I am going to suggest that you use what is called proportional allocation (proportional to area) to estimate the county level crop data for your "sub-watersheds". I am going to assume three things: one is that subwatersheds and counties do not have coterminous boundaries (i.e. sub-watersheds are simply not a few counties concatenated together), two both supports are polygons, and three, sub-watersheds are completely covered by the counties layer. (Even if some of these aren't true it may only slightly change your approach anyway).
Basically what I am going to have you do is find out the proportion of county x that is in sub-watershed y. You then calculate the product of whatever variable you are interested in in county x and that proportion. You then sum all of these areas (county x's) that fall within sub-watershed y.
In arcmap, after you have both shapefiles in your map document, open up the arctoolbox. Within the toolbox navigate to Analysis Tools -> Overlay -> Intersect. In the intersect GUI, place both your sub-watersheds feature and your counties feature, and then designate the output shapefile. This output shapefile will essentially be all of the slices of sub-watersheds and counties that fall on top of each other (if you expand the help in the intersect GUI it has a picture of what I mean).
In this new shapefile you just created, you will have all of the attributes listed for both the original watersheds and county files, as well as the area of the polygons in the new layer. What you then want to do is add a new field, and say X is the variable you want to allocate to subwatersheds, the field calculation syntax will look somthing like this, (I did a test run and on my machine it just named the new area Shape_Area)
What this does is weight the value of X according to the area proportion that polygon is in relation to the original county.
After you calculate those fields for all of the variables you are interested in, you will then sum those fields to the sub-watershed level. If you right click the unique identifier field for your sub-watersheds (still working within the intersected shapefile), click summarize, then define sum for each of your new fields you just calculated, and then specify the output dbf file. You can then use a table join to join the information within that dbf and your original sub-watershed shapefile.
As an aside it is good to know that this technique assumes that all of the variables are equally distributed within counties. This may or may not be reasonable, and depending upon how small counties are and how great the spatial mis-match is such an assumption may only cause trivial error anyway. (Also if counties do not entirely cover sub-watersheds say so, as you may want to calculate the density of your variable within that missing region).
Of course their are other alternatives, I can think of 1 easier solution and 2 more complicated ones. The easier solution would be to calculate the polygon centroids and just use a spatial join to aggregate county variables within sub-watersheds. I do not suggest you do this unless the boundaries between watersheds and counties are damn near coterminous.
Other suggestions may be to either use aerial imagery/remotely sensed data that has smaller units than counties. It wouldn't surprise me if their was relevant information for your project in this raster based format (such as land use cover). Another option would be to use geostatistical kriging techniques to estimate your variables.
If you have any other questions feel free to ask.