6 Evaluating Antipoverty Policy

Define what success means

Begin by specifying the policy’s goals, the population it is intended to help, and the time horizon for assessing results. Possible goals include reducing the number of people below a poverty threshold, alleviating deep poverty, improving food or housing security, supporting employment, and improving health or children’s development.

These goals are related but not interchangeable. For example, a food benefit may reduce food hardship without lifting a family above an income-based poverty line. A short-term income boost may reduce measured poverty without resolving chronic housing insecurity. Report the intended outcome as well as relevant broader outcomes.

Choose measures that fit the goals

The Official poverty measure compares pretax cash income with a threshold. The Supplemental Poverty Measure (SPM) also counts tax credits and benefits such as SNAP and housing assistance, subtracts certain necessary expenses, and adjusts thresholds for housing costs and geographic differences. These measures answer different questions, so a policy may appear more or less effective depending on which measure is used.

Use several kinds of outcomes to avoid relying on a single statistic:

• Poverty rate: the share of people below the chosen threshold.

• Poverty gap and depth: how far below the threshold people are, including the share in deep poverty.

• Material hardship: food insecurity, housing instability, or difficulty paying essential bills.

• Longer-term outcomes: health, education, employment, and earnings.

• Distribution and cost: who benefits, who is missed, and the resources required per outcome.

A policy’s effects should be considered alongside who receives its benefits and what resources are required. An evaluation that reports only an overall poverty rate may miss changes in hardship, deep poverty, or longer-term outcomes.

Assess what the evidence can establish

A basic accounting analysis estimates how many people would fall below a poverty line with and without counting a program’s benefits. This shows the program’s immediate contribution to measured resources. It does not, by itself, establish the Counterfactual: families may change work, spending, or program participation in response to the policy.

To estimate causal effects, researchers compare outcomes with a credible counterfactual. They may use randomized evaluations where feasible or strong quasi-experimental designs, such as policy changes that affect similar groups differently. Evaluations should distinguish immediate effects from longer-term effects and state the uncertainty, population, time period, and limits of the research design.

Evidence on child poverty illustrates why findings should be reported across outcomes and approaches. A National Academies review found that EITC expansions improved some child health and educational outcomes, and that SNAP improved birth outcomes and other health outcomes. Evidence on housing assistance was mixed, though some children who moved to lower-poverty neighborhoods had better adult outcomes. The review also concluded that none of the single policy options it assessed was estimated to cut child poverty in half; combinations of policies were more promising.

Examine access and implementation

Written eligibility rules do not guarantee that eligible people receive help. Application complexity, documentation requirements, recertification, office hours, digital access, language barriers, stigma, and limited outreach can affect participation. Institutions shape who can successfully claim a benefit, while social status and local conditions shape how people experience those institutions.

Examine the path from eligibility to outcome: who is eligible, who applies, who is approved, how quickly aid arrives, and who continues to receive it. Disaggregate results by relevant groups and locations because an overall average can hide low take-up or unequal access. Burdensome application processes may deter participation, and there can be gaps in connecting potentially eligible students to SNAP.

Implementation involves trade-offs. Simplifying rules can reduce burdens and delays, but weak verification may increase incorrect payments. Tighter checks may improve accuracy yet make access harder for eligible households. Evaluate access, timeliness, administrative workload, decision accuracy, and payment errors together rather than treating any one dimension as sufficient.

Weigh trade-offs and unintended effects

Assess effects beyond a policy’s stated purpose. Examples to test include benefit phase-outs that reduce the gain from additional earnings, rules that discourage saving or reporting changing income, and complicated tax-credit requirements that lead to errors or unclaimed benefits. These effects are not inevitable; their size depends on program design and household circumstances. A possible trade-off is not proof that a program fails: measure the actual effect and compare it with the benefits.

The Earned Income Tax Credit (EITC) is designed to supplement earnings, but its eligibility and filing rules can be complex. Evaluation should consider poverty reduction and take-up, as well as errors and the burdens households face in claiming the credit. Work requirements should be assessed for their employment effects and for whether reporting or compliance barriers cause eligible people to lose assistance.

Program design and local administration may vary. The National Academies noted that variation in how states used TANF funds limited the evidence available to estimate the effects of particular approaches on family income and child poverty. When implementation differs, document what was delivered, where, and to whom before drawing conclusions about the policy itself.

Bring the evaluation together

A well-rounded evaluation asks six connected questions:

1. What problem and population does the policy target? Define eligibility, goals, and the time horizon.

2. What changed because of the policy? Choose suitable measures of poverty, hardship, and longer-term outcomes.

3. Compared with what? Identify the counterfactual and the limits of the research design.

4. Did intended recipients receive the benefit? Examine take-up, timeliness, access barriers, and variation in local delivery.

5. What else changed? Measure behavioral responses, errors, stigma, unmet need, and other unintended effects.

6. For whom did it work, and at what cost? Report distributional effects, uncertainty, and costs alongside average impact.

Antipoverty policy is both a resource intervention and an institutional process. Poverty measures define what counts as success, causal evidence estimates what changed, and implementation analysis shows who actually benefited. A credible judgment combines poverty reduction with hardship, access, longer-term outcomes, unintended effects, distribution, and cost. Because results can differ across groups and settings, evidence supports careful comparison rather than a simple label of “works” or “doesn’t work.”