Family Stability: What Works, What Falls Short

The strongest starting point

A family should be able to get help with the bill, the food, the care or the housing gap that is putting it at risk. The most useful combined approach starts with adequate material support, makes that support easy to receive, and adds specialized services where a family wants and needs them.

This article ranks ten priorities for reducing preventable family hardship in the United States, with California implementation in mind. It also maps 60 approaches, including older models, established programs and proposals still being developed. The ranking and combined design are editorial judgments informed by the cited research.

The reviewed evidence does not establish that these ten ideas, separately or together, eliminate 90% of family instability. That is a demanding goal for a defined program to test. A measured reduction in one outcome—for example, emergency hospital use—cannot stand in for solving every problem a household faces.

What we compared—and how to read the ranking

The problem here is preventable material hardship and its consequences: insufficient food, loss of housing or utilities, unaffordable care, interrupted work, and family stress connected to unmet needs. Family safety and access to appropriate treatment also matter. Families facing immediate danger need a response matched to that danger, regardless of this policy ranking.

This is a focused evidence review, not an exhaustive systematic review of thousands of programs. It combines a major research synthesis with selected randomized trials, credible quasi-experiments and more recent follow-ups. The 60-item catalogue is a map of options; inclusion does not mean that every variant received an individual impact evaluation here.

Priority reflects five considerations: relevance to severe unmet needs; credibility of causal evidence; reach; durability; and practical delivery. We favor reliable evidence of a useful outcome over an attractive program name. Costs, eligibility, local capacity and the household’s choices can change the order. No numeric score or universal success percentage is assigned because the studies measure different populations and outcomes.

“Strong for the measured outcome” means a rigorous study supports that particular result. “Mixed” means results vary by design, site or outcome. Neither label guarantees that California would reproduce the finding. The order is a policy investment sequence, not instructions to make an individual family wait for lower-ranked care.

Ten priorities to combine

Use the table as a starting order for investment. The sections below explain the evidence, limits and ingredient to carry into a combined service.

1. Reliable income for essential expenses

A household with a continuing monthly shortfall needs continuing resources. The policy options include refundable tax credits, a child allowance and cash assistance, with benefits designed to reach families with very low or no earnings. A work-linked credit and an unconditional child payment reach different groups.

The National Academies review supports combining income support with policies that encourage employment. A more recent cash trial adds useful limits: OpenResearch’s three-year experiment increased income including transfers and spending on basic needs, while employment was about two percentage points lower. It did not establish broad improvements in physical health or children’s short-run academic outcomes.

Carry forward: predictable payments, accessible enrollment and a clear accounting of interactions with other benefits. Price the recurring commitment. Do not promise that cash alone supplies a scarce apartment, a child care place or treatment. The recommendation is to close the resource gap while making those services available.

Sources: National Academies · A Roadmap to Reducing Child Poverty (2019) · OpenResearch · Unconditional Cash Study

2. Affordable, stable housing

For a family whose rent consistently exceeds its resources, durable affordability is the central intervention. In HUD’s randomized Family Options Study, long-term subsidies produced the clearest improvements in housing stability and several family outcomes. Short-term rehousing and service-intensive transitional programs generally did not match those gains relative to usual care.

The study enrolled 2,282 families already experiencing homelessness. It is strong evidence for that population, not a direct estimate for every family at risk. Some subsidies also reduced work and earnings; assess housing security and labor outcomes separately.

Carry forward: a subsidy that lasts as long as the affordability gap requires, practical help finding an acceptable unit, and voluntary services. California also needs available homes and willing participating landlords. Construction, preservation and land-use measures belong in the longer-term supply plan, but this review does not assign their effects the voucher trial’s results.

Sources: HUD · Family Options Study

3. Dependable food support

Build a year-round food plan: grocery benefits, school meals, summer support and age-appropriate nutrition programs, with emergency food providers available when a payment or application is delayed.

A randomized evaluation of the 2011–2013 Summer EBT demonstrations surveyed approximately 52,000 households. It found that the intervention reduced very low food security among children by about one-third and improved several nutrition measures. This is a substantial reduction in a specific form of hardship; it is not a one-third reduction in all family problems, nor an estimate for every food program.

