Land Value for Rental Property Depreciation Is the Number Nobody Checks.

An investor who closed on a duplex in January sent me his first return for a second look. Schedule E showed $18,296 of depreciation on a $525,000 purchase, which is the whole price over 27.5 years. There is a lot under that duplex, and no part of it depreciates. The land value for rental property depreciation is the one input nobody checks, it is fixed the day you close, and it rides every year for the next 27.

Land is not depreciable. Treas. Reg. §1.167(a)-2 allows depreciation on the improvements and physical development added to land but not on the land itself. Buy both in one transaction and Treas. Reg. §1.167(a)-5 caps depreciable basis at the same proportion of the lump sum as the depreciable property's value bears to the whole property's value at acquisition. Capitalized closing costs ride in that same ratio. Basis is then fixed at acquisition under §1012, so this is a one-time decision that compounds through the 27.5-year recovery period in §168(c), the repair ceiling you get each year, and the character of your gain on the sale.

The modern case is Nielsen v. Commissioner, T.C. Summ. Op. 2017-31. The taxpayers allocated effectively all of their Los Angeles County purchase to improvements, called the assessor's data extraordinarily inaccurate, and offered a land sales method and an insurance replacement method. The court used the county assessor's allocation and said it could find no authority that a taxpayer is qualified to allocate value between land and improvements without supporting evidence. Summary opinions carry no precedential weight under §7463(b), so this is not law. It is the IRS's opening position.

The other direction exists. In Meiers v. Commissioner, T.C. Memo. 1982-51, the IRS allocated solely by the local assessor's relative valuation of two condominiums and the taxpayer allocated by an investigation of building replacement costs. The court went with the taxpayer, allowed 80% to the buildings, and called the question purely a factual one. The IRS's own position fits: you cannot allocate solely by assessed values when better evidence of fair market value exists. Neither case is about the method. Both are about the file.

So rank your support by what is in that file. Best is three or four sales of comparable vacant lots in the same neighborhood, dated near your closing. Next is an appraisal stating a separate site value, which a lender's appraisal usually does not. Then the assessor's ratio. Last is a percentage somebody called standard, which is what the Cost Segregation Audit Techniques Guide in Publication 5653 tells examiners to look for. Use the assessor's ratio, never the dollars: a Utah notice stating taxable value rather than market value gives the same ratio anyway, because the 45% residential exemption in Utah Code §59-2-103 reduces the dwelling and up to an acre of land together.

Almost everything written on this treats the non-land half as one bucket. It is two. Site work is land improvements: the driveway, the flatwork, the fence, the retaining wall, the sprinkler system. Asset class 00.3 of Rev. Proc. 87-56 gives them a 15-year recovery period, inside the 20-year ceiling for bonus depreciation, and §168(k) is back at 100% for property acquired after January 19, 2025. On this duplex that is $38,000 deducted in 2026 instead of $1,900. Used property qualifies, so whoever poured the driveway in 1998 is not your problem.

Two honest limits. The carve-out needs support of its own, which in practice means a cost segregation study or a contractor's itemization, and a percentage off a website will not hold. And a deduction is not a refund: the loss meets §469 before it meets your W-2, and a household over $150,000 of modified adjusted gross income without material participation carries it forward. Whether a study pays for itself here is is cost segregation worth it on a single rental, and whether the loss gets through at all is deducting rental losses against W-2 income.

Depreciation comes back. Straight-line depreciation on the building returns as unrecaptured §1250 gain at a rate capped at 25% under §1(h)(1)(E). The site work is worse on exit than it looks: most land improvements are §1250 property, and bonus depreciation on them is depreciation in excess of straight line, which §1250(a) recaptures as ordinary income to the extent of gain. So moving $90,000 out of land is a rate and timing trade, 28.45% now against 25% or ordinary rates later. Still the right trade for most owners, but it is not free.

There is a quieter benefit on the other side. The small-taxpayer safe harbor in Treas. Reg. §1.263(a)-3(h) deducts repairs up to the lesser of $10,000 or 2% of the building's unadjusted basis: $6,300 at $315,000 of building, $7,340 at $367,000. The land number sets your annual repair budget, which is the argument in repairs versus improvements on a rental.

One return out: amend it before the next one is filed. Two or more consecutive returns on the same impermissible schedule: you have adopted a method of accounting under Rev. Rul. 90-38, and the fix is Form 3115 with a §481(a) adjustment, the same machinery as catching up missed depreciation. Depreciating land runs that machinery backward, though. The §481(a) adjustment is income rather than a deduction, and under Rev. Proc. 2015-13 a positive adjustment is spread over four years, the year of change plus the next three, unless it comes to less than $50,000 and you elect to take the whole thing in the year of change.

In multi-year reviews the single wrong line is rarely the exposure. Inconsistency is. I have reviewed returns where the years the owner prepared himself were cleaner than the year he paid a service for, and the only finding worth real money was that the method moved year to year. Pick a defensible figure, write down where it came from, and use it every year.

I would spend $500 on the land support and take the $120,000. Three vacant-lot comps and a one-page memo is a few hours of work, and against the assessor's $210,000 it buys $3,272 of deduction a year for 27 years, about $931 a year at 24% federal and 4.45% Utah. Hard return to beat on a $500 document. Under roughly $300,000 of price, use the assessor's ratio and move on.

What I would not do is pick the land number because it is low. Zero to land is the Nielsen fact pattern, and a thin land share where dirt sells for real money will not survive the first question, and the examiner who replaces your figure will not pick a generous one. Price the lot, then do the arithmetic.