Allocating purchase price to land: the first depreciation decision

Land does not depreciate — Treas. Reg. §1.167(a)-2 has said so since the beginning — so the first tax decision on any property purchase is how much of the price is land. Buy a rental for $1,000,000 and the land allocation directly sets the depreciable base: a 20% allocation leaves $800,000 to recover; 40% leaves $600,000. Nothing else in the depreciation chain — recovery periods, bonus, cost segregation — matters until this split is defensible, because every downstream deduction is computed on the building number.

The accepted methods

The regulation's rule is relative fair market value at acquisition (Treas. Reg. §1.61-6(a) for the allocation principle; Publication 527 restates it for landlords). Three evidentiary routes, in ascending order of strength:

1. Assessor ratio. Apply the county assessment's land-to-total ratio to your purchase price. Cheap, contemporaneous, third-party — and the method courts and examiners most often accept as reasonable, even while acknowledging assessments lag markets.
2. Qualified appraisal. A purchase appraisal that separately states land value is the strongest support, especially where assessments are stale or the market's land share is unusual.
3. Replacement cost. Value the improvements at depreciated replacement cost and treat the residual as land — common inside formal studies, useful as a cross-check.

What fails: copying the seller's schedule, using the mortgage appraisal's insurance value (which excludes land entirely and proves nothing), or picking a percentage because it was fine on the last deal three states away.

Illustrative land shares by setting — local evidence controls, not these ranges.

Rural / industrial10%

Suburban rental25%

Urban infill40%

Coastal high-value60%

Illustrative mid-points; actual ratios vary widely within each category and by market cycle.

Why aggressive allocations sink studies

A cost segregation study reallocates the building basis among 5-, 15-, and 39-year classes — it inherits the land split rather than fixing it. An examiner working the Cost Segregation Audit Techniques Guide checks the land allocation early, because it is the easiest adjustment in the file: no engineering dispute, just an assessor printout showing 35% land where the return claimed 15%. The adjustment then cascades — every class's basis shrinks pro rata, bonus depreciation reverses in part, and the understatement can carry accuracy-related penalty exposure. Quality studies document the land number for exactly this reason; a study that accepts the client's unexamined split is a marker of the study-quality problems worth screening for.

Land improvements are not land

The intro-level trap runs the other direction too. Paving, curbs, sidewalks, fencing, exterior lighting poles, storm drainage, and landscaping are land improvements — 15-year MACRS property under Asset Class 00.3, eligible for bonus depreciation — not part of the non-depreciable land at all. A bare land/building split typically leaves them stranded in the land bucket; identifying and pricing them is one of the core deliverables of a study, detailed in land improvements explained. Only genuinely permanent site work — general grading, clearing — stays with the land. The classification rules and recovery tables are in Publication 946.

One last neutrality note: sometimes the honest land number is simply large. High-land-value markets are where depreciation strategies underperform on arithmetic alone — a 60%-land San Francisco duplex has little basis to accelerate, and that fact belongs in the is-it-worth-it analysis before any study is commissioned.