Operationalizing family structure and auxiliary benefits

Why this chapter exists

Family structure is not a side table in Social Security modeling. It is one of the main reasons a dynamic model is needed at all.

That is especially true for:

spousal benefits
survivor and widow(er) benefits
divorced spouse benefits
dual entitlement
dependent and child-in-care benefits
reform proposals aimed at widows, caregivers, or households with interrupted work histories

A model can match own-worker benefit distributions and still be weak on some of the most policy-relevant adequacy questions if it treats family history as an afterthought. This chapter therefore does for family structure what the earlier operational chapters do for earnings, disability, and claiming: it turns a broad requirement into a concrete build plan.

The main modeling distinction

The proposal should keep three objects separate.

Current family status whether a person is married, divorced, widowed, never married, or living with dependent children in the current year
Relationship history the timing, duration, and ordering of marriages, divorces, deaths, remarriages, and parent-child links over the life course
Benefit-facing auxiliary status whether current law would make the person eligible for spouse, survivor, divorced spouse, or dependent benefits and whether those benefits top up or replace an own-worker entitlement

Those objects overlap, but they are not the same.

For example:

two currently unmarried women of the same age can have very different benefit prospects if one is never married and the other is widowed
two currently married couples can imply very different spouse and survivor outcomes depending on the earnings asymmetry within the pair
divorced spouse benefits depend on marriage duration and remarriage history, not just current marital status
child-in-care and dependent benefits depend on family links and child ages, not just on the worker’s earnings record

The model should therefore carry explicit relationship history and benefit-facing family states rather than relying on current marital status alone.

Why this matters for policy analysis

Some of the most visible Social Security adequacy debates run through family pathways rather than through the retired-worker benefit formula alone (Whitman and Reznik 2011; Tamborini and Cupito 2013).

That includes:

widow benefit proposals
divorced spouse eligibility and adequacy concerns
gender disparities driven by interrupted work histories
proposals that raise or lower own-worker benefits but have different effects on dually entitled beneficiaries
caregiver credit proposals that interact with later spouse or survivor benefit receipt

This is exactly where a public model can otherwise overstate its usefulness. If the family-history layer is weak, the model may still look fine on aggregate OASDI payments while misrepresenting adequacy for widows, divorced beneficiaries, and low-own-earnings spouses.

What the public benchmark models tell us

The public record on existing models gives useful guidance about what needs to be explicit.

DYNASIM

Public descriptions of DYNASIM indicate that it carries family structure, marriage, divorce, disability, and Social Security benefit logic inside a larger retirement-income model (Favreault et al. 2015; Urban Institute 2024).

That is the right benchmark for seriousness. It implies that a model meant to analyze spouse and survivor benefits cannot stop at person-only earnings paths.

But DYNASIM does not solve the transparency problem for us:

the full code is not open
the public record is stronger on model scope than on exact relationship-history mechanics
outside researchers cannot fully inspect how family-history construction and auxiliary-benefit logic interact

MINT

MINT is a useful public benchmark because its documentation is clearer about what is included and what is simplified (Smith et al. 2021; Social Security Administration 2024).

The main lesson here is that a credible phase 1 does not need to claim the full combinatorial complexity of every family-benefit path. But it does need to be explicit about:

which auxiliary-benefit categories are in scope
how marriage histories are carried
whether claimant timing is simplified
which exceptions and edge cases are deferred

CBO

The public CBO record remains relatively thin on person-level family construction and auxiliary-benefit pathways. That is itself useful. It means the proposal should not pretend we know more about public CBO micro-implementation than the documentation supports.

Recommended state representation

The proposal should specify a family-history layer as a first-class state block inside longitudinal populace.

Current marital-status state

At the annual level, the model should distinguish at least:

never_married
currently_married_first
currently_married_remarried
divorced_not_remarried
widowed_not_remarried
separated_or_transitioning

The exact categories can change, but the main point should remain: current status needs to distinguish remarried and previously married people from never-married people because their auxiliary-benefit exposure differs materially.

