or, "Who has won 'the Democratic wing of the Democratic Party'?"
This post begins with a hypothesis: that Obama's wins have been bigger in red states than in blue states. It was based on the recollection that Clinton had fared best in the northeast and California, including the Big Blue threesome (CA, MA, NY) while Obama had been sweeping the south and west. If true, it seemed to challenge the conventional wisdom that one must run left in a Democratic primary to win, and the perception that Clinton is the more centrist candidate. If true, it might also bring interest to how the coin will fall in the remaining Big Blue states, Vermont and Rhode Island, whose March 4 primaries are otherwise overshadowed by the more delegate-rich contests in Texas and Ohio.
But, as you'll see, the hypothesis does not hold up well to the data. (Q: "What's the difference between a blogger and a pundit?" A: "One tests their ideas against the data, the other tests the data against their ideas.") That is, it is not entirely true -- Obama's performance in a state is probably explained better by geography, by timing, and by whether it is a caucus or primary state, than it is by how red or blue the state is -- but, as you'll see, it's not entirely false either.
Let's start with the data.
What you're looking at is a scatter plot in which each state (plus DC) that has already held its primary or caucus is represented by a dot. The dot's position on the X axis ("Bush-Kerry") shows what Kerry's percent margin of victory was in that state, and its position on the Y axis ("Obama-Clinton") shows what Clinton's percent margin of victory was in that state. (Undoubtedly, someone will complain that Kerry should be on the left or Obama should be on the top. Whatever.) So, the reddest of the red states are on the left, and the bluest of the blues are on the right.
I expected Clinton and Obama's "home states" -- AR and NY for Clinton, HI and IL for Obama -- to vote strongly for their daughter and son, regardless of national trends, so I put each in a special color and I recommend ignoring them. I've labeled some of the other dots to highlight the states that sometimes have a strong opinion -- the ones that went by large margins for any one of Bush, Kerry, Clinton, or Obama.
(The chart data I used came from electoral-vote.com (2004 data, 2008 data from the main page's map). The data I used can be downloaded as a comma-separated list you can open in your favorite spreadsheet program by clicking here.)
There are a few ways to read the chart. The first, and easiest, is to imagine a diagonal line from the lower left to the upper right, and ask yourself if you think the dots fall along that line. If they do, it would suggest that Obama's success in a given state can be explained, at least partly, by how red the state is. Next, imagine a line from the top left to the bottom right and ask yourself the same question. If the dots fall along that line, it means that you think Obama's success can be explained, at least partly, by how blue a state is. To my eye, there is a weak trend for the black dots to fall along the lower-left to upper-right line, but it is thrown off quite a bit by OK, TN, UT, and DC.
(As an aside, yes, one could use statistics to make this argument more rigorously. Personally, my approach to data is first to ask how it looks and then to try to quantify those perceptions using statistics -- because no matter what arguments I find I can make using statistics, if they don't look qualitatively right to my eye, I don't really believe in them. That said, the black dots in the chart above fit a terrible linear regression with an r-squared of less than .01, but if you let me remove OK, TN, UT and DC -- for no good reason other than that they're screwing up my fit -- then you get a reasonable linear regression that can be expressed as "Clinton = .8 x Kerry - 13", with an r-squared of .4, which is to say that if you figure that Clinton gets about 80 percent of the Kerry voters and then loses 13 points overall to Obamania, you'll be able to explain about 40 percent of the variation in the state-to-state vote totals so far. The important points here are that the slope of the regression line, .8, is positive and close to 1, consistent with my eyeball take that Clinton tends to do better in bluer states (although often that just means "loses by less"), and that the r-squared value, .4, is midrange between 0 (no correlation) and 1 (perfect correlation), consistent with my take that the fit is ok but not spectacular. And, for what it's worth, my impressions were not written post hoc -- I ran the regressions after writing the above paragraph about how the chart looked to me qualitatively.)
Another way to look at the data is to take only the extreme outliers. For example, where Kerry had his biggest wins (the bluest blue states), how did Clinton do? Kerry's biggest wins were, in order, DC, MA, RI, VT, NY, MD, CA, CT, and IL. He beat Bush by 10 points or more in each of those states. Of those states, we're ignoring the candidates' home states of NY and IL, and RI and VT have yet to vote. That leaves five contests: DC, MA, MD, CA, and CT. Of those, Obama had a blow-out in DC which may be explained partly by racial demographics, Clinton had some of her most solid wins in MA and CA, and they nearly split CT (it was Obama's second-smallest win, after MO). But what about MD? It was solidly pro-Kerry, and now it's solidly pro-Obama. Perhaps racial demographics again come into play. Perhaps it's the timing of the primary (it went Feb 12, a week after Obama's Super Tuesday victories). The point is, the perception that Clinton did better in blue states is often true -- but demographics, geography, and timing can trump it.
