Page 86 of Harrison & Rubinfeld (1978) mentioned that the authors had hypothesised that PART would affect housing prices but - in the Boston data - it was too highly correlated with NOX to use.
The NOX and PART data come from TASSIM (Ingram and Fauth, 1974), which worked with 122 zones. Harrison & Rubinfeld seemed to then reallocate those data to their 506 census tracts.
Has anyone seen a version of the Harrison & Rubinfeld dataset including PART data, assigned to 506 census tracts?