Relationship-history layer

The annual status is not enough. The panel should also carry a relationship-history object with:

current spouse_id where applicable
prior spouse links
marriage start date or year
marriage end date or year
end reason: divorce, death, separation, or unresolved
marriage order
marriage duration in months or month buckets
age at marriage
spouse age gap
spouse education pairing
spouse lifetime-earnings rank proxy

This is the layer that makes divorced spouse rules, widowhood rules, and dual-entitlement logic possible.

Parent-child and dependent layer

Auxiliary benefits are not only about spouses. The family-history layer should also support:

child links
child year of birth or age
child disability-before-22 flag where feasible
child-in-care indicator
co-residence where relevant to dependent pathways

Not all of those need to be in the first funded release, but the schema should anticipate them.

Benefit-facing auxiliary state

The proposal should also define an explicit benefit-facing family state. At minimum, that state should distinguish:

own_worker_only
spouse_eligible_topup
divorced_spouse_eligible
survivor_eligible
dually_entitled
family_maximum_relevant
child_or_dependent_pathway

These are not all mutually exclusive in a legal sense, but they are useful operational categories for simulation and validation.

Event overlay

As with claiming, monthly timing can matter even if the main panel is annual. The family-history layer should therefore include lightweight event timing for:

marriage
divorce
spouse death
remarriage
child birth where in scope

This does not require a fully monthly model for every process. It does require more than a single annual marital-status flag.

Phase 1 scope

A fundable phase 1 should be narrower than “all family pathways,” but it should still be strong enough to support the main auxiliary-benefit questions.

Core phase-1 family objects

The first funded version should aim to support:

current and prior marriage histories
marriage duration sufficient for the 10-year divorced spouse rule
spouse links for current couples
widowhood timing
pair-level own-worker versus spouse-benefit relevance
dual-entitlement logic for spouse top-ups
survivor-benefit relevance for older adults

That is already a meaningful capability set. It covers many of the questions policy organizations actually ask about widows, low earners, and households with uneven earnings histories.

What phase 1 can simplify

The proposal should state plainly that phase 1 may simplify:

precise monthly sequencing for every remarriage interaction
parent benefits
some dependent-child and child-in-care pathways
simultaneous management of many current and former spouse claim paths
very detailed family-maximum interactions outside the main older-adult cases
rare exception cases that matter little for aggregate incidence

These simplifications are acceptable if they are disclosed and bounded. They are not acceptable if the proposal implies full household-history completeness without actually funding it.

What phase 1 should not simplify away

Phase 1 should not collapse:

widowhood into generic non-married status
divorced and never-married people into one category
current couples into random pairings that ignore earnings asymmetry
spouse and own-worker benefit receipt into a single “beneficiary” flag

Those shortcuts would erase exactly the parts of the program that make family structure policy-relevant.

Why older-adult history deserves priority

The Social Security-first version of the project does not need to model every family process equally well at every age.

If resources force prioritization, the model should favor the histories that matter most for near-retirement and beneficiary analysis:

marriage duration for older adults
widowhood timing
remarriage after widowhood or divorce
spouse earnings asymmetry
current-law exposure to spouse and survivor benefits

That is a better phase-1 priority than trying to perfect general fertility dynamics or every child-benefit path before the older-adult auxiliary layer is credible.

Recommended construction strategy

The family-history layer should be built in a way that respects both the existing populace cross section and the needs of Social Security benefit logic.

1. Start from the base-year household network

populace already provides a cross-sectional household and family structure. That gives the project a real starting point for:

current couples
children in household
household composition
current relationship links

The longitudinal build should preserve this observed or synthesized current network where possible rather than redrawing it from scratch.

