Democrats rode a wave of wins in 2006, after coming up short in 2004. What did we do right?
In a recent exit poll analysis, Chris Bowers tries to show that we energized the base and boosted Democratic turnout. Markos highlights this result in a DLC-bashing front page post. Unfortunately, Chris didn't finish the maths, or select the most comparable data.
Perfecting his model, we learn that 78% of the swing came from Independents and Republicans. ("Base" gains finished third.) This tends to weigh against "base election" schemes, in favor of persuasion, conversion and oppo attrition strategies, with a primary focus on "center court".
Details [and bonus updates by Ideology and Region!] below the fold.
First, the maths. Elections are decided by vote margin, not vote total. Every political act wins votes and loses votes, for your side and the opposition. Capable strategists instinctively size up every bright idea that's thrown on the table by this four-part rule. Unless our political "income statement" covers both sides of the equation, we're "budgeting votes" with play money.
Bowers uses exit poll data to estimate Democratic gains between 2004 and 2006 -- but not Republican losses.
For instance, 26% of respondents exit polled in 2006 House races identified themselves as Independents (same as the 2004 Presidential sample). In 2004, 49% of these reported voting for Kerry. 57% said they voted for House Dem's in 2006. 0.26 x (0.57 - 0.49) = 0.0208, or 2.08% So far, we confirm Bowers figure.
But Republicans also lost 7% of these Indies (from 46% for Bush in 2004, to 39% I's for Republican House candidates in 2006). This adds 1.82% to our net margin attributable to I's. Total margin gain from Independents? 3.90%, not 2.08%.
In similar fashion, using the same data, net margin gain from self-identified Democrats pencils out to 3.82%, and from Republicans, 2.31%.
Thus I's gave us more swing than D's, which in turn gave us more than R's. 3.82% net margin gain came from turning out our "base", versus 6.21% from non-party and opposing party segments of the electorate. (We could also do this with self-identified liberals, moderates, conservatives.)
Next, the data. The National Election Pool collects separate samples for Presidential and House races, and publishes separate results. The original analysis calculated swing between 2004 Presidential exit poll results and 2006 "National House" results. That's apples-to-lugnuts.
Using the more comparable NEP 2004 National House sample, we derive strikingly different results.
On a House-to-House, margin-to-margin basis, we derive:
3.93% Margin growth from Independents
2.80% Margin growth from Republicans
1.90% Margin growth from Democrats
Thus, of our 8.63% net margin swing between 2004 and 2006, our greatest yield came from self-described I's. The next most came from R's ... with D's a distant third (22% of net margin gain).
Do these more comparable data and more complete calculations lead you to appreciate the exit poll analysis less? We can offer the following grains of salt:
- Party ID is short term mutable. Dem-leaning centrists may identify as Democrats today and Independents tomorrow. Similar for "out and proud" Republicans and Republican-leaning Independents.
- Any comparison of a presidential election cycle (2004) and a midterm is automatically suspect. They are not the same electorates, either nationally or locally. (Neither are there competitive race drawing cards in all the same districts in both years.)
- NEP protocols were substantially reformed after the 2004 exit polling debacle. They aren't same polls.
- Exit poll results are published in whole percents. A reported "37%" could be anything from 36.5% to 37.5%. The actual number behind a reported "49% of 26%" could be 12.74%, or 12.3675%, or 13.1175% ... anywhere within a 0.75% range. A comparison of two such products (for 2004 and 2006, for instance) could be 2.08% plus or minus 1.5%!
- Exit polls are not high-acuity instruments in any case, especially on deltas between subsamples.
With all those caveats, how seriously should we take the result? Not entirely seriously ... but it's a nice topic for further further study and analysis.
Update: We can do likewise with self-described "Liberal", "Moderate" and "Conservative" exit poll respondents. In brief:
0.71% Lib's
4.49% Mod's
3.20% Con's
... for 8.40% net margin swing, and 92% of that achieved outside our liberal base.
By Region?
- Democrats accounted for a bare majority (50.39%) of our 10.36% swing in the East.
- Southern D's swung the wrong direction (-1.49%), accounting for a whopping minus 27% of their region's favorable 5.61% swing.
- Midwest D's accounted for 27% of the region's 5.24% swing.
- In the West, D's swung 33% of a net 11.79% swing.
