Earnings-Participation Paradox

Definition

The Earnings-Participation Paradox is the empirical pattern in which a program intervention reduces labor force participation while simultaneously increasing average earnings among those who remain employed. The apparent contradiction is resolved by compositional selection: when the margin of labor force exit is concentrated among lower-earning workers, the conditional mean earnings of the remaining employed population rises even as total participation falls. Chen (2012) documents this pattern for wives of Disability Insurance (DI) beneficiaries: DI crowds out wife's labor force participation by $-6\%$ yet wives of beneficiaries earn $\approx \$2,479$ more per year by year 5 than wives of rejected applicants.

Key Ideas

Mechanism: Not all workers are equally likely to exit in response to a crowd-out shock. If the marginal exits are drawn from the lower end of the earnings distribution, the average earnings of the surviving employed population rises. This is a compositional effect, not an intensive-margin wage gain for any individual worker
Extensive vs. intensive margin: The crowd-out operates almost entirely on the extensive margin (who participates), not the intensive margin (hours or wages among those who remain). Conditional earnings rise because low earners leave — not because remaining wives earn more per hour
Policy implications: Aggregate earnings statistics, used without participation data, lead to the incorrect conclusion that DI benefits wives' labor market outcomes. In fact, the wives most harmed are the lower earners who exit — the group least visible in conditional-mean statistics
Mathematical structure: Mean earnings = (fraction employed) × (mean earnings | employed). DI reduces the fraction employed; if low earners are selected out, mean earnings | employed rises enough to overcome the participation decline in the product formula. Under Chen's estimates, the selection effect dominates
Not unique to DI: Any program that induces selective exit from the labor force — along a dimension correlated with earnings — will produce the same pattern. Unemployment Insurance (UI) crowd-out of spousal labor supply, welfare-to-work programs, or job-training exits could all generate this paradox if the margin of response is lower-earning participants
Caution for crowd-out studies: Studies using only earnings as the outcome (rather than both earnings and participation separately) will systematically misread the direction and magnitude of crowd-out effects whenever selection is non-random

How It Works

Formally, let $\bar{E}$ be mean earnings for the spousal group, $p$ be participation rate, and $\bar{e}$ be mean earnings conditional on employment:

$\bar{E} = p \cdot \bar{e}$

A crowd-out shock that lowers $p$ by selecting out the bottom of the earnings distribution simultaneously raises $\bar{e}$ . If the selection is strong enough:

$\Delta \bar{E} = \Delta p \cdot \bar{e}_0 + p_1 \cdot \Delta \bar{e} > 0$

even though $\Delta p < 0$ . In Chen (2012), $\Delta p \approx -0.06$ but $\Delta \bar{e} \approx +\$2,479$ — the intensive-margin rise dominates the direct participation effect on mean earnings.

Why It Matters

Welfare analysis of crowd-out: The paradox shows that measuring crowd-out through earnings alone will give a misleading (and wrong-signed) estimate of the welfare effect on spouses. Proper evaluation requires the full distribution: who exits, at what earnings level, and what happens to their consumption
Distributional consequences: The wives most adversely affected by DI crowd-out are lower earners. Policies focused on aggregate earnings as the outcome metric will fail to detect harm to this group and may even report a spurious benefit
Program evaluation methodology: The paradox is a general econometric warning. Outcome aggregation can mask distributional harm whenever the treated-into-exit population differs systematically in the outcome from the treated-to-stay population

Sources