2. Reconstruct missing histories backward

For current adults, the model needs plausible prior marriage and family events that are not directly observed in the base-year cross section. Backward reconstruction should infer:

likely prior marriages
timing of divorce or widowhood
duration of prior marriages
likely spouse characteristics where the former spouse is not present

This is a synthetic-history problem similar in spirit to earnings reconstruction, but with stronger relational constraints.

3. Simulate forward transitions

After base-year history construction, the model should simulate:

first marriage
divorce
widowhood
remarriage
dependent-child transitions where in scope

These can be implemented as hazard models or related reduced-form transition models. The important thing is not sophistication for its own sake. It is preservation of the joint relationship structure that benefit rules depend on.

4. Match couples jointly, not independently

Person-level marriage hazards alone are not enough. A family-history layer also needs a couple-formation mechanism.

That mechanism should preserve:

age gaps
educational homogamy
race and ethnicity patterns where modeled
geography where sample quality permits
spousal earnings correlation
dual-earner versus single-earner household patterns

This is one of the strongest reasons to think in terms of longitudinal populace rather than a loose collection of independent hazards.

5. Enforce relational consistency

The build needs an explicit repair or constraint pass so that:

marriage links are two-sided
no person has impossible overlapping marriages
widowhood requires a spouse death event
child ages are feasible relative to parent ages
remarriage timing does not violate prior relationship end dates

Without this pass, a synthetic panel can look reasonable in marginal statistics while still containing relational impossibilities that break benefit calculations.

6. Derive benefit-facing family variables

Only after the relationship layer is coherent should the model derive:

spouse PIA relevance
ex-spouse PIA relevance
deceased-spouse PIA relevance
current spouse top-up eligibility
divorced spouse eligibility
widow(er) eligibility
family-maximum relevance
dual-entitlement categories

That ordering matters. Auxiliary-benefit variables should be outputs of the relationship-history layer, not substitutes for it.

Matching and household synthesis

The proposal should go one level deeper than “we will estimate marriage hazards.”

Why hazards alone are not enough

A person-level hazard can say who is likely to marry or divorce, but it does not by itself determine:

whom they marry
whether couples look realistic jointly
whether spouse earnings asymmetry is preserved
how remarriage affects later survivor exposure

Those are pair-level and household-level problems.

Recommended practical approach

The simplest credible phase-1 approach is a hybrid:

estimate person-level entry and exit hazards from panel data
when a union is formed, perform constrained matching using a distance-based or score-based pairing algorithm
run a consistency repair pass on the resulting relationship network
benchmark the resulting couples against cross-sectional and panel targets

This is less elegant than a fully joint generative household model, but it is likely more fundable and easier to validate in phase 1.

Higher-upside extension

If longitudinal populace advances enough, the project can later move toward hierarchical or household-first generation that jointly models:

household composition
partner assignment
child links
earnings correlation within couples
tax-unit and family-unit consistency

That is a plausible phase-2 or methodology-R&D direction. It should not be the phase-1 dependency.

Estimation inputs and benchmark sources

The family-history layer needs both panel data and benefit-facing targets.

Main empirical inputs

The main sources should include:

PSID for marriage, divorce, remarriage, labor-force history, and pair-level earnings relationships
HRS for widowhood, older-adult family status, retirement transitions, and benefit-facing older-cohort family histories
CPS and ACS for current marital-status distributions, household structure, age gaps, educational pairing, and dual-earner patterns
SSA published statistics for spouse, widow(er), and other auxiliary-beneficiary counts and average benefits
MINT and DYNASIM documentation as public benchmarks for what a serious but still simplified family-history layer should contain

Where public data are weak

The project should also be transparent about weak spots:

exact histories for ex-spouses not present in the household
rare auxiliary-benefit categories
exact timing of sequential marriages and claims for every record
parent-benefit pathways
some child-in-care pathways outside the main older-adult use cases

These are reasons for scoped promises, not reasons to leave the family layer vague.

Evaluation metrics for family structure

The family-history layer should be judged on both demographic realism and benefit relevance.

