I looked over the 'working paper' and what struck me is the amount of regurgitation of numbers, over slicing and dicing. But a couple of fundamental problems with the analysis, again its based on 'census tracts' and appraisers don't define neighborhoods by census tracts and every report in the last twenty years should name the streets. So it would be beneficial for the authors to do the work right. Second, there seems to be a conclusion of 'correlation' but I couldn't see a stated R2 (so flawed statistical analysis) or hiding the validity of the analysis).
And as Judges are wont to say, "Appraisal isn't science, its as much art as science."