The IRS publishes a dataset called Statistics of Income Migration. It tracks county-to-county flows of tax filers, including their aggregate Adjusted Gross Income. It is the closest public-records analog to seeing the wealth map redrawn in real time. Almost no one uses it for CRE.
When you do — and especially when you join it to local earnings (ACS PUMS), housing cost (HUD FMR + observed rents), education quality (NCES + GreatSchools), and walkability (OSM POI density) — you get an arbitrage signal: neighborhoods receiving disproportionate high-AGI inflow where the cost-of-living number hasn't moved yet.
The flow signal: who's moving where, with how much
The IRS publishes net flows at the county level. We disaggregate to ZIP using ACS commuting and worksite data, then to H3 cells using NCOA address-change density. The output is a per-cell estimate of: net filers in/out, average AGI of inbound vs outbound, and the AGI delta — which is the actual signal.
The AGI delta — average inbound AGI minus average outbound AGI — leads observed rent change by ~14 months on the median, with correlation 0.71 across the 22-metro panel.
The triangle: education × walkability × wage
Migration data on its own is too coarse. The high-AGI inflow chooses among cells inside the destination metro using three filters: walkable amenity density, school quality (where children are involved; less so for early-career), and wage opportunity (job-posting density).
The cells where rents move fastest aren't the trendiest by individual metric. They're the ones balanced across all three. Single-dimensional 'top walkability' or 'top schools' lists are noise compared to the intersection set.
Named examples
| Cell | Inbound AGI | Outbound AGI | Delta | Triangle rank |
|---|---|---|---|---|
| Atlanta — West Midtown | $167K | $92K | +$75K | Top 5% |
| Denver — RiNo / Five Points | $148K | $84K | +$64K | Top 5% |
| Raleigh — North Hills | $152K | $98K | +$54K | Top 10% |
| Philadelphia — Brewerytown | $118K | $74K | +$44K | Top 15% |
| Cleveland — Tremont | $104K | $71K | +$33K | Top 15% |
| Detroit — West Village | $96K | $68K | +$28K | Top 20% |
These cells have one thing in common with each other and nothing in common with the metro's headline narrative. Detroit is supposed to be flat; Cleveland is supposed to be declining; Philadelphia's story is uneven. The IRS data doesn't care about the narrative. It tracks the cash.
Where this signal fails
Three failure modes worth naming. First, retirement metros: high inbound AGI from retirees doesn't translate to rent pressure because the move is final, not transitional. Second, single-employer towns: a corporate relocation can spike the AGI delta in a cell without representing organic flow. Third, college towns: graduate students show up as low-AGI inbound, masking the real underlying earnings power.
The Explorer flags these structural anomalies rather than silently filtering them. An analyst looking at the Sun Belt retirement belt sees the IRS signal AND the retirement-flag note that explains why it won't translate to rent.
Why this isn't priced in
Two reasons the arbitrage persists. The IRS publishes with an 18-month lag; the data exists but most operators ignore it because it's not Q-over-Q fresh. And the disaggregation work — county-to-ZIP-to-hex — is non-trivial. Most platforms stop at the county level, which is too coarse to act on for site-specific underwriting.