no finite wage could ever deter shirking. Consequently, w^NSC approaches infinity as L → N, guaranteeing that equilibrium unemployment is positive.

Could upward-sloping age-earnings profiles or performance bonds replace unemployment as a discipline device?

In principle, Edward Lazear (1979, 1981) showed that if workers post upfront performance bonds or accept deferred compensation (underpayment early in tenure followed by overpayment and pensions later), the threat of losing the bond or deferred pension can deter shirking without aggregate unemployment. However, Shapiro and Stiglitz note that capital market imperfections (liquidity constraints on young workers) and the moral hazard risk of firms firing older workers just before pension vesting limit the complete elimination of efficiency wages in practice.

1. Shapiro, C., & Stiglitz, J. E. (1984). Equilibrium unemployment as a worker discipline device. The American Economic Review, 74(3), 433–444. https://doi.org/10.7916/D89P3b0H
2. Yellen, J. L. (1984). Efficiency wage models of unemployment. The American Economic Review, 74(2), 200–205. https://doi.org/10.1017/CBO9780511559594.002
3. Lazear, E. P. (1981). Agency, earnings profiles, productivity, and hours restrictions. The American Economic Review, 71(4), 606–620. https://doi.org/10.3386/w0463
4. Raff, D. M. G., & Summers, L. H. (1987). Did Henry Ford pay efficiency wages? Journal of Labor Economics, 5(4, Part 2), S57–S86. https://doi.org/10.1086/298165
5. Krueger, A. B. (1991). Ownership, agency, and wages: An examination of franchising in the fast food industry. The Quarterly Journal of Economics, 106(1), 75–101. https://doi.org/10.2307/2937907