Specifically, our model
has a standard deviation in house price valuations
of +/-15 percent, meaning that any valuation
between 15 percent overvalued and 15 percent
undervalued should be considered statistically
normal.

Basically they are saying that because historical valuations swing so wide, we cant control for it. So a 500k house could be 575k and still be considered fairly valued. It could also be that a house at 425k would meet the same valuation of “fair”. So basically they cant differeniate between 425k and 575k as that falls in their margine of error. That is a rather wide spread don’t ya think?

Plus, just a guess, but they are most likey using the median house price, and we all know the problems with using that as your gage. (if I am wrong please correct, I couldnt find how they identified a price in each area)