This repository was archived by the owner on Jul 2, 2026. It is now read-only.
VA state dataset overestimates high-income tax units by ~4x (impacts HB979 scoring) #548
PavelMakarchuk
started this conversation in
General
Replies: 2 comments
Update: Ultra-High Income Analysis Across VA, AZ, and NYCritical finding: All three state datasets cap out at ~$6.5MNo PE state dataset has a single record above $10M in taxable income. The entire ultra-high-income tail is missing.
Top 20 records show hard ceilingAll three states have the same pattern — taxable income clusters in a narrow band near the maximum and then stops:
This looks like a systematic cap in the Enhanced CPS imputation — possibly CPS top-coding propagated into the state datasets. Comparison with actual tax return data$1M+ filers (PE vs actual):
$5M+ filers (PE vs actual):
$10M+ filers (PE vs actual):
The double distortionThe PE datasets have two simultaneous problems in the high-income tail:
This means:
Revenue impact on pending bills
Actual ultra-high income data (from state agencies)Arizona (AZ ADOR, TY 2022, residents only):
New York (NY DTF, TY 2022, all filers):
Note that NY's $10M+ filers alone hold $127.4 billion in income and pay $6.1B in state tax — and PE has zero representation of this group. Sources |
0 replies
Data sources used in this analysisPE state datasets
IRS Statistics of Income (SOI)
State tax agency data
News / policy org sources (VA-specific)
NYC-specific
Other
|
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Summary
While scoring Virginia HB979 (new 8% and 10% income tax brackets on high earners), we found that the Enhanced CPS state dataset for Virginia significantly overstates the number of high-income tax units, which inflates revenue estimates from high-earner tax provisions.
Context: state-legislative-tracker PR #70
The discrepancy
Source for actual data: Cardinal News citing VA tax return data (13,170 out of 4,063,070 returns report income over $1M).
Granular income distribution from PE's VA dataset (2027 projection)
Key observations:
Revenue impact
This overcount directly inflates revenue estimates for any policy targeting high earners:
The scaling ratio (13,170 / 56,413 = 0.23) is a rough adjustment -- the true correction depends on whether PE is overweighting existing high-income records vs imputing income that doesn't exist. But the magnitude of the discrepancy (4.3x) suggests this is a material data quality issue for high-income policy analysis.
Marginal income in each new bracket
For the HB979 reform specifically:
The >$1M bracket captures $112B in marginal income. If the actual figure is closer to $112B * 0.23 = ~$26B, the revenue from this bracket drops from $4.76B to ~$1.1B.
External validation points
PE's bracket-only estimate is 1.9x to 2.8x higher than external estimates.
Questions
All reactions