Carry forward: bridge the periods when school meals are unavailable, help families obtain the benefits for which they qualify, and verify that food is accessible by location, hours and transportation. Count meals missed and food insecurity, rather than merely counting cards issued or referrals made.

Sources: Collins et al. · Summer nutrition benefit trial (2018)

4. Health coverage and accessible care

Medical expenses can destabilize a family even when its income has not changed. Coverage should connect to available primary, dental, mental-health and other appropriate care, with help understanding bills and renewal requirements.

The Oregon Medicaid lottery study found that coverage nearly eliminated catastrophic out-of-pocket spending and reduced positive depression screens by 9.15 percentage points after approximately two years. It did not detect significant improvements in several measured physical-health indicators. The study concerned low-income adults and does not establish identical effects for every family or insurance design.

Carry forward: financial protection, continuity of coverage and timely appointments. Budget for services actually used; improved access does not automatically lower total medical spending. Judge specific treatments using their own evidence, rather than assuming that an insurance finding validates every clinical intervention.

Sources: Baicker et al. · Oregon Medicaid experiment (2013)

5. Affordable, dependable child care

A usable child care arrangement must fit a child’s needs, a parent’s schedule and the family’s transportation options. Payment assistance matters most when a suitable place is actually available.

A randomized Cook County study found more stable care, greater satisfaction and fewer work-related child care problems among families offered subsidies. It did not find higher employment or earnings over two years; most parents were already working. A Washington trial also found that lower copayments lengthened subsidy use without increasing employment or earnings.

Carry forward: affordable fees, reliable providers, continuity during renewal and help covering gaps in care. Expand provider capacity where shortages are the binding constraint. Track missed work, care disruptions and parent choice alongside earnings. The recommendation is for usable care, with quality and workforce costs included.

Sources: MDRC · Child care subsidies in Cook County (2010) · MDRC · Reducing child care copayments in Washington (2010)

6. Timely help with a solvable financial shock

A temporary setback may need a fast, limited payment: rent arrears, a utility balance or another essential expense. A continuing income gap requires a longer-term response.

In a Chicago study of 4,448 callers, people seeking help when funding was available were 76% less likely to enter a homeless shelter. Funding variation supported a quasi-experimental comparison. The result is about this targeted prevention program; it is not proof that any small grant prevents homelessness at that rate.

Context changes results. Five pandemic-era rental assistance programs in four urban areas increased rent payment but had little effect on housing stability overall. Carry forward: enough funding to resolve the identified shortfall before its deadline, verification proportionate to the payment, and a route to continuing support when one payment will not work. Applying the same flexible fund to car repairs or diapers is a design proposal, not a result established by these rent studies.

Sources: Evans, Sullivan & Wallskog · Homelessness prevention (2016) · Collinson et al. · Pandemic rental assistance (2025 manuscript)

7. Hands-on enrollment and renewal help

An unreceived benefit cannot help a household. Offer a person who can explain eligibility, assist with the application, help obtain documents, follow up and troubleshoot renewal. Use reminders and software to support that work.

In an experiment involving approximately 30,000 older adults likely eligible for SNAP, nine-month enrollment was 6% with usual practice, 11% with eligibility information, and 18% with information plus application assistance. Assistance improved take-up, but most people still did not enroll. Those newly reached were not necessarily the neediest.

Carry forward: a named contact and confirmation that help was actually received. The numerical result should not be transferred to parents or every benefit program. Automatic enrollment, shared forms and data reuse need their own legal authority, privacy protections and evaluation. A website directory is a useful entry point; its existence is not evidence that a need has been resolved.

Sources: Finkelstein & Notowidigdo · SNAP enrollment experiment (2018/2019)

8. Training linked to real jobs and advancement

Useful training connects a person to an occupation with demand, adequate pay and a realistic path to advancement. Include the care, transport and time needed to complete it.

The ten-year WorkAdvance follow-up provides both encouragement and a warning. St. Nicks Alliance increased average year-ten earnings by 32%; the other three providers had no significant effect on the two prespecified year-ten outcomes. All four had increased earnings at some earlier point. These are provider-specific results, not a guaranteed return from training.

