Presidential elections versus mid-terms. Schaller says it is a "major no-no" to compare a presidential to an off-year election. As we acknowledge in our paper and mentioned up above, turnout varies between off-years and presidential elections. And if we had written a paper looking at historical turnout trends, we would differentiate the two.

But we are writing in the realm of the practical, where the entire House and one-third of the Senate is at stake every two years, along with a few dozen state houses and state and local offices. Politicians don't have the luxury of choosing the cycle in which they will run. And if one's aim is to take over the House, the Senate or both – you can't rely only on presidential years to create your majority.

And that is really what the heart of Looking Red, Voting Blue is about. The significance is not that the electorate was redder in an off-year, but that Democrats won in spite of it. In 2004, it was a bluer electorate that voted red by re-electing George Bush and reaffirming the Republican majority in Congress. In 2006, a redder electorate turned the tables by voting blue, and our paper tried to determine where those new Democratic votes came from.

It turns out they came from predominantly red-looking voters -- mostly men, mostly from folks who are financially better off, mostly from people who felt good about the economy, and lots of religious and rural people. These new Democratic voters had turned on Bush, soured on the war, and were sickened by Republican corruption. The Democratic majority did not come from a massive turnout of the base, but defections from GOP-type voters.

Methodology. Schaller further argues that we derived "artifactual" results by using the normalization technique. Not only does he argue that normalizing the 2004 and 2006 elections is as misleading as "comparing a slugging centerfielder with a singles-hitting second baseman," he implies that without normalization, the 2006 gains would in large measure be due to better performance among base groups. We disagree on both points.

As to validity, normalization is a standard statistical procedure employed not just in baseball but by many election analysts as well. In the same way that it allows comparisons between pitchers who face unequal numbers of batters, it allows election analysts to compare elections with unequal vote totals (including mid-term to mid-term).

But even setting that aside, our results -- non-normalized -- would remain the same. We chose to normalize because it illustrates our results in a vivid and accessible way by translating percentages of voters (pretty dry) into numbers of actual people (much more illuminating). It's also the best way to simultaneously capture the effect of changes in both electoral composition and performance.

Take, for example, the following two facts from the exit polls:

• Democratic Congressional candidates won 28 percent of voters who rated the economy "good or excellent" in 2006, versus only 16 percent of these voters in 2004.
• The percentage of voters who thought the economy was "good or excellent" increased by two percentage points.

So Democrats not only gained 12-points among people who thought the economy was in decent shape, these voters were a bigger share of the electorate. To really grasp the impact of both shifts, you want to translate this gain into actual votes. The only fair way to do this is to normalize -- i.e., by "shrinking" the 2004 election totals by a constant ratio so they are comparable to 2006. As a result, we find that the 12-point gain that Democrats made with people who think the economy is good or excellent proportionately translates into 6.1 million new votes.

But to be absolutely fair, here are some non-normalized findings from the exit polls that restate our original results:

• Democrats won a larger share of higher-income voters in 2006 than in 2004. Democrats gained 6-points among voters making between $50,000 and $100,000 and 5-points among voters making more than $100,000. (They also gained 4 points among voters making less than $15,000, but that group was less than half the size of the upper-income group in 2004. As a share of the total votes received by Democrats, the under $15,000 group dropped by nearly 2-points).
• Democrats wooed more men and married people. Democrats won 52 percent of male voters in 2006, versus 46 percent in 2004, for a gain of 6 points. The gain among women was 3 points. Democrats won 48 percent of married voters in 2006, versus 43 percent in 2004 -- a gain of 5 points. The gain among single voters was also 5 points, but married voters outnumbered single voters by more than half again. And there was a much greater turnout among the married.
• Democrats won more whites and Latinos. Democrats made a 5-point gain among whites (47 versus 42 percent) and a 14-point gain among Latinos (70 percent versus 56 percent). The share of the African-American vote won by Democrats remained unchanged. In 2006, 77 percent of the total Democratic votes received came from whites, versus 66 percent in 2004. While the Latino gains were impressive, whites outnumbered Latinos nine-to-one in both 2004 and 2006.

2002 versus 2006. Both Schaller and Schmitt argue that we should have compared 2002 to 2006. We disagree, for the reasons stated above, but for the purposes of this response, we ran the calculations anyway.

