For those who read my response to Nate Silver's comments on Herman Cain, I definitely recommend reading Silver's follow-up, with a response to me and others.
Silver makes several excellent points -- as I said, read the whole thing. Perhaps the most important is just how difficult it is to resolve these things empirically because we have so few cases. As I pointed out in the previous go-round, we're talking at absolute best about twenty nomination contests, but it's really quite a bit fewer than that.
Silver argues that the difference between his results and my criticisms is, at core, about different ways of understanding the process. I think that's mostly right, and the rest of this post is about those understandings. So if you're looking for a review of how political scientists think about the nomination process, keep reading; if you want Herman Cain stuff, though, I'm afraid I don't have much today, just a bit at the bottom.
Silver argues for a bottom-up interpretation of the process. In this version, rank-and-file voters are autonomous agents. They're like those focus group participants that the networks sometimes turn to after debates. They start as undecided, spend time watching the candidates, and then make decisions informed by what they see and hear. While voter behavior isn't exactly my specialty (you want award-winning blogger Sides for that), in general I'd say that's not consistent with what we know about voting. To be sure, it's more realistic for primary elections than for general elections, and I don't rule out some role for unmediated interactions between candidates and voters. Indeed, I've argued that the role of primaries and caucuses now has become somewhat similar to what it was before reform. Back then, party leaders used primaries for information about how possible nominees would do with mass electorates (perhaps the most famous example is how JFK used the West Virginia primary in 1960 to "prove" that Southerners would vote for a Catholic). I think that's how it works now; for example, in 2004 I suspect that party leaders were willing to adapt to Howard Dean, but once he demonstrated a limited appeal to voters in Iowa they quickly turned on him and marginalized his campaign.
He calls the alternative "top-down." I have no particular problem with that, but I do need to clarify several things. First of all, not all political scientists agree with a "party decides" version of what happens in presidential nomination politics. Older research on (the reformed) nomination system emphasized candidates, the media, and voters. Voters still weren't really Silver's autonomous actors, but they were influenced mainly by candidates through their campaign, and by the press. So questions about media norms and biases were important. For example, one important story from that perspective was that "Jimmy Carter" rather than "undeclared" was the winner of the 1976 Iowa Caucuses, even though the latter had more votes, because the networks wanted to interview the winner and you can't interview an empty chair; and that decision matters because voters in New Hampshire are swayed by the positive publicity that Carter gets as a result. Or, in 2000, John McCain gets positive publicity because of some combination of his open-door policy (which appeals to the self-interest of reporters to get access) and the self-interest of the press to portray nominations as close, exciting contests, rather than just accept that a candidate with a large lead is very likely to be nominated (and, again, voters turn to McCain because of that positive publicity). Also important were questions about how campaign spending affected vote choice.
That perspective is still common, but party network researchers have introduced a new perspective, which finds that political parties play a far more central role in the process. Party network research begins by understanding parties to be not just formal party organizations (such as the Democratic National Committee or the California Republican Party) but also what I call an "expanded party" that includes campaign and governing professionals, party-aligned interest groups, candidates and their campaign organizations, and activists. Note, first of all, that this party broadly defined is not exactly the same thing as "insiders," and is certainly not limited to Washington-based partisans. Second, at least as I interpret it, this perspective is neutral with regard to which people actually do have more influence within the party -- it could be activists, it could be party-aligned interest groups, it could even be the formal party organizations. The answer will depend on lots of things, including political regulation (such as campaign finance laws), electioneering technologies, and...well, we don't really know.
At any rate, it seems clear to most of us working in that sub-sub-field that parties, properly understood, play major roles in nominations. For state and local nominations, see Seth Masket's No Middle Ground. And for presidential nominations, you should read The Party Decides by Mary Cohen, David Karol, Hans Noel, and John Zaller (you can also read my own contributions; here's -- albeit gated -- a short article I did with the great Casey Dominguez, and here's my narrative of the 2004 Democratic process).
My response to Nate Silver is that there is, in fact, considerable empirical evidence that expanded parties play a major role in nomination politics. Parties affect voters through fundraising, through how party-aligned and neutral media portray the candidates, and through direct cues to voters. On the other hand, parties haven't always controlled the process (the evidence is that they mostly didn't in the immediate post-reform cycles), and things can always change. And even within the current system, I wouldn't rule out direct voter effects. That is, if one of the things that various party members take into account is how candidates do with rank-and-file voters, then early good polling certainly could be a factor in winning party support, which translates into resources (money, prime spots on Fox News) that then produce higher vote totals in Iowa and New Hampshire.
Now, I don't think that's even remotely likely with Herman Cain, because in my view party people would be nuts to trust someone with no record in public office. But it certainly is possible, I suppose; that's what I call Cain (and the others -- Bachmann, Paul, Johnson, Huntsman, etc.) implausible, not impossible, nominees.
And remember, what I'm calling "party people" in that last paragraph includes a lot of folks, not all of whom have identical interests or points of view. It includes activists who might be "purists" and just want the best candidate on their issues, regardless of what happens in November; pollsters and media specialists who might be heavily influenced by financial incentives; GOP candidates who mostly care about winning and little else; organized groups who may care about both winning and a record of loyalty on their issues; and many more.
So my advice to those trying to figure out what to make of early polling is to think about what each of the people in those groups would make of early polling. Because the odds are that it will eventually drive voters in Iowa, New Hampshire, and the rest of it if and only if enough of the party gets on board.