Methodology & limitations

Sources, construction and limits

Every figure on these pages comes from federal administrative sources. Nothing is estimated except where explicitly flagged. This page sets out the sources, the joins, the validation, and the limits of the data.

Sources

Reconstructing eligibility

Eligibility required poverty, so the comparison used throughout is against the tracts a state was eligible to designate and did not. The eligible universe was rebuilt from the statutory test.

A tract is a low-income community under §45D(e) if its poverty rate is at least 20%, or its median family income is at most 80% of the relevant benchmark — the statewide figure for non-metropolitan tracts, and the greater of the statewide and metropolitan figures for metropolitan ones. Applying that test to ACS 2011–2015 reproduces 31,562 eligible tracts against Treasury's published 31,680: agreement to within 0.4%.

The residual comes from tracts sitting on the threshold, where Treasury used a slightly different ACS vintage or rounding than the published tables expose. Because the gap is four-tenths of one percent and does not cluster in any state, it does not move any conclusion here. Note that 86 designated tracts fail our reproduction of the test; they are carried through as designated but left unranked.

The 2010 / 2020 boundary problem

Opportunity Zones were designated on 2010-vintage census tracts and the designation is fixed to those boundaries for the life of the zone. The Census Bureau redrew tracts for 2020. Any "current conditions" figure therefore requires bridging two incompatible geographies, and pulling present-day tract data without doing so silently mismatches a meaningful share of zones.

Tracts are classified into two tiers via the Census relationship file:

• Tier A — the 2010 and 2020 boundaries coincide exactly. Current values are read directly. 75.2% of the comparison universe.
• Tier B — the tract was split or merged. Current values are interpolated by land-area weight across the overlapping 2020 tracts.

All change-over-time figures on these pages use Tier A only. No interpolated estimate enters a headline number. Tier B tracts still appear on the maps, flagged with an asterisk in their detail panel.

Absence of tract-level investment data

A second limit sits inside the published totals. Treasury's state table sums to $104.2 billion while the headline national figure is $112 billion. The difference is investment reported without usable location information, plus the territories. So roughly $7.8 billion of qualified investment cannot be placed anywhere on a map — about 7% of the programme.

The within-state distress percentile

A national poverty cutoff would conflate "poor for Mississippi" with "poor for Connecticut". Governors chose from their own state's pool, so each eligible tract is ranked within its state, and a state's targeting score is the mean percentile of the tracts it designated. 0.50 is what drawing at random would produce.

Two cautions apply, and both cut against over-reading a low score:

• The 25-zone floor. The statute capped designations at 25% of a state's low-income tracts but guaranteed a minimum of 25 zones to states with fewer than 100. Wyoming designated 24 of its 33 eligible tracts and Alaska 25 of 57. Where a state designates most of its pool the percentile is constrained toward 0.50. Those states are excluded from the correlation analysis.
• Pool dispersion. Where eligible tracts within a state are similar to one another the measure has less to separate, so a value near 0.50 can reflect a narrow distribution rather than the selection of affluent tracts.

Limits of the analysis

• Not causal. Comparisons between designated and non-designated tracts are raw differences between groups that were not equivalent at baseline. Designated tracts started with higher poverty and more room to move. Nothing here identifies the effect of designation.
• State-level investment does not describe within-state allocation. With about 50 observations the state panel has low statistical power, and it is an ecological comparison: it does not identify which tracts within a state received investment. Investment is concentrated within states as well as between them.
• ACS margins of error are not carried through. Tract-level five-year estimates for small populations can be imprecise, particularly for median values. Differences of a percentage point or two between individual tracts should not be treated as firm.
• The 2027 comparison is an area-weighted overlap, not a lookup. The two rounds use different tract vintages, so current zones are matched to the new eligible universe through the relationship file. A zone is called ineligible when essentially none of its land falls in an OZ 2.0-eligible tract; partial cases are reported separately rather than forced either way.

The OZ 2.0 income benchmark

One change in the One Big Beautiful Bill Act is not described in the published summaries. Under the previous §45D(e) test, a metropolitan tract was measured against the greater of the statewide and metropolitan median family income, while micropolitan tracts were treated as non-metropolitan and measured against the statewide figure.

Treasury's OZ 2.0 eligible file uses the CBSA figure for every tract in any CBSA, metropolitan or micropolitan, and the statewide figure only for tracts in no CBSA at all. We verified this directly against the file: of 71,368 tracts with a CBSA, the benchmark equals the CBSA figure 100% of the time, and equals the greater of the two only 62.3% of the time. In 26,882 tracts the new benchmark is lower than the old one.

So the income test tightened three ways at once, not one: the ratio fell from 80% to 70%, the poverty route gained a 125% income cap, and the benchmark itself became less generous. Together these cut the eligible pool from 31,680 tracts to 25,332.

Analysis by NTONTAN. Prepared July 2026 from data current as of Treasury's 30 April 2026 extract. Iowa analysis · National picture