20% +/- Square footage "Rule" for Comparing Comps

[Artemis Fowl]

Hmmm. Well aside from whatever guidelines you may be working with, the only rule I believe is that you get a good model that can adjust for the GLA range you want to use. For example, let's say you have 120 sales comps for your area for the past two years, you run MARS or Earth and get an R2 of 0.80 on a GLA range of 800sf-3500sf, first run. You score the comps based on first model, create the residuals, rank and score the residuals and then rerun MARS on the residual scores. You will then likely get an R2 of 0.98 or better. Then you select your best 12 comps based on all factors, in particular Quality/Condition/Appeal/etc - really whatever your model tells you are the most important features for value. After running adjustments against the subject, you get 3-4 comps within 0.5% of the average, 3-4 comps between 0.5 and 1% of the average and the rest between 1 and 2% of the average. If your comps range from 800sf to 3500sf, it's demonstrating the power of your model, it's confirmation you know what you are doing. .... IMO JUST REMEMBER: Same treatment for all comps, - be consistent.

Well the OP shouldn't have any questions after that...

 

[ZMonet]

Well the OP shouldn't have any questions after that...

Thanks for everyone's input. I didn't realize that it wasn't so clean, but that is understandable. In this case, the subject property is 4800 square feet and COMP#1 is 3100 square feet and COMP#2 is 2800 square feet. "Comps" are within six months, There is an almost exact copy of the subject (across the street) that is within 5% of size, but the sale is two years old. Does the two-year-old sale carry any weight in a market that hasn't really changed that much in the last 5 years? Thanks again!

 

[Overimprovement]

Since it sounds like there are few 'good' comps for the subject, I would be inclined to at least consider the 2 year old sale, if not give it a good chunk of weight.

50% of my reports are rural, so I may be more used to going back in time than others, who may have those 6 sales in the last 36 hours on the same block. Many large metro areas have subdivisions which frankly approach that level of comp saturation--cookie cutter homes, active market. For those, 20% may be high, yes, but that doesn't sound like your scenario at all.

If the median GLA for that area is 4800, then there should be some other more similar sales out there. If the median GLA is much lower, then the report may truly be using the best, most recent sales available. Still, I would consider the 2 year old sale from the info known from these posts.