The 2027 update widened the fairness gap by home value.
Gold-standard fairness tests (PRD, PRB, and VEI) found our model less regressive than OPA, making it a useful independent benchmark—not ground truth. The new 2027 reassessment moved cheaper homes farther from that benchmark while moving more expensive homes closer to that benchmark. The cheaper-home shift was larger. All three tests show that regressivity got worse.
What happened
The cheaper the home, the farther above our estimate.
We ranked homes by our estimate and split them into five equal groups. A ratio above 100% means OPA assessed the group above our estimate. Below 100% means OPA assessed it below.
OPA assessment as a share of our estimate
Cheapest 20%
136% → 146%
46% above our estimate
Lower-middle 20%
109% → 113%
13% above our estimate
Middle 20%
99% → 101%
1% above our estimate
Upper-middle 20%
93% → 96%
4% below our estimate
Most expensive 20%
87% → 91%
9% below our estimate
“Cheapest” means the 20% with the lowest model estimates—not the cheapest recent sales. The chart includes 376,234 homes without a known data warning. We left the other 19% out of this fairness comparison.
The measures answer different questions. Slightly more homes moved closer to our estimate than farther away. Overall consistency improved. The fairness gap by home value grew.
Your home
Check your address before drawing a conclusion.
The citywide pattern does not decide whether one home is assessed correctly.
Think OPA’s value is wrong?
See the appeal steps →Need help paying the bill?
See City tax-relief programs →Methods, technical measures, and limits
We compared 464,760 single-family homes with values in both TY2026 and TY2027. We left known data-warning records out of the fairness totals.
Why use our estimate as the benchmark? On the same out-of-time financed-sales test, our model was less regressive than OPA on all three vertical-equity measures: PRD 1.021 versus 1.065, PRB 0.004 versus -0.057, and VEI 0.7% versus -16.3%. Values closer to 1.000 for PRD and zero for PRB and VEI are fairer by home value. The model is still a check, not ground truth.
The new assessments take effect January 1, 2027. There are no sales from that tax year yet, so we compared both years with the same model estimate, dated July 14, 2026.
OPA’s own and outside sales studies found its citywide 2027 measures within recommended ranges. Our check asks a different question: using one independent estimate for both years, did the result become more or less fair by home value? The two studies use different benchmarks and can reach different conclusions.
Against that benchmark, all three standard vertical-equity measures worsened: PRD 1.107 → 1.114, PRB -0.151 → -0.181, and VEI -50.0% → -60.1%. COD, which measures consistency rather than fairness by value, changed from 21.5 to 21.2.
Model run 20260714T190028Z-baseline · report generated 2026-07-16 · Philadelphia OPA open data.