About this Blog
About The Author
Email Me

RealClearPolitics HorseRaceBlog

By Jay Cost

« The Pivotal Politics of Health Care Reform, Part II | HorseRaceBlog Home Page | Why Can't Obama Stop "Renegade" Democrats? »

Mike Murphy's Strange Math

Mike Murphy's new column in Time recycles many of the arguments proffered by Democrats who have asserted that their majority will be enduring. I've dealt with these at length, and rather than rehash them here, I'll point you in the direction of my essays on the subject. See here, here, and here.

Instead, I want to point out the peculiar argument that's contained in this snippet:

Despairing Republican friends have been asking me what I think we should do to rebuild the GOP and begin our certain and inevitable comeback. My answer disappoints them: "Build an ark."

I say this because I've made a career out of counting votes, and the numbers tell a clear story; the demographics of America are changing in a way that is deadly for the Republican Party as it exists today. A GOP ice age is on the way....

It was a huge shock to the GOP when Barack Obama won Republican Indiana last year. The bigger news was how he did it. Latino voters delivered the state. Exit polls showed that they provided Obama with a margin of more than 58,000 votes in a state he carried by a slim 26,000 votes. That's right, GOP, you've entered a brave new world ruled by Latino Hoosiers, and you're losing.

Of course, it was just four years ago that George W. Bush pulled in a historically large number of Latinos to the Republican Party. The exit poll had it at 44%. Some thought that was overestimated, and other estimates had it around 39%. Either way, a significant pull. This is something that seems to me to be worth mentioning when making an argument about the enduring Democratic majority. Yet it rarely is.

Anyway, Murphy's math is correct on Latinos in Indiana. The exit poll estimate has them giving Obama a plurality of something like 58,000 votes. However, his conclusion - "a brave new world ruled by Latino Hoosiers" - is completely overdrawn, which I think is characteristic of these demographics-mean-GOP-doom arguments.

The reason is...drumroll please...white voters. Shock of shocks! Who would have thought that white voters would make the biggest difference in Indiana? Yet, I can assure you that it's true! In 2004, John Kerry won 34% of the white vote. In 2008, Obama won 45%. That's an 11-point improvement, and it made a significant difference. McCain won about 218k more white Hoosiers than Obama did. Bush won 681k more whites than Kerry.

So, Hispanics moved. But so also did white voters, and their movement was much more substantial.

Like I said, I'm not terribly interested in rehashing all the various arguments for why the enduring Democratic majority argument is problematic. I've done it already. I'll only say that Murphy's argument is consistent with what I've seen many times. The proponents of this hypothesis end up putting forward numbers that somehow don't tell the full story. If it's allocating all non-white voters to the Democrats, doing an apples-to-oranges comparison of 1988 to 2008, ignoring recent elections that cut against the hypothesis, inventing ad hoc psychological concepts to explain falsifying evidence away, or whatever - there often seems to be something a little askew in the presentation of this argument. That's not to imply that anybody is cooking the books. Far from it! I have great respect for Murphy, as well as those with whom I've argued on this subject. And to a certain extent, I think they're on to something. Maybe the trick is how do you approach the data. Do you do so looking to test your hypothesis, or to find instances that support it? The latter is a dangerous endeavor, for in a data set as large as American national elections (!), you can always find something, somewhere that appears to support your theory. But that's not how data should be used to evaluate arguments.

For what it's worth, my take is that this theory relies far too heavily on the concept of realignment - something that political scientists have begun to move beyond, and for good reason. I think realignment is a highly problematic category. Almost inevitably, the data needs to be squeezed here and stretched there to fit into the proper form.

One final objection to Murphy's piece. He writes:

In 1980, Reagan beat Jimmy Carter by 10 points. If that contest were held again today, under the current demographics of the electorate per exit polls, the election would be much closer, with Reagan probably winning by about 3 points.

I object to this type of analysis - re-running old elections with contemporary demography to argue for some new alignment. I've seen this before. The problem is that each candidate's share per demographic group is locked in the past while the size of the group is updated for today. This is arbitrary! For instance, Reagan won about 60% of the two-party vote among whites that year. Carter won about 40%. A lot of Carter's white support came from the South - where he ran extremely close in Alabama, Arkansas, Kentucky, Louisiana, Mississippi, North Carolina, South Carolina, and Tennessee. Plus, he won Georgia and West Virginia. Since 1980, all of these states have shifted rightward. Other states in our fine Republic have shifted leftward, and the preferences of different types of voters have shifted as well. So, if we were to rerun the exact same election in 2008, the candidates' shares of each group would surely be different, in ways we cannot predict. So, maybe if we re-ran 1980, Reagan's overall lead would have dropped. Maybe it would have increased. Who knows? This is why I think this intellectual exercise has little analytical payoff.

-Jay Cost