Estimating Median Household Income by Block

I'm trying to make a higher resolution map of median household income. The Census Bureau releases income data by census tract (around 4000 people) and also by block group. Relatively few people use block groups because the confidence interval is too high. Even when you pool five years of data from the American Community Survey the confidence interval for census tracts (which have about 3 times more people than block groups) is high (and can be too high to use).

So I'm trying to estimate the income for census blocks by using race. I tentatively assumed that the difference between income for different racial groups in a census tract would be equal (in relative terms - eg. as a proportion) to the differences in US national median household income by race. For instance, if Hispanics are earning 85% of the national median - then in a census tract they'd earn 85% of the census tract's median IF the census tract had the same demographics as the US. (In practice, I'm guessing that the racial income inequality within a census tract is smaller.)
Estimated Block Income Map

I might also try to use smoothing (so that a block which is on the edge of a census tract is influenced by the income of the surrounding tracts), using block groups as my starting point, or another method.