Accounting for unequal sample sizes - Crosstabs Chi Square?

If you are using statistical software, this is pretty straightforward. One independent variable is zip code, one is time period. The dependent variable is enrollment.

Thanks, that's actually what I did, but it appears that the unequal sample sizes aren't being accounted for (unless it's done automatically). To be more specific, I'm leaving all the cells with missing data in the analysis for the second time period. The output graphs shows the opposite of what I would expect. I had 54 cases respond to every question in time period one, but only about half of them responded in time period two (and I'm leaving all the cases in the analysis for time period two, even those that we could not reach...so all of the variables for each of these cases is blank).

How I would do it: set up a table that looks like:

Person, Zip Code, Time Period, Food Stamps
------------------------------------------------
1, 70001, old, no
2, 70002, new, yes
...

You should have around 80 records, 54 old and around 26 new. There should be no blanks. Then build a model with Zip Code and Time Period as the independent variables and Food Stamps as the dependent variable. Ensure that zip is set up as a nominal value instead of a continuous variable.

Thanks! Tried it. Regardless if I eliminate the blanks or not, I get the exact same result. So I guess leaving in the blanks for those that didn't respond in the second time period doesn't matter. But the results are still odd and opposite of what I expect.

Sometimes a good experiment will do that to you.