Match Quality and Marital Dissolution

Definition

Match quality is the unobserved component of the gains from marriage — the degree to which a particular couple's combination of traits, preferences, and circumstances generates household surplus beyond what either partner could achieve alone or in an alternative marriage. Because match quality is only partially revealed before marriage (through courtship and search) and continues to be revealed afterward (through cohabitation, joint production, and the evolution of each partner's traits), marital dissolution is fundamentally an information problem: couples accumulate evidence about their match quality over time and exit the marriage when the expected value of continuing falls below the sum of their outside options. Weiss and Willis (1997) provide the canonical empirical formalization of this framework.

Key Ideas

Two causes of divorce (Becker, Landes, Michael 1977): (1) a superior outside match is encountered (search); (2) traits that determine the gains from marriage change unexpectedly (surprises). Both operate through the same mechanism — the current marriage's value falling below the combined outside options.
Surprises dominate initial sorting on income: Unexpected changes in earning capacity during marriage strongly influence the divorce hazard; expected earning capacity at the time of marriage does not. This means it is what happens to spouses' economic circumstances after marriage, not who they were when they married, that drives most dissolution.
Earning capacity asymmetry: An unexpected rise in the husband's earning capacity stabilizes marriage (his market productivity is more valuable within the household due to division of labor); an unexpected rise in the wife's earning capacity destabilizes it (her market productivity is more valuable in the single state — it raises her outside option more than her contribution to household production). The most disruptive scenario: husband's predicted earnings fall and wife's rise simultaneously.
Rising hazard early in marriage: The divorce hazard initially increases with marriage duration, then declines. The rise reflects learning about unobserved match quality (Jovanovic 1979): since divorce is costly, couples experiment before separating, and the hazard peaks when accumulated information is sufficient to act on a bad draw. The subsequent decline reflects selective exit — divorce-prone couples leave earlier, leaving a survivor pool of more stable marriages.
Friction mitigates shocks: Small shocks to match quality or earning capacity do not trigger divorce if the marital surplus is large. This means couples with strong matches (educational homogamy, same ethnicity/religion, from intact families) are buffered against economic surprises that would dissolve weaker marriages.
Children and property as commitment devices: The presence of children (especially ages $3$ – $6$ ) and marital property both reduce the divorce hazard by raising the costs of dissolution — child support misallocation, legal costs of property division, loss of shared public goods.

How It Works

The Value Function Framework

At each period, a couple stays married if:

$V(\text{marriage state}) \geq A_{\text{husband}} + A_{\text{wife}} - C_{\text{divorce}}$

where $V$ is the expected value of continuing the marriage (discounted future household production), $A_i$ is each partner's outside option (value of single state plus option to remarry), and $C$ is the cost of divorce (legal costs, lost marital capital, child welfare inefficiency).

Divorce occurs endogenously when this condition fails — whenever the couple cannot find an allocation within the marriage that both prefer to divorce. Under transferable utility this rule is efficient regardless of divorce law (consent-based or unilateral), because the partner who gains from continuing can compensate the one who wants to leave.

The Identification Strategy (Weiss and Willis 1997)

The key empirical problem: unobserved match quality $\theta$ is correlated with observed characteristics at marriage (couples select partners partly on the basis of $\theta$ ). To purge this, Weiss and Willis assume that at the time of marriage couples are just indifferent — ex ante gains from marriage are zero. This means any favorable observed characteristic at marriage must be offset by a lower draw of unobserved match quality, allowing the researcher to back out $\theta$ from the marriage decision itself. Subsequent divorce hazard is then estimated conditional on this inferred initial match quality.

Earning Capacity Predictions as Surprises

Long-run earning capacity (predicted 1985 earnings) is estimated recursively using year-by-year information on work history, schooling, degree attainment, and personal background. The prediction at year $t$ of the marriage is based on the information set available at $t$ ; the prediction at year $t-1$ was based on the smaller information set available then. The difference between successive predictions — the revision — constitutes the "surprise." By construction, surprises are orthogonal to initial expectations, allowing clean separation of sorting from shocks.

Endogeneity Control

A woman anticipating divorce may invest more in market activities (defensive investment), inflating her measured earning capacity change. Weiss and Willis control for this using retrospective happiness ratings ( $1$ – $7$ scale) for the first five years of marriage: happiness is added to the prediction equation, then predictions are generated from the remaining variables, so that the measured "surprise" in the wife's earnings is the component unrelated to marital quality deterioration.

Why It Matters

Marriage Market Design

The finding that initial sorting on expected earning capacity is weak (spousal income correlation $= 0.2$ ) while sorting on education is strong (correlation $= 0.57$ ) reflects that education and schooling are complements in household production (shared leisure, child quality investment, preference similarity) while earning capacity is a substitute (division of labor). This has implications for how the marriage market equilibrium responds to changes in the female earnings distribution: rising female earnings raise the outside option of wives more than they raise the within-marriage contribution, mechanically destabilizing existing matches even absent any preference change.

Connection to Nonmarital Fertility and Multi-Partnered Fertility

The rising instability of marriages as female earning capacity increases is a structural feature of the household production model — not a cultural or preference-based change. This provides a formal economic mechanism for part of the rise in marital dissolution that underlies the trends in Nonmarital Fertility and Multi-Partnered Fertility: as women's labor market outcomes improve and earnings distributions widen, the surprise component of earning capacity revisions grows, mechanically raising the frequency of dissolution-triggering shocks.

Connection to the Deserving Poor and Welfare Policy

The structure of household dissolution — driven partly by the wife's outside option improving — is directly relevant to the composition shift in Aid to Families with Dependent Children (AFDC) caseloads documented by Moffitt 2015 — The Deserving Poor, the Family, and the U.S. Welfare System. As divorce rates rose and never-married motherhood increased (partly endogenously to rising female earnings and declining male earnings in certain communities), the composition of the welfare caseload shifted from "deserving" widows to "undeserving" never-married and divorced mothers — ultimately triggering the political realignment behind the Personal Responsibility and Work Opportunity Reconciliation Act (PRWORA) 1996. See Deserving Poor.

Open Questions

Does the earning capacity asymmetry persist at later marriage ages and later career stages? The National Longitudinal Survey of the High School Class of 1972 (NLS-72) sample is truncated at age 32, when many marriages are still in early stages.
How does the effect change as gender earnings gaps narrow? The asymmetry is predicted to shrink as female earning capacity becomes more valuable within marriage (if complementarities replace division of labor).
Can the framework be extended to incorporate fertility endogeneity? Dissolution risk affects fertility decisions, and children affect dissolution — the recursive structure is not fully modeled.
Is the "just indifferent at marriage" assumption empirically defensible? If there are substantial search frictions and rents to a given match, the assumption fails and the identification breaks down.