I would be very cautious performing a direct aggregation between census block and watersheds. This is a classic example of the Modifiable Areal Unit Problem (Openshaw 1984; Cressie 1996) issue and could yield unstable results. One option to mitigate MAUP is to use an areal interpolation technique such as dasymetric mapping (Reibel & Agrawal 2007).
USGS Dasymetric mapping ArcGIS tool
http://geography.wr.usgs.gov/science/dasymetric/data.htm
References:
Cressie, N. (1996) Change of Support and the Modifiable Areal Unit Problem. Geographical Systems, 3:159-180.
Openshaw, S. (1984) The Modifiable Areal Unit Problem. Norwich, Geo Books ISBN 0-86094-134-5.
Reibel, M., and A. Agrawal (2007) Areal Interpolation of Population Counts Using Pre-classified Land Cover Data. Population Research and Policy Review, 26(5):619-633