Well I can't dispute Fred and Hud. I wonder if the discrepancy has to do with the harvard data only considering metro areas, and metro areas having significantly higher incomes but possibly not significantly higher median house price.
If two partners are buying a house together (a fairly common occurrence) per capita ratios would treat both partners as an individual, but a household would help fix for that discrepancy.
Median household incomes take into account households of all sizes (1 person households to n-person households), and if you compare against median price, you can help reduce the risks of outliers tainting any comparison.