I read a lot and take in multiple data points. Most of the time, the information is predominantly about investments. These days, quite a bit it concerns the upcoming Presidential election.
I’m always interested in how diverse data points integrate. Sometimes the data points appear consistent. Here’s an example, pace Monty Python: If I know that all fish live in the sea, and that all herring are fish, and if I then see a herring swimming beneath my glass-bottom boat, the data points line up. But if I note the first two data points (fish/sea, herring/fish) and then observe a herring riding a bicycle outside my window, I have a “What’s wrong with this picture?” moment.
Right now here I see six election data points I find tough to reconcile. I list them below in what seems to me to be order of declining credibility:
1) InTrade odds show Obama with a 67 to 33 advantage. In other words, if you buy $100 worth of Romney shares, and he wins, you will be paid $303. If you buy $100 of Obama shares and he wins, you get only $149. For any believer in efficient markets, this is pretty much dispositive. Not impossible odds, but Romney getting elected would be an upset equivalent to football’s New York Giants beating the Patriots in the 2008 Super Bowl.
2) Nate Silver of the New York Times, who correctly forecast the winner in 49 out of 50 states in 2008, gives Obama over an 80% chance of being re-elected, which is darn close to a lock. His complex model uses state-by-state polling to project confidence levels in each state, then combines individual state predictions into projections about likely Electoral College totals.
3) Recent polling shows Romney with an advantage among independents of between 5 and 12 points. (Obama won independents in 2008 by a margin of 8 points.) It is a political truism that independents decide Ohio, and that Ohio decides the election – at least, if the Republican wins.
4) Obama has rarely been above 50% in any national poll over the last few weeks. An incumbent President who polls below 50% with less than a week to go in the race has never been re-elected. In large part this is because, historically, late-deciders break against the incumbent in all but the most favorable economic conditions.
5) Per Gallup, early voting shows Romney leading by 52% to 45%, a seven-point advantage, compared to a 15% Obama advantage at the same point in 2008, a 22% net swing toward the Republican.
6) In the television era, no President has been re-elected with unemployment above 7.3%. (It is now 7.8%.)
What’s wrong with this picture? How can you reconcile data points 1 & 2 (Intrade and Nate Silver) with data points 3, 4, 5 & 6 (advantage independents, below 50% polling, advantage early voting, lousy economy)?
Any possible answer must take account of the mechanics of modern polling. Polling is not like the weather report. You don’t simply stick a thermometer out the window and read the temperature. The reason for this is that many polling calls go unanswered. Was your final sample representative? Did you under-poll young Democrats with cell phones, or did most Republicans refuse to pick up the phone? You are left with a dataset that may be different from those who actually cast ballots.
It’s kind of like me and my wife. Unless pregnant, Amy is usually cold where I am warm. I’m happy in a 60° house in the winter, miserable in a 75° house in the summer. She is unhappy at 65° in the winter, perfectly comfortable at 80° in summer. If you call our house and ask, “Is it warm or cold?” in order to estimate the temperature, you had better know if you are talking to me or Amy. If you make 1000 phone calls, to a neighborhood with 500 Amys and 500 Jims, but you end up talking to 50 Amys and only 30 Jims, you will think it is colder than it really is.
Same way with polling. If you talk to 1000 voters, some self-identified Democrats, some Independents and some Republicans, you are going to plug them into your model according to what you believe the actual electorate will look like on November 6. You will assume a certain number of Democrats, a certain number of Independents and a certain number of Republicans, and work to make sure your sample matches that likely electorate. So here are two possible explanations of how you can reconcile the conflicting clusters of data points:
1) Markets are efficient. Intrade and Nate Silver are both right. Polls that show Romney with an advantage among Independents are really counting a bunch of closet Republicans (different from Log Cabin Republicans) as Independents. Romney may be tied with Obama in the overall polls, but that is because he is way ahead in red states (which he was always going to win), closer in blue states (which he will still lose), but behind in so many battleground states that he can’t plausibly get to 270 electoral votes. Republicans may have taken a page from the Democrats early-voting playbook, but Obama’s ground game will still swamp Romney on Election Day.
2) Markets are inefficient. This election is much more like 2004 or even 2010, with Republicans and Democrats even among likely voters, or perhaps with Republicans slightly ahead. Intrade is responding to Nate Silver, and both are relying on polls that systematically over-sample Democrat voters, because pollsters assume in error that the 2012 electorate will be even more liberal and non-white than the 2008 electorate. It becomes a “garbage in-garbage out” problem, with under-counted Republicans in multiple polls creating an apparently insurmountable Obama advantage in the Electoral College. Romney wins, and possibly by larger margins than most people expect.
Reading comments from both the Left and Right, I am struck more than ever about the degree to which we are not dealing simply with two different sets of principles, but two different realities.
There are potential lessons here on market efficiency, the power or limits of systems, and whether the behavior of large numbers of human beings can be predicted. I’ll touch on this again next week, once we know the outcome, and we’ll see if there are any lessons that we can persuasively apply to our investments.