Pardon my PhDweeb moment as I go theoretical and methodological. I want to discuss (briefly) whether political knowledge is a worthy concept standing alone or whether it should be folded into some overall concept.
Candidates include expertise and sophistication and awareness, all which have some history. Let's pluck out sophistication, just for the sake of argument. What makes someone sophisticated in a realm such as, say, politics? Obviously knowing stuff about the topic, thus we get knowledge, but also we need to get more theoretical. We need motivation and ability, for example.
Political interest probably deserves mention. It suggests motivation. We usually measure this with a single item that asks if a respondent cares about or is following politics or a campaign or whatever it is we're studying. High in this, people have the motivation to keep up with politics. Doesn't mean they have the ability to comprehend, so with that we might add education, which both gives one the motivation to keep up (you see the stakes and how they affect you) and the ability to do so (literacy, background info, etc.).
Okay, then there's knowledge. This suggests the ability to translate what one has been exposed to into long-term memory, and it suggests one can access that information on that domain. Knowing stuff matters.
And finally we have media use. This is the channel that delivers the information, and exposure to it says something about motivation (I keep up, therefore I must care) and potential ability (I hear/read about politics, therefore do I actually process the info?). Knowledge is the end result of media use, then.
So do we gain anything by studying sophistication rather than the individual factors that might be thought of as factors in a multidimensional variable? I dunno. I believe we lose something by lumping it together in a single variable, in part because not all media exposure is the same (CNN vs Fox vs Print, etc.), in part because not all knowledge is the same. Textbook versus civics versus current versus recall versus recognition, all these matter when studying what people know, and how they know it.
My gut feeling is that sophistication is a lot of work to fold in a lot of variables and losing some rich data in the process. This might have worked in a more homogeneous media environment, but I'm not sure it works today in our fragmented marketplace.