About the Dashboard

Data Source: The dashboard runs on the local-area arm of Microcosm, PolicyEngine's calibrated national microdata (policyengine/populace-us): about 1.6 million households on a hybrid survey spine (the CPS ASEC carrying detailed program and tax information, plus the 2024 American Community Survey for local-area density), reweighted against roughly 4,500 administrative targets: the full IRS Statistics of Income surface by state (including federal CTC, ACTC, and EITC amounts), USDA SNAP benefits and caseloads, Medicaid enrollment, and state and congressional-district populations. One file covers all 50 states and DC, pre-partitioned into per-state slices so each analysis simulates exactly the selected state's households. Both the dataset release and the PolicyEngine US model version are pinned per deployment (shown at the backend's health endpoint) and bumped deliberately.

Microsimulation:PolicyEngine's open-source tax-benefit model computes federal and state taxes and benefit programs for every household under current law and under your reform; every reported impact is the difference between those two simulations for the analysis year.

Poverty Measurement:We use the Supplemental Poverty Measure (SPM), which accounts for geographic cost-of-living differences, taxes, and the value of government benefits. State child-SPM rates on the current dataset track the Census Bureau's published state figures with a median deviation under 5%, though individual small states can differ more; see Validation below.

Fiscal Cost: Costs are calculated as the difference in total government spending and tax revenue between baseline and reform scenarios, split into federal tax, state tax, and benefit-outlay components, with per-program attributions for the credits the dashboard models.

The dataset's calibrated surfaces reproduce their administrative targets essentially exactly: federal CTC, ACTC, and EITC amounts by state (IRS Statistics of Income), SNAP benefits and caseloads (USDA, FY2024), Medicaid enrollment, and state populations (our audit measures the population fit at a median +0.9%). Each release is also scored against benchmarks it was notcalibrated to, published on PolicyEngine's calibration dashboard.

State child poverty: for 2024, the year the microdata represents, simulated state child-SPM rates run modestly above the Census figures on net (a median signeddeviation of about +5%; absolute per-state errors are larger, with 28 of 51 states within ±25% and real outliers beyond that). The dashboard's 2026 baseline rates run roughly 20% below the latest (2023) Census print, largely because 2026 law is genuinely different: the $2,200 OBBBA Child Tax Credit, SNAP rule changes, and new state childcare programs all reduce projected child poverty relative to 2023 law. Individual small states can deviate substantially in either direction (Vermont and Maine high; Arkansas, Hawaii, and Michigan low).

State credit costs (2026 vs official outlays):state EITC totals land at a median +2.4% of official figures (23 of 29 states within ±25%). State CTC totals run about a third above official outlays, chiefly because the model assumes full take-up of refundable child credits, where real-world participation among low- and no-liability filers runs roughly 50–75%; several benchmarks also predate recent program restructures.

Dataset transition (August 2026): the dashboard moved from an earlier 57,000-household file to the current 1.6-million-household dataset with exactly-calibrated federal credit surfaces. Reform impacts changed with the recalibration, generally toward smaller, better-anchored poverty effects (the earlier file concentrated credit-responsive households too heavily), so results predating the switch are not comparable.