[jyl]

Just because I am a geek, I collected data for 66 possible comparable houses. These were all arguably legitimate comps: same desirable residential neighborhood (all w/in 10 block radius), similar age of house (1910 to 1920s), similar sq ft (avg 3300 sq ft) and #bedrooms (avg 4), similar views (none), all sold w/in past 12 months (4/05 to present).

The data I collected was: date of sale, total listed sq ft, #bedrooms, #bathrooms, sq ft main floors, sq ft finished attic, sq ft finished basement, sq ft unfinished space, sq ft garage, lot size, sale price. Then I ran some regressions to see if there is a "scientific" way to predict the sale price of a house.

I was a bit surprised at which variables were statistically significant and which were not. Care to guess?

Okay, the ONLY variables which were actually statistically significant (t-stat > 2) were date of sale and #bathrooms.

Any real estate pros have a reaction to this? Is this why we see so many new houses with more bathrooms than bedrooms?

P.S. The R-square of the best regression was only 0.55.