Summary
Site value comes from vacant land sales where you have them, and from allocation or extraction where you don't. The value of the whole site and the rate the market pays for extra lot area are two different numbers, and the second is much smaller. Using the first as a lot-size adjustment is one of the biggest errors you can make on the grid.
Two questions that get one answer by mistake
Site value comes up twice in an appraisal, and it means something slightly different each time. The cost approach needs the value of the whole site as if vacant. The sales grid needs what the market pays for a comparable's lot being bigger or smaller than the subject's.
Those are related, and they aren't the same number. A quarter-acre lot worth $90,000 doesn't mean another tenth of an acre is worth $36,000. In most residential markets it is worth a small fraction of that. What the buyer is paying for is a location with a usable building site, and area beyond what is usable adds very little.
The first is a rate across a whole site. The adjustment needs a marginal rate. Use one for the other and you inflate the lot line badly on any comparable whose site is much bigger or smaller. It's easy to spot in review, because the implied land value soon runs past the value of the whole property.
Vacant land sales, where you have them
This is the main method and the most defensible. Sales of vacant lots in the subject's market, time-adjusted to the effective date, screened for comparable zoning, utilities, topography and buildability.
A few things worth naming in the file:
- Land sells rarely, so your set often covers a longer period than your improved comparables and needs more time adjustment, not less.
- Land moves differently from improved property. Applying your improved-sale trend to land sales is an assumption. Where you can measure land separately, do it.
- Buildability matters more than area. Two lots of the same size with different setbacks, access, utilities or floodplain aren't comparable, and area is what most often hides that.
- A teardown isn't a land sale unless the improvement really did contribute nothing. Treating it as one needs a sentence of justification.
When there are no vacant land sales
In a built-out neighborhood there may not have been one in years. Two accepted methods fill the gap, and both should be labeled for what they are.
- Allocation. Apply a land-to-value ratio observed in a comparable market where land does sell. Simple, and only as good as the similarity between the two markets. Derive the ratio and cite it rather than assuming it.
- Extraction. Take improved sales, subtract the depreciated cost of the improvements, and call what is left land. This works where the improvements are fairly new and depreciation is small. It gets less reliable as depreciation grows, because every error in the depreciation estimate lands entirely on the land figure.
Both are inferences from something other than land sales, and both deserve a stated range rather than a single number. Where allocation and extraction land near each other, the conclusion is reasonably supported. Where they split, report the split.
Deriving the lot-size adjustment itself
Once you have site value, the grid still needs the marginal rate. Derive it the way you derive every other physical rate: from the market, not from the total.
- Paired sales that differ mainly in lot size, time-adjusted, with the improvements matched as closely as the data allows.
- Grouped medians across lot-size bands within one segment, reading the step between neighboring bands.
- A regression including site area alongside living area and the other major variables, where your set is big enough to support one.
Expect a modest number, and expect it to flatten out. The market usually pays real money for the difference between a small lot and a typical one, and very little for the difference between a typical one and a large one. One flat rate across that whole span overstates your large-lot comparables. Where the data shows the flattening, band the rate and say so. Where it doesn't, keep the adjustment inside the range your data actually covers.
Excess and surplus land is its own case. Area beyond what the market will use isn't priced like the main site, and area that could be developed separately may be priced far higher. Neither is captured by a rate per square foot.
Keep the two figures consistent
The site value in your cost approach and the reasoning behind the lot line on the grid should read as one analysis. A reviewer comparing them is checking exactly that, and an unexplained mismatch is one of the easier findings to make.
Derive it once and carry it to both places, with the marginal rate stated separately from the total, and the problem goes away. More on that pass in writing support a reviewer can follow.
Common questions
How old can a land sale be before it is unusable?
Age isn't what disqualifies it. An adjustment you can't support is. An older sale you can time-adjust with evidence is more useful than a recent one from a different submarket. Where you can't measure the land trend, say the adjustment carries more uncertainty and weight the conclusion accordingly.
Can I use assessed land-to-value ratios for allocation?
They are a starting point and a weak source. Assessment practice varies, and the ratios are often stale or formulaic. If you use them, put them alongside a market-derived ratio as a check and say where they came from.
What if site value exceeds what extraction supports?
That usually means the improvements carry more depreciation than your estimate allowed, or the market is pricing redevelopment potential the cost approach can't see. Either is worth looking into rather than reconciling away, and it often changes the highest and best use analysis.
