There has been some good recent scholarship (here and here) in political science challenging the use of the hypothetico-deductive model to explain how race impacts the political process. Traditionally, political scientists have taken race or ethnic identification to logically precede group-based interest-formation and mobilization.

The reality of race and ethnicity is that they are multifaceted, inter-sectional and contextual constructs that cannot be captured through survey research that asks respondents to check a box next to the ethnicity with which they identify.

Attempts by statistical researchers to ‘control for third variables’… ignore the ontological embeddedness of locatedness of entities within actual situation contexts.” (Emirbayer 1997, 289).

This is true, but then the question remains, how do you validly and reliably study identity in the political process. One interesting approach might be to employ folksonomies to race questions in political science. Rather than asking people to classify themselves according to the controlled vocabulary of the survey researcher, a folksonomy would allow the respondent to use as many self-identifiers they want to describe themselves. You can use social network analysis to group respondents based on the similarity of their self-tagging structures into clusters and then test whether cluster membership is related to a desired political outcome.