You are right, Stan, when I run it only for homes priced at less than $600k, the correlation is a lot higher (r-squared = 43% for all San Diego county). But since I’m charting the pr/sqft for all houses, and there is no correlation for the dataset as a whole, I don’t think this particular issue presents any big distortions. (At least, any distortions from this issue are probably minor in comparison to some of the other distortions caused by the fact that square footage is an extremely rough proxy for overall quality — a square foot in La Jolla is different than a square foot in La Mesa, and all that.)
