What explains the behavior of legislators on bills that restrict the rights of marginalized minorities? Studies of representation typically focus on factors like party or public opinion but seldom account for theories of minority representation like electoral capture or subconstituency politics. One reason for this is that data allowing for the comparison of these theories are seldom available for U.S. House districts. We overcome this hurdle by implementing multilevel regression with post-stratification to estimate opinion on gay marriage during the 1996 Defense of Marriage Act vote. We show that subconstituency politics explains legislators’ behavior better than electoral capture, party, or public opinion.