I just completed an appraisal of a home for a refinance where the subject had sold eight months prior to date of value. I included the prior sale of the subject as Comp#4 and had two very recent, reasonably similar comparable sales that I included in the sales comparison grid. The subject is located in a suburban neighborhood built out in the 1960s-1970s on a ridge with some homes featuring city light views. Many recent sales feature remodeling or recent updating while others are relatively original.
In most suburban neighborhoods, using sale price per square foot (PSF) is the best way to track market trends because using PSF reduces the impact of changes in floor area in your sales comparable data. Generally speaking, the sale price model is subject to variations in floor area in this case and is not the best model. So, following my normal practice, I used the PSF linear regression model to determine time adjustments to comparables.
The PSF linear regression formula is:
Slope of the Trendline * Subject’s GLA = Daily Time Adjustment
So, in this case:
0.0992 * 1794 sf = $178/day
Once I applied these time adjustments, my range of value was from $447,000 to $466,000 with an indicated value of $460,000. However, the most recent sale indicated $450,000 at most as did homes in contract competitive with the subject. Something wasn’t right.
So I tried a sale price linear regression model instead.
Sale Price Linear Regression Model:
Slope of the Trendline = Daily Time Adjustment
In this case, indicated daily time adjustment is $117/day or
Indicated range of value for the subject ranged from $443,000 to $451,000, in line with the most recent sale and competitive homes in contract.
In this case, the Sale Price Linear Regression Model explained my data better than the PSF model.
I’ll have posts explaining more about time and real estate in the near future and will be teaching a class this July in Sacramento. Or you can learn straight from the national expert, George Dell. Sign up for his website and classes here. Thanks George for sharing with the rest of us.
Why do you think the linear model worked better in this case?