Tract-level or neighborhood-level data on Race for 1940/1950

I am looking to construct a dissimilarity index by race for all (or as many as I can manage) cities or metropolitan areas for 1940 and 1950. I would prefer to use census tracts or “neighborhoods” as the unit of analysis within the city but I’m having trouble finding the appropriate data. All I need is population for each race by tract (or neighborhood) within a city. Consistent boundaries between years would be nice but not essential. I understand the use of census tracts were not widespread until 1940/1950 (not sure exactly on this) so maybe it’s not the best geography, but I thought I would ask - does any source do this or is this even possible? It seems relatively straightforward but for some reason I can’t find what I’m looking for.

What I’ve found thus far:

  • NHGIS has a table using census tracts for a very limited number of cities for 1940 and 1950 (although these may be the only cities that were divided by census tract at that time? I haven’t been able to find this out).

  • IPUMS has a ward-level sample for 1940 and a enumeration district-level sample for 1940/1950.

Enumeration-level seems like it would be too small a geography, especially for the big cities where an ED may only be one block of a city. Ward-level seems a bit better although this might be poor for smaller cities - In general I’m having trouble finding information on how wards were set and how they change (if they do), so if anyone has any tips on this it would be greatly appreciated.