Charlie Cook from the the National Journal discusses the practice of combining polls. Essentially the idea is to combine the data of two or more surveys with the exact same question wording in order to increase the sample size. By doing this you increase the accuracy and decrease the margin of error for crosstab categories.
Maybe we ought to try it. For example, it could help explain the racial dynamics in the Democratic primary races.
I think that the idea could run into a lot of problems. The presidential race has seen a lot of fluctuation, with people dropping out, the Edwards announcement, coverage of money etc. To combine polls done 6 weeks apart in that case seems a little odd.
ReplyDeleteHowever, for races like the Lt Gov here in NC (for example) I think the idea makes a lot of sense.
You're right about the presidential race. For example, the African-American crosstabs have fluctuated wildly. Partly because everyone is just getting introduced to Obama.
ReplyDeleteBut the fluctuations may be precisely because we haven't got a large enough sample.