Carry forward: employers involved in course design, demonstrated placement and earnings outcomes, and follow-up when sectors change. Publish completion, job quality, earnings and cost by provider. Offer support to people who cannot currently work; training cannot substitute for the resources they need today.

Sources: MDRC · WorkAdvance after ten years (2025)

9. Paid leave families can actually use

A birth or serious care responsibility should not force an avoidable choice between family care and immediate income. Payment levels, awareness, application burden and protection of the job all affect whether leave is usable.

A historical California study using a difference-in-differences design estimated that the original paid-leave program increased new mothers’ average maternity leave from roughly three weeks to six. This supports the ability to take leave. It is not proof of a permanent earnings increase for every parent or a 90% reduction in family hardship.

Carry forward: adequate wage replacement, understandable access and coordination with applicable job protections. The exact law, benefits and eligibility require current program guidance. This article uses the historical evaluation to assess the idea, not to give a current benefits determination.

Sources: Rossin-Slater, Ruhm & Waldfogel · California paid leave (2013)

10. Voluntary early family support with a tested model

Use a specific evaluated model, trained staff and appropriate referrals. Generic “family support” is too broad a label to inherit a program’s evidence.

The Family Connects trial offered brief postpartum nurse home visits and connections to community services. In the five-year impact follow-up of 531 children, assignment to the program was associated with 33% less emergency medical care use. The estimated 39% reduction in CPS investigations was less certain: its 95% confidence interval included no effect. An investigation is not the same as confirmed maltreatment.

Carry forward: voluntary participation, competent clinical supervision, clear consent and available referral partners. Benefits depend on the model and quality of delivery. Keep practical assistance available to families who decline home visits, and do not turn a request for food or diapers into an unnecessary surveillance requirement.

Sources: Goodman et al. · Family Connects five-year trial (2021)

What has been less effective—or remains unproven?

“Least effective” must name an outcome and comparison. A program can have value while failing to produce the particular result used to justify it. These are the clearest reasons in the reviewed evidence to avoid overpromising.

Sources: HUD · Family Options Study · Finkelstein & Notowidigdo · SNAP enrollment experiment (2018/2019) · MDRC · WorkAdvance after ten years (2025) · Kluender et al. · Medical debt relief experiments (2024 manuscript) · Collinson et al. · Pandemic rental assistance (2025 manuscript) · MDRC · Reducing child care copayments in Washington (2010)

A working catalogue of 60 approaches

The following map preserves the range of ideas worth comparing. Each row contains five distinct approaches. “Historical,” “established” and “developing” describe the broad lineage or current research status, not a claim that a program is available in every county. Some old ideas are still being tested in new forms.

Evidence in the numbered review above applies to the named intervention and population. The other catalogue entries are candidates for further, separate appraisal; this list does not certify their effectiveness or imply that all 60 have been evaluated here.

What newer research should change

New cash experiments make the discussion more precise. OpenResearch’s findings show that additional money can improve choice and some parenting practices without producing measurable improvement in every child or health outcome during the study. That argues for specifying what a cash program is meant to accomplish, rather than labeling it a universal cure or dismissing it because one outcome did not change.

Baby’s First Years continues to follow children after a randomized cash intervention begun at birth; its research team reports age-eight data collection underway in 2026. Longer-term outcomes deserve continued review. Early measures and later measures answer different questions.

For this proposed service, useful technology candidates are document checklists, opt-in reminders, translation support with review, and updates on referral status. These are design suggestions. Before expansion, test accuracy, time saved, accessibility and whether assistance actually arrives. Automated eligibility denial and secret family-risk scoring are not necessary ingredients.

Sources: OpenResearch · Unconditional Cash Study · Baby’s First Years · Study design and ongoing follow-up

Combine the strongest ingredients into one practical service

The proposed service would offer one voluntary entry point through county offices and trusted community partners. A family names its priorities and time-sensitive deadlines, chooses how to be contacted, and receives a short plan with a responsible person for each next step.

