Election 2006 Congressional Ballot
Charles Franklin at Political Arithmetik has really become the must-read polling related blog in this election cycle.
Prof. Franklin has one part of the data that I've been longing to see for months - the trend of the generic congressional ballot difference between Democrats (D) and Republicans (R) in opinion polls taken in the 1-2 year period prior to each election, starting with the 1994 election. If you are a reader interested in the 2006 election results, you should drop everything and click through and read his post and absorb the wonderful chart that accompanies it.
But don't jump to any conclusions on the chart just yet.
After you read his post, click over to this must-read post by Prof. Matthew Soberg Shugart at Fruits and Votes, who compares the generic congressional ballot difference (D-R) in polls conducted just before the election in each year (starting in 1994) to the actual D-R vote difference in the election.
I won't spoil the fun for you by sharing all the lovely data that Prof. Franklin and Prof. Shugart have posted. But I'll make a few of my own preliminary observations here and invite those amongst you who are somewhat statistically inclined to share your thoughts in the comments. (Note: In addition to keeping in mind the error bars associated with the polls, one obviously needs to analyze the directionality of the trend lines. It is insufficient to merely look at the absolute values in the generic Congressional ballot polls.)
1. Congressional ballot opinion polls routinely and significantly overstate the D-R (Dem minus Rep) vote percentage difference expected in the elections. A quick review of Prof. Shugart's data suggests that on average the overestimate of the Democratic vote percentage in polls taken right before the election is roughly 6.4 + 4.0%. (I'm not entirely surprised by this).
2. The trend in the D-R difference in Congressional ballot polls for 2006 is somewhat unusual in comparison to polls from 1994, 1998, 2000, 2002 and 2004. In all of the latter years, as Prog. Shugart has noted, the two-party vote ended up being in favor of the Republican party.
3. The only year in which the Congressional ballot poll D-R difference trend shows some qualitative similarities right up to the election, in comparison to what we see so far in 2006, is the year 1996. Interestingly, 1996 was the only national election in the last decade in which Democrats won a greater percentage (albeit a very small percentage of ~0.3%) of votes than Republicans in the 2-party Congressional vote.
4. Is 2006 going to turn out like 1996 or like 2004? I have no idea. All we can say is that if the current trend in the polls of the Congressional ballot continues, then Democrats may actually have a good chance of doing well in the elections this year.
I hope Prof. Franklin will continue to update the Congressional ballot poll chart, and add the approximate error bars associated with the polling data and the actual electoral 2-party results to all of his charts. That would be an even greater public service than what he has been doing this year - and he has been doing a tremendous public service already!