Methodology

Every number on this map is computed from public parcel data with a transparent formula. Here is exactly how, and what the limits are.

The data

Parcel records come from Regrid for tax year 2021, City of Ann Arbor. We include only taxable parcels — residential, commercial, industrial, agricultural, and their vacant variants — with a positive assessed and taxable value. Exempt, municipal, and zero-value parcels are excluded because they carry no levy.

Key terms

• SEV (State Equalized Value) — the assessor's value, by law about half of market value. We show estimated market value as SEV × 2 (an approximation; assessors do not hit exactly 50%).
• Taxable Value (TV) — the Proposal A–capped base your tax is actually charged on. Proposal A (1994) limits how fast TV can grow each year and resets it to market only when a property sells ("uncapping").
• Prop A discount — (SEV − TV) / SEV. How far the capped base sits below assessed value.

Modeled tax

The data has no per-parcel tax amount, so we model it: tax = TV × millage, using the 2021 City of Ann Arbor combined rates — 52.02 mills for principal residences and 70.02 mills for non-homestead property (the homestead rate plus the 18-mill school-operating tax that principal residences are exempt from). These are the all-in rates (city, county, schools, library, transit, college, state education).

We calibrated and checked these against real tax bills from the county's BS&A system: three principal-residence parcels spanning a 5× range of value all reconcile to within 0.01% of their actual 2021 tax. A commercial sample runs about 8% under our model because large downtown parcels carry special assessments beyond the base millage — so treat non-homestead and commercial dollar figures as slightly conservative.

The revenue-neutral "fair rate"

We model a single reform: remove only the Proposal A cap, tax every parcel on its full SEV, and reset the rate so the total amount collected is unchanged. We do not touch the homestead exemption — that is a separate policy this project does not contest. Concretely, within each class we scale by k = ΣTV / ΣSEV and compute fair tax = SEV × millage × k. Because the rate is reset downward to be revenue-neutral, this is not a tax increase in aggregate — it is a redistribution. In 2021 the multipliers were k = 0.747 for principal residences and k = 0.719 for non-homestead property.

A parcel's fair-rate change is fair tax − current tax. Positive means it is taxed below a fair rate today (a beneficiary of the cap); negative means it already pays more than a fair rate would charge — often a recently-sold property.

Limitations

• Single snapshot. 2021 only. We cannot show how a parcel's gap grew over time, only where it stands.
• No sale dates in the dataset. The size of a parcel's current discount is our proxy for how long it has been capped.
• Modeled dollars. Tax is computed from millage, not read from bills; treat dollar figures as close estimates.
• Blended rate. We use one combined rate per class rather than modeling each overlapping taxing jurisdiction separately.