Value landlike land.

How the numbers are made

The records

OPA's certified roll: one row per parcel, with its land and building values, category, exemptions and abatements.¹ "Built properties" are parcels with a building value (541,000 of them). "Abated" means a full or partial construction abatement on the roll.³

The one-fifth rule

A built property is counted at the rule when its land, divided by its total assessment and rounded to four decimal places, is one-fifth. 9 in 10 are.

Lockstep, the abatement premium and matched pairs

Each compares like with like inside OPA's own market areas. The correlation is taken among built parcels in the same market area and the same quarter of that area's lot sizes, and the figure is its median across those groups. The premium is abated parcels' median land value per square foot over unabated built parcels', in the same market area, and the figure is its median across market areas. The pairs match each vacant lot to the nearest built lot within 150 meters whose lot area is within 20% of its own.²

What abated lots sold for

Arm's-length deed transfers⁵ of lots sold vacant and since built on, or bought as teardowns, validated using OpenAVMKit's sales tools.¹⁶ A sale counts only where the same lot, its area within 10%, is on today's roll: 814 lots, 684 of them abated. Each ratio is OPA's land per square foot over the lot's own sale price per square foot, brought to the valuation date with a repeat-sales index: the change in price of the same single-family home sold twice, at least three years apart, across the city, by quarter, leaving out homes first sold while a new building's ten-year tax abatement was running (the buyer pays for the years left, so those pairs would read as slower growth).⁵ One index for the whole city moves every ratio together, so it can shift their level but not the gradient across neighborhoods. Across the six adjustments tested (this one, pairs held a shorter time, a separate index for each fifth, an earlier valuation date and OpenAVMKit's own index), the most non-white fifth's ratio runs from 0.97× to 2.01× and the whitest's from 0.52× to 0.82×, and the first is 1.6 to 2.4 times the second.²

Neighborhood fifths

Census block groups ranked by non-white share of residents and by median household income,¹⁴ cut so that each fifth holds a fifth of the city's taxable homes. These are neighborhood averages, not households.

Land valued like vacant land

Each built parcel's land is the median value per square foot of the ten nearest lots OPA assesses as vacant, weighted toward the closest, times its lot area. Lots larger than 10,000 square feet, which few land sales test, are scaled by how far the method misses on the sales of such lots. Vacant lots keep OPA's value.²

The average miss

Scored on 1,114 sales of lots since built on or torn down, each predicted without the assessment of the parcel that sold. The coefficient of dispersion is the average gap between each value-to-price ratio and the median ratio, as a share of the median,⁶ here with the most extreme ratios at each end trimmed. Some of these sales may have informed OPA's vacant values, which would flatter this method and OPA's roll alike.

The bills

Each built parcel keeps OPA's total assessment, re-split into land and building on the new land value, at today's rate of 1.3998% for the City and School District.⁸ Because the total and the rate stay put, a bill moves only where the split matters: chiefly abated properties, which pay on land. The Homestead Exemption comes off the total, so a home that only has the exemption keeps its bill.⁷

The tax the fix moves

On each abated property, the bill at today's rate on the new land, against today's bill. Bills that fall and bills that rise are summed separately, so $29 million is the tax that changes hands and $3.2 million is what is left over. On new construction alone, leaving out rehabilitation, it is $22.0 million. The fix misses land sales too (37% on average), so part of that sum is its own error: it is what the fix moves, not a count of what OPA gets wrong.²

Homes the rule takes back

PPI's model of whether owners build goes lot by lot through private land where housing is allowed, vacant or worth rebuilding, and asks how likely its owner is to build, with and without today's abatement, and with land valued by OPA's rule (a fifth of the total, a tenth for a newer apartment building of five or more units) or by the fix. How strongly building answers profit is calibrated to the building permits of 2023–25.⁴ The share is what the abatement adds over 20 years with land valued by the fix, less what it adds under OPA's rule, over the first. The model's two least certain settings, how widely outcomes vary and how strongly building answers profit, move both sides alike, so the share hardly moves with them: 6.7% to 6.9% across the range of variation between owner groups, and 5.9% to 8.4% across every other setting tried one at a time. The number of homes moves far more, so it is not quoted. Under the original abatement, which fully exempted a new building for ten years and when OPA set land at a fifth on apartment buildings too, the rule took back 13.6%. The model credits that abatement with 12.1% of construction on the permits of 2010–2019, well below the 30–40% of development a study for the City found could have been lost or delayed without it.¹⁷ A stronger response to the abatement would raise both sides alike and leave the share where it is.²

The land value tax

Land taxed at four times the building rate, with the two rates set to raise today's revenue, after today's abatements have run out, and with the Homestead Exemption taken from land and building in proportion to value, which state law would need amending to allow.¹⁸ Each chart bar is the change in homes' total bills in that fifth.²