Alternative value estimates of owner-occupied housing: Evidence on sample selection bias and systematic errors