Underwriting tax abatements and conversion incentives in office-to-residential pro formas

A recurring tension in adaptive reuse underwriting. Nearly every conversion pro forma these days carries a policy line somewhere: a tax abatement, a conversion overlay that waives parking, a density bonus, a state program covering part of the gap for income-restricted units. Sometimes the number is large. On a 140 unit model, that kind of incentive can represent several million dollars of present value, and the deal can fail to work without it. There are two positions commonly argued, both from experienced operators. Underwrite it in. Policy support for conversions is broad and growing, jurisdictions are actively rewriting zoning, parking, and height rules to pull housing out of empty offices, and refusing to count any of it means losing every competitive bid, since other bidders are counting it. These programs typically have application windows and criteria that can be read in advance, so the risk is knowable rather than speculative. Count nothing until it is awarded in writing. Abatements get modified, application windows close, criteria change with a new council, and the programs that survive often carry income restrictions or wage requirements that eat into the underwritten benefit. Timeline matters too: an award landing in month twenty of a twenty-eight month build is worth far less than a static model implies, and models rarely discount for that. If a deal only works with the incentive, the deal does not really work, it has a coupon attached to a political decision. A middle approach carries the incentive at a probability weight, say fifty to sixty percent of face value, sizing the loan off the unabated numbers. It sounds sensible on paper and is rarely seen executed in an actual model, which suggests it may be more often said than done. Program specifics vary considerably by state and city, so anything counted in a model should be confirmed with the administering agency and counsel before it goes in. How this gets carried in practice is worth comparing across operators.