In your difference$-$in$-$difference analysis of the "policy" that affects college graduates, you $($should have$)$ used information about non$-$college graduates, even though they aren't affected by the policy. Why would we use information on an unaffected group?