Adjusting polls for party identification

There has been some discussion about adjusting public opinion polls for party identification (for example, see this page by Alan Reifman, which I found in a Google search). Apparently there has been some controversy over the idea as it was applied in the 2004 Presidential election campaign. Setting aside details of recent implementations, adjusting for party ID is in general a good idea, although it’s not as easy as adjusting for characteristics such as sex, age, and ethnicity, whose population proportions are well-estimated from the Census (and which change very little, if at all, during an election campaign).

The basic idea

Here’s a simple example to illustrate party-id adjustment. Suppose that from one week to the next, there is a 5% shift in the poll response for some opinion question (for example, support for the Republican candidate in an upcoming election). Then it would be natural to estimate that there had been an underlying 5% shift in national opinion. But now suppose that you learned that the second poll had 5% more Republicans compared to the first poll. Then it would be more reasonable to assume that there had been no national opinion change, but rather that the first poll just had too many Democrats (or the second poll had too many Republicans, or some combination of these).

Statistical modeling

As the above example illustrates, in some cases an adjustment for party ID can be simple. More generally, one can do the adjustment using a series of polls, fitting a time-series model for party ID and then using the estimated proportion of Democrats, Republicans, and Independents to weight (formally, to post-stratify) the respondents in each individual poll.

Details appear in our paper from 2001 in the Journal of the American Statistical Association, by Cavan Reilly, Jonathan Katz, and myself, where we illustrate with the example of using party ID to improve estimates of Presidential approval.

See also here, here, here, and here, for more on this topic.