In non-normalized terms, Democrats gained 8-points among men and 7-points among whites (versus 6-points among women and a net loss among blacks). They gained 9-points among voters with household incomes above $50,000 and 12-points among voters making over $100,000 (versus 3-points among voters making less than $15,000). They made a 10-point gain with suburbanites and a 6-point gain with rural voters -- both greater than the increase among urban voters.

In normalized terms, Democrats in 2006 picked up 5.3 million new votes over 2002 -- more than the 4.7 million new votes they picked up from 2004 to 2006.

Where did these new votes come from? 4.2 million, or 4 out of 5 were white. 5.1 million, or 96 out of 100, had incomes greater than $75,000. 820,000, or 1 out of 6 were regular church attendees. In other words -- it's a similar pattern to the one we showed between 2004 and 2006. Some of the gains -- like the 2.9 million men or the 2.3 million married voters -- were very impressive but not as spectacular as the gains of 2006 over 2004.

Turnout. We address a couple points here. First, Schaller seems to argue that performance is more important than turnout. The hypothetical he uses in his critique implies that a candidate is better off winning a larger percentage of a smaller base turnout than a smaller percentage of a larger turnout.

This is actually a false choice. What matters is the actual votes -- cast or not. Let's take two elections with 1,250,000 and 1,000,000 voters where the white performance remains constant (50 percent). In the first race, 150,000 African-Americans turn out, and 80 percent vote for the Democrat, thereby giving the Democrat a victory margin of 90,000 votes. In the second race, 100,000 African-Americans turn out, 90 percent of whom vote for the Democrat, thereby providing a margin of 80,000 votes.

Which Democratic candidate would you rather be? Schaller would argue that the second candidate is better off, despite losing 10,000 votes. But that 10,000 votes amounts to more than the difference in the Montana and Virginia Senate races, which were won by Democratic candidates Jim Webb and Jon Tester. That's the difference between Sen. Harry Reid and Sen. Mitch McConnell running the show. How many House races were decided by less than a few thousand votes? Enough to hand Speaker Nancy Pelosi, and now-Committee Chairmen Henry Waxman, Charlie Rangel, and David Obey their gavels.

Second, Schmitt argues that the principal reason the 2006 electorate was whiter and better off is because all of the most contested Congressional elections were in suburban, white, affluent districts. We have two responses to this. First, the number of contested districts was actually shockingly small. Stuart Rothenberg's last pre-election analysis, for example, rated only 62 races competitive (that's 14 percent of the House). And of these non-competitive races, many more were in white districts than in minority ones -- for every Chaka Fattah, there were at least two Randy Neugebauers. Second, Congressional races weren't the only thing driving turnout. There were at least a dozen competitive Senate races, plus at least that many gubernatorial races, thousands of local elections, ballot initiatives, etc. In fact, in five of the six states with hot Senate races for which exit polls were available (Arizona, Ohio, Maryland, Missouri, Connecticut), 2006 African-American turnout was the same or higher than in 2004.

And a final point on turnout: Schaller argues that "it would not matter if Democrats returned to 2004 performance levels among privileged whites, males and so forth because they could still win the White House" if Democrats "can simply reproduce in 2008 the performance rates among their traditionally favorable, non-privileged groups." That is in fact what happened in 2004. The Democratic base turned out in droves -- and Democrats lost.

"NASCAR men" and "office park dads." Schaller says that we used statistics to create a "pre-ordained outcome" and we are "intent on proving that 'NASCAR men' or 'office park dads' are key targets." Those are his words, not ours. In our report we did not counsel Democrats about what to do to keep these voters or get more of them. We didn't urge Democrats to be centrists or appeal to moderates. We simply made the calculations and let everyone draw their own conclusions. Someone from MoveOn or VoteVets could easily conclude that the answer is not to back down one inch on Iraq. A political consultant might conclude that they should tie every Republican to George Bush. The people at Citizens for Responsible Ethics in Washington might conclude that Democrats must push for, and adhere to, ethics reforms.

In conclusion, we feel that Looking Red, Voting Blue is an honest and straightforward analysis. When we have briefed several folks who are in the business of winning elections or running themselves, and when we explain the methodology -- they quickly get it and they see the value of the analysis. We hope other readers will take the time to read it as well.

--Jim Kessler, Anne Kim, and Scott Winship