Summary
A GLA adjustment prices a difference in gross living area between your subject and a comparable. Sale price divided by square footage gives you the average value of every square foot, including the land, the kitchen and the roof. Your adjustment needs the marginal value: what the next hundred square feet added. Those are two different numbers, and the second one is always smaller.
The usual way to get this number is the wrong way
A house sells for $400,000 at 2,000 square feet. That's $200 a square foot, so a 200 square foot difference is a $40,000 adjustment. The arithmetic is fine. The rate isn't.
That $200 is an average across everything the buyer bought: the lot, the kitchen, the bathrooms, the furnace, the roof over all of it. None of that doubles when the house gets bigger. A 2,200 square foot version of the same house doesn't come with a second lot or a second furnace. It has more floor area and a little more of the finishes that go with it.
What the adjustment grid needs is what the extra area contributes, and that runs well below the overall price per square foot. Use the average and you overstate the adjustment. You overstate it worst on the comparables that differ most in size, which are the ones the adjustment was there to fix.
Getting a marginal rate
Three ways to do it, and they work best together.
- Paired sales. Two sales alike in everything but size, time-adjusted, price difference divided by area difference. This gives you the marginal rate directly and it is the cleanest evidence available. It's also the hardest data to find, because size comes bundled with almost everything else. See paired sales analysis.
- Grouped medians. Sort your sales into size bands, hold the other features as steady as the data allows, and look at the median price of each band. The step between neighboring bands is a marginal rate. Noisier than a pair, and it uses all of your sales.
- Regression on living area. Fit price against area with the other major variables in the model, and the coefficient on area is what the extra space contributes. Strong when you have enough sales and misleading when you don't. On a thin set it will hand you a confident number fitted to noise, and it won't tell you the model is wrong.
Where two of these land close together, you have a rate you can support and a range to report. Where they disagree, the disagreement usually points to a feature that moves with size, which one method controlled for and the other didn't.
Size comes bundled with everything else
Bigger houses aren't just bigger. They tend to be newer, on bigger lots, with more bathrooms, bigger garages and better finishes. Any rate you pull out of raw size differences is carrying all of that with it.
That cuts both ways on the grid. Take an inflated living area rate out of bundled data, then adjust separately for the extra bathroom and the third garage bay, and you've counted the same difference twice. The result is a grid that overcorrects your largest comparables and reconciles worse than it would have unadjusted.
There are two ways to protect yourself. Match on those other features when you build pairs, so the rate really does come from sales that differ mainly in size. Then run the range test. If the full set of adjustments spread your indications further apart, something is being counted twice. Living area is the usual suspect, because it is the biggest line.
Measure the same thing everywhere
A rate taken from one definition of living area and applied to another is wrong before the analysis even starts. Below-grade finished space is where this usually goes wrong. Some agents count it as living area in the MLS record and some don't, and the form treats it differently again. Finished attics and converted space cause the same trouble.
Decide what counts. Apply that to the subject and to every sale you derive from. Fix the records that disagree, and say which convention you used. Below-grade area belongs on its own line at its own rate rather than folded into living area, because the market prices it differently, often by a lot.
Test it before it goes on the grid
This is the same check that catches most adjustment errors. Look at the range of your comparable indications before and after the living area adjustment. A correct rate brings them closer together. That's what the adjustment is for.
If the spread got wider, the likely causes, in order, are: the rate is an average rather than a marginal one, your set covers two different markets, or another line is doing the same work. All three show up in the comparison. None of them show up in the number by itself.
What to record
- The definition of living area you used, and any records you corrected to match it.
- The sales the rate came from, and what you left out.
- What each method indicated, and the range across them.
- The rate you selected and why it sits where it does.
- The range test, before and after.
Where the data won't carry a rate you can defend, that is a conclusion worth reporting, not a reason to fall back on price per square foot. See sometimes the right adjustment is no adjustment.
Common questions
Is there a typical ratio between the marginal rate and price per square foot?
Rules of thumb get passed around, and quoting one here would mean inventing a number. The ratio varies by market, price point and construction type. Measure it in your own data. Once you have, it makes a useful sanity check on the next assignment in the same area.
Should the rate be the same for a comparable smaller than the subject?
Usually yes, within a reasonable range, since the same market is pricing the same space. When the size difference gets big enough that the two houses appeal to different buyers, the comparable is a weak one. The answer then is a better comparable, not a different rate.
How do I handle below-grade finished area?
Put it on its own line with its own rate, and treat finished and unfinished separately. Fold it into living area at the above-grade rate and you overstate it in most markets. The grid gives it a place because buyers price it differently.