How about Bush's biggest winners? Reading the chart from left to right, you can see that Bush's second-biggest win, ID, was Obama's biggest blowout. Bush's third biggest win, NE, was also among Obama's best performances, as were Bush's #5 and #6, AK and KS. But Bush's #4 state, OK, was Clinton's biggest win after her home state of AR, and his #1 state, UT, was much closer than a simple "red staters like Obama" model would have predicted.
Who has won "the Democratic wing of the Democratic Party"?
The phrase was popularized by Howard Dean during his 2004 campaign, and it's a reference to his opponents at the time who, he argued, were running as "Bush Lite." Making a strong stand to represent traditional Democratic principles of a balanced budget, tolerance and open-mindedness, and a government dedicated to the public good, Dean sought to contrast himself with centrist candidates in the field. As Democrats often do in a primary, he ran left -- but, startlingly for many of us, he lost.
Of the two Democratic candidates left in the race today, Obama may draw the more obvious comparison to Dean. He is running as an outsider, who wants to change the tone of politics in Washington (although, for that matter, Obama is equally echoing George W. Bush's "reformer with results" outsider campaign from 2000). Obama, like Dean, is turning young people into activists and provoking their passions on a scale rarely encountered in politics. Yet, as I've shown above, he is certainly not performing best in the states that are the strongholds for the Democratic wing of the Democratic Party. Although he is winning many of them, and geography and timing may be better explanatory factors than the state's blueness, Obama is doing better in red states while Clinton is outperforming (at least, losing by less) in blue states.
How can this be? One explanation is that Clinton is the establishment candidate, and the solidly blue states are the states where the Democratic establishment is strongest -- that she's not the favorite of progressives, she's the favorite of party faithful. But perhaps another explanation is that Clinton really is the more progressive candidate.
Don't look at the win, look at the margin of victory
Another lesson here is that the widely-reported numbers about who has "won" more primaries are, in fact, fairly meaningless. They may be hold-overs from the organized media's comfortable method of reporting general election results, but in primaries where delegates are apportioned proportionally -- and, per DNC rules, all Democratic primaries are run that way -- it's the size of the victory that counts. Whether Obama won MO 49-48 or lost it 48-49 doesn't matter: it will hardly affect the delegate count. And, as I think the chart above shows, the most useful information is often in the states where a candidate won by 10 points or more. To lump together Obama's 4-point win in CT and his 63-point win in ID is to disregard common sense, and to throw away useful information.
Taking into account the size of the victories also allows one to test ideas about why a candidate has won, as I've done here, and to try to use those ideas to make quantitative predictions. If you take at face value the model that "Clinton = .8 x Kerry - 13" then you can predict the following for March 4:
TX: Obama 65 Clinton 34
OH: Obama 57 Clinton 43
VT: Clinton 52 Obama 48
RI: Clinton 52 Obama 48
and, for the week after,
WY: Obama 73 Clinton 27
MS: Obama 65 Clinton 35
While all eyes remain on TX and OH, it will be worth keeping tuned in to VT and RI as indicators of whether the idea that Clinton does better with the true-blue Democratic left holds water.
There really are states that "don't matter"
The Clinton campaign has been castigated for referring to states she has lost as states that didn't really matter anyway. And they've somewhat deserved it, because it stinks of sour grapes and a loss of message discipline. But, as the chart above subtly shows, they have something of a point.
One of the more obvious features of the chart is that there are a LOT more points on the bottom than the top -- Obama has simply racked up many more wins. But just as obvious is that there are
even MORE also more points on the left than the right -- Bush won a far greater fraction of states than Kerry did. Yet that election was close enough that a few diehards still insist Kerry won it, and most of us spent a great deal of election day morning expecting that to be true.
The cold electoral fact is that many of those dots just don't matter much. You can sweep the western bloc of ID, MT, WY, ND, and SD and still get only 16 electoral votes, where if you had just taken NJ you'd have gotten 15. In fact, to become president, all you need to do is win 11 states (NJ, NC, MI, OH, IL, PA, FL, NY, TX, GA and CA). The other 39 states wouldn't matter. In fact, Gore won only 20 states, Kerry only 19. Saying that some states don't matter is terrible politics -- it's rude, it's undemocratic, it's poor sportsmanship. (It also stupidly ignores the importance of bringing one's coattails to Senate, House, and local races.) But, as is implicit in the chart above, it's not just how many states you win, but if you're able to win the ones that "matter."
Obama has won the ones that matter. He has also won blue states. But his margin of victory has been bigger in states that are smaller and redder -- he has performed best where Kerry performed worst. That should give us great hope for the prospect of having Obama as the nominee heading into November. It may also give us pause about whether he represents the Democratic wing of the Democratic Party -- and whether, if he did, he would perform as well as he has.
I'd like to acknowledge the tremendous help that it was to have the electoral-vote.com databases available during the writing of this post. If you like quantitative analysis, I strongly recommend adding the site to your daily reads.