Core demographic and relationship metrics

Metric	Why it matters
Married, divorced, widowed, never-married shares by age and sex	Basic family-status realism
Share ever married by cohort	Tests life-course history, not only current status
Remarriage prevalence by age and sex	Important for divorced spouse and widow pathways
Widowhood prevalence by age and sex	Directly linked to survivor exposure
Mean spouse age gap	Couple realism
Educational homogamy within couples	Pairing realism
Dual-earner share among couples	Important for spouse-benefit exposure
Spousal earnings correlation or rank correlation	Central for adequacy and top-up patterns

Core auxiliary-benefit metrics

Metric	Why it matters
Spouse beneficiaries by age and sex	Direct test of auxiliary exposure
Widow(er) beneficiaries by age and sex	Direct test of survivor pathways
Divorced spouse beneficiary counts where available	Tests duration and remarriage logic
Share of beneficiaries who are own-worker only versus dually entitled	Important for interpreting reforms
Average spouse and widow(er) benefit	Links history quality to actual benefit levels
Family-maximum incidence where in scope	Tests dependent-benefit logic

Distributional stress tests

If the project wants a stronger adequacy story, it should also examine:

benefit-source decomposition for low-own-earnings women
poverty and near-poverty among widow(er) households
benefit outcomes for divorced versus never-married older adults
reform incidence for households with uneven lifetime earnings

This is where the family-history layer becomes directly legible to funders and policy partners.

Suggested stage-1 thresholds

The proposal should not dodge stage-1 numeric expectations for this layer either.

Metric	Draft stage-1 threshold
Major marital-status shares by age-sex cells	within 2 percentage points
Widowhood and divorce prevalence at older ages	within 2 percentage points
Mean spouse age gap	within 0.5 years
Educational homogamy shares	within 2 percentage points
Spousal earnings rank correlation	within 0.05
Spouse and widow(er) beneficiary counts	within 2-3 percent
Average spouse and widow(er) benefit	within 2-3 percent
Share dually entitled among older beneficiaries	within 3 percentage points

The point is not that these exact cutoffs are sacred. The point is that the proposal should make the family-history layer falsifiable.

Recommended operational sequence

The build sequence should reflect both policy value and implementation difficulty.

Older-adult marriage histories enough to distinguish never married, divorced, widowed, remarried, and currently married records in benefit-relevant ways
Current spouse links and dual-entitlement logic enough to support spouse top-ups and own-worker versus auxiliary decomposition
Survivor and divorced spouse pathways using widowhood timing, marriage duration, and remarriage history
Dependent and child-in-care pathways where they materially improve coverage of auxiliary categories
Rare and complex edge cases including parent benefits and more exact sequential claim interactions

This is realistic, fundable, and much easier to validate than promising the whole family-benefit codebook at once.

How this should be positioned in the proposal

The proposal should make three points clearly.

1. Family structure is a core state block, not downstream polish

Auxiliary benefits are not an optional enhancement to add after the earnings model is “done.” They are part of the minimum Social Security story.

2. A narrower phase 1 can still be serious

The project can credibly promise:

marriage histories good enough for spouse, survivor, and divorced spouse analysis
couple matching good enough to preserve earnings asymmetry
explicit dual-entitlement treatment
bounded simplifications on rare auxiliary categories

That is already a strong and fundable claim.

3. This layer should have its own stage gate

If the model cannot produce directionally credible widow, spouse, and dual-entitlement outcomes, the project should narrow its public policy claims accordingly. Family structure should not be treated as something that can silently underperform while the proposal still advertises full adequacy analysis.

Bottom line

The proposal should not describe family structure as a few marriage hazards plus a spouse-benefit rule call.

It should describe an explicit relationship-history layer inside longitudinal populace, say what phase 1 will and will not include, benchmark those choices against DYNASIM and MINT, and evaluate the result against the auxiliary-benefit outcomes that policy users actually care about.