Polls Don't Reflect GOP's Real Chance of Taking Senate
8:25 AM, Oct 28, 2010 • By JEFFREY H. ANDERSON
A close look at the Senate polls suggests that they are likely understating the probability of Republican victories. Most seem to be under-sampling either Republicans, independents, or both. As a result, in 10 of 12 key Senate races, the Republican candidate’s likelihood of winning appears to be greater than what the polls are registering.To be sure, such conjectures are as much an art as a science. But this much we know:
Colorado Republican Senate candidate Ken Buck
First, both major parties’ ranks have thinned somewhat from 2004, while the pool of independent voters (Tea Party or otherwise) has expanded. Rasmussen shows 32 percent of voters now identifying themselves as not being a member of either major party, up from 24 percent in the fall of 2004. That year, exit polling showed that 26 percent of voters were not affiliated with either party, and that number rose to 29 percent in 2008. It is likely to rise further.
Second, in terms of party affiliation (aside from the growth in independents), we’re right back where we were in 2004. Both Gallup and Rasmussen now show Democrats enjoying an edge in party identification of just 1 to 2 percentage points — a spread that’s almost identical to what both polls registered in 2004. In that election, according to exit polls, turnout was split evenly between the two parties: 37 percent apiece.
But by 2008, things had changed dramatically, as exit polling from 2008 showed a 7-point advantage in turnout for Democrats over Republicans.
Since the Democratic advantage in turnout was 7 points higher in 2008 than in 2004, and since party identification is now essentially identical to 2004, a reasonable guess is that (all other things being equal), this year will be about 7 points worse for Democrats, in terms of turnout margin, than 2008 was.
Thus, by adjusting the margin in each state by 7 points in the Republican direction, we can approximate the party split of 2004. And by increasing the percentage of independents by 10 percent, we can approximate their increased impact.
For example, in Wisconsin, 29 percent of voters in 2008 were independents. Increasing that number by 10 percent (29 times 1.10) gives us 32 percent (rounding to the nearest whole number). And in 2008 the Democrats enjoyed a 6-point advantage over Republicans in the Badger State. Shifting that margin by 7 points leaves us with a Republican advantage of 1 point and gives us a projected turnout of 34 percent Republicans, 33 percent Democrats, and 32 percent independents, with 1 percent remaining (which can be added to the independent tally). Aside from the growth in independents, that projection is very close to 2004, when exit polls showed that 38 percent of Wisconsin voters were Republican, 35 percent were Democrats, and 27 percent were independents.
For those who might wonder, the reason why this approach is more accurate than just using the 2004 split in party turnout for a given state and adjusting the percentage of independents upward, is that in some states there’s been a marked change in party affiliation across the past six years. In West Virginia, for example, an 18-point gap between Democratic and Republican voters in 2004 actually shrank to 16 points in 2008, at the same time that the gap in party allegiance moved 7 points in the Democrats’ direction nationally. In contrast, shifting the 2008 margin between Democrats and Republicans in each state by 7 points in the Republican direction allows the resulting estimate to reflect the 2004 turnout (nationally) while also taking account of unique developments in particular states.
In the states that are home to the most hotly contested Senate races, the projected turnout under this approach would favor the Democrats by an average of between 3 and 4 points — same as the Democrats’ average advantage in those states in 2004.
The polls, however, are generally forecasting very different turnouts than this. For example, the approach explained herein would project that the Democrats would enjoy a 5-point advantage in turnout in California, very close to the Democrats’ advantage in turnout in the Golden State in 2004 (7 points), as measured by exit polls conducted that year. And it would project that 33 percent of California’s voters this time around would be independents.
Meanwhile, among the four polls listed by Real Clear Politics from the past week that show their projected party breakdowns for California, the average projected turnout advantage for the Democrats is 11 points—almost identical to the 12-point advantage that the Democrats enjoyed in 2008, the best Democratic election in recent memory. And on average these polls project that just 22 percent of the California turnout will be made up of independents — well below the 2008 figure of 29 percent and even the 2004 figure of 27 percent.
Such examples abound. In Colorado, the most recent polls suggest that the Democrats will enjoy a 1-point advantage in turnout, just like in 2008. And these polls say that only 29 percent of Colorado voters this time around will be independents, even though 39 percent of Colorado voters were independents in 2008, and even though that percentage, by all accounts, is growing.
The expectation that Democrats will come anywhere close to matching their 2008 turnout, while independent turnout will suffer, is not rooted in reality. In truth, turnout should look very much like it did in 2004, except that independents will make up a higher percentage of the voters.
Fortunately, many of the polls show each candidate’s level of support by party. So we can see, for example, that the latest CNN/Time poll for Colorado shows Republican challenger Ken Buck with 89 percent support among Republicans, to 6 percent for Democratic incumbent Michael Bennet; 6 percent support among Democrats, to Bennet’s 90; and 49 percent support among independents, to Bennet’s 36. Using the projected turnout figures explained herein (and more or less mirroring 2004), these tallies would mean an 11-point lead for Buck. Because of its radically different turnout projections, CNN/Time calls it a 1-point lead for Buck.
On the whole — using every poll from the past ten days that breaks down support by party (and in Washington, Florida, Delaware, and Wisconsin, where such polls are more scarce, going back a bit further) — here are the tallies that these polls would yield if they were to use the turnout projections outlined herein (with the actual RCP averages for each state listed alongside):
—California: Boxer (D) by 1 (RCP average: Boxer by 6)
—Colorado: Buck (R) by 8 (RCP: Buck by 2)
—Connecticut: Blumenthal (D) by 12.5 (RCP: Blumenthal by 12.5)
—Delaware: Coons (D) by 16 (RCP: Coons by 17)
—Florida: Rubio (R) by 10, over Crist) (RCP: Rubio by 12, over Crist)
—Kentucky: Paul (R) by 12 (RCP: Paul by 8)
—Illinois: Kirk (R) by 4 (RCP: Kirk by 3)
—Nevada: Angle (R) by 3.5 (RCP: Angle by 2)
—Pennsylvania: Toomey (R) by 5 (RCP: Toomey by 3)
—Washington: Rossi (R) by 1 (RCP: Murray (D) by 2)
—West Virginia: Manchin (D) by 1.5 (RCP: Manchin by 5)
—Wisconsin: Johnson (R) by 10 (RCP: Johnson by 6)
In all but Florida and Connecticut, the polls appear to be inflating the Democratic candidates’ prospects by inflating Democratic turnout.
In all likelihood, however, Republican candidates have a shot of doing slightly better on the whole than even the above tallies suggest, for the turnout projections that inform these tallies really only take into account party identification, not party enthusiasm. And party enthusiasm certainly seems to favor the GOP. In this year’s primaries, according to American University researcher Curtis Gans, Republican turnout outnumbered Democratic turnout for the first time in 80 years.
On the whole, Republicans look like they have a reasonably good chance to maintain their advantage in the races where they’re leading by at least three or four points in the tallies above, which means they would need to take two out of three among West Virginia, Washington, and California — all nearly toss-ups — to become the majority party in the Senate.