First, address imminent loss of an essential: food today, a rent deadline, a utility shutoff or a gap in care. Then distinguish a one-time disruption from an ongoing shortage. Use flexible assistance for a bounded shock and recurring support for a recurring gap. These are proposed delivery rules, not a tested ten-part intervention.

Second, assemble the benefits and services the household wants and can qualify for. Connect the family to housing, food, coverage and child care; add job advancement, leave assistance or a specific family-support model where relevant. Confirm receipt and check whether the need stayed resolved. Families should not have to participate in all ten components.

Third, maintain capacity behind the referral: available appointment slots, responsive agencies, transport options and actual assistance funds. A shared list cannot create these. Counties can coordinate many services, but changes to tax credits, benefit eligibility or paid-leave law may need state or federal action.

Finally, use a consent-based record that contains only what is needed for the service. Explain who can see it, let families correct errors and offer a human route for disputes. Evaluate rural access, language access and disability access from the start. Never count a family as stable simply because it stopped answering.

Make 90% a measurable ambition

The best-known broad benchmark in this review is much more modest than 90%. The National Academies’ 2019 simulations estimated child-poverty reductions of 50.7% and 52.3% for two large policy packages. Those are historical modeled estimates, not observed outcomes, current costings or forecasts for our proposed service.

Define the goal before operating the pilot. A useful service target would be: at least 90% of enrolled households with an identified, in-scope essential need have that need resolved by the agreed deadline and still resolved at 90 days. Report all enrolled households in the denominator; show unresolved and unknown outcomes separately. This is a proposed operational target, not a predicted success rate.

Also report households with all of their identified essential needs resolved. Resolving one of three problems should not qualify as complete household stability. Track residents who requested help but could not enroll, including the reason, so a high success rate cannot be manufactured by excluding harder cases.

A 90% reduction in crises is a different and much stronger claim. It requires a credible comparison group and a prespecified outcome, such as shelter entry or an essential-service loss over 12 months. Calculate relative reduction as 1 minus the intervention-group rate divided by the comparison-group rate; also publish the absolute rates, uncertainty and follow-up losses.

For example, a fall from 10% to 1% would be a 90% relative reduction and a nine-percentage-point absolute reduction. That arithmetic is illustrative. Do not add the percentages from unrelated studies: their populations overlap, outcomes differ, and the effects may interact.

Sources: National Academies · A Roadmap to Reducing Child Poverty (2019)

A pilot that can earn the right to expand

Start with a defined county or service area, eligibility rules, baseline needs and funded referral capacity. Pay participating parents and providers for helping design the process. Agree on urgent response standards, a recurring-support pathway and a realistic follow-up schedule.

Before recruiting, set the assistance budget, staffing capacity, evaluation design and stopping rules. A defensible budget separates direct support, navigator time, provider payments, administration, technology, evaluation and contingency. This review does not establish a local price per family or a net savings estimate.

Use an independent evaluator to choose an ethical randomized rollout where feasible, or a credible comparison design otherwise. Determine sample size from the expected baseline rate and detectable change; a small convenience pilot cannot establish a rare-outcome reduction of 90%. Do not withhold an existing entitlement or urgent protective service to create a comparison.

Publish household outcomes, cost per resolved need, recurring hardship, time to receipt, participant burden, complaints and access gaps. Expand only when the gains are meaningful, delivery is reliable and funding can continue. If the service mainly increases referrals without reducing unmet needs, change the service.

The recommendation

Invest first in the resources families need, connect them to those resources with capable people, and add specialized support based on the household’s circumstances. Keep the components that improve a clearly defined outcome; redesign or retire those that do not.

The ten priorities are a defensible starting package for development and testing. They are not ten independently proven 90% solutions. The aim should remain ambitious, while every claimed result stays tied to the people, outcome, comparison and time period actually studied.

Review notes

Prepared September 19, 2026, as a separate companion to “Family Stability, Before Crisis.” Research and synthesis prepared with AI assistance for Kevin G.’s requested publication. The practical California contact directory remains on the idea page.

This edition is selective, US-focused and intended for public discussion. It is not a pooled meta-analysis, a current eligibility guide or a complete inventory of every intervention worldwide. Future editions should add local evaluations and full cost comparisons. Corrections and stronger contrary evidence are welcome.