Obama's toast.

Or, he's not.

History proves it.

I don't watch preseason football or pay much attention to this kind of stuff, but I'm not judging you if you do.

Comments

[Andrew Brod]

Every four years I write a column on what the economic models are predicting. I haven't done my research on that for what'll probably be my end-of-October column in the BizJ, but I expect to find that they're all pointing toward a Romney win.

Up to now, I've focused on the political-economic model run by what's now Moody's Analytics. For what it's worth, in 2000 that model pointed strongly to a Gore win. In 2004, it got things mostly right by predicting a Bush win but overestimated the margin of victory. In 2008, the model correctly predicted Obama's win, including a narrow victory in North Carolina.

Wait a minute. This year the Moody's model is pointing toward Obama. I guess I see why, on one level. Most of these models are based on changes rather than levels, and there has been improvement. Even so, I'm not sure I buy it. If Obama wins, it'll be in spite of current economic conditions, not because of them.

[justcorbly]

Who knows? Americans are woefully stupid about economics and their personal finances. (See this.)

Voting for some rich guy because you think he can get you a job makes about as much sense as voting for a retired general because he won some battles. But, we've done that often enough in our history. (All praise Zachary Taylor!)

GOP social policy seems to resonant with about 30 percent of the populace. That policy convinces some unknown number of people to vote for the Dems, no matter what. We will see what happens.

[David Wharton]

Just to be sure I understand you, Andrew. In 2000, your model was quite wrong ("And according to the Dismal Scientist, it’ll be Al Gore in an electoral-college landslide"). In 2004, your model was (somewhat) right.

How is your model superior to mine, which I'm running right now and (*flips coin*) is predicting a narrow Romney victory?

[polifrog]

Voter Identification (Rasmussen)

2008 - 41% D to 33% R
2010 - 35% D to 35% R
2012 - 34% D to 36% R

[Andrew Brod]

DW, I'm not claiming anything about the economic model's superiority over other models.

Having said that, the Moody's model correctly predicted 4 of 5 elections since 1992. The probability that a coin flip would be right 4 or 5 times in 5 tries is under 1/5. So there's less than a 20% chance that your coin would have done as well as or better than the Moody's model.

Moreover, as I noted in each of those columns, the economic model tries to capture the way economic factors affect an election. It doesn't capture, and isn't intended to capture, the way other factors come into play. Sometimes the other factors dominate, as in 2000, but usually they don't.

We'll see how it plays out this year.

[Account Deleted]

Interesting to note that Ford, Carter and GHWB lost with unemployment above 7 percent, but Reagan pulled out a win in 1984 with similar numbers. His GDP numbers on the linked chart pointed to coming growth and the Dems ran an anemic candidate in that election.

We've become incredibly polarized since then and a ham sandwich running as a Republican would still get 34 percent of the vote or so. Romney, however, may have trouble attracting independent votes given the hard right drift of the GOP and his lack of specific economic proposals.

That being said, does anyone have data showing how the independents are breaking thus far?

I watched Obama's speech last night with a critical mind and I have to say I could see him winning the likability contest against Romney hands down.

[Andrew Brod]

Working in Reagan's favor was the fact that the economy was improving significantly. That's what's great about a V-shaped recession. Unemployment had fallen nearly 3 percentage points since the beginning of 1983. In contrast, the unemployment rate in August of this year was only a single percentage point lower than it was in January 2011.

I don't know what the polling data are saying about independents this year, Jeff, but my understanding is that we got down to a low rate of undecideds much earlier than usual.

[Spag]

"GOP social policy seems to resonant with about 30 percent of the populace."

You base that on what?

[Spag]

"The unemployment rate has been above 8 percent since February 2009, a month after Obama’s inauguration—the longest period of such elevated joblessness since the Great Depression of the 1930s." - Business Insider.

Didn't Obama tell us back in 2008 that we were facing the "worst economic crisis since the Great Depression"? He blamed that on his predecessor. Perhaps Romney is justified in doing the same thing.

Now we could pull ourselves out of this if we would only raise taxes on the rich another 4.9%...But those Republicans, they just don't get it. The reason we keep losing jobs is because the "rich" don't pay that extra 4.9%...

[Andrew Brod]

"The reason we keep losing jobs is because the "rich" don't pay that extra 4.9%."

In case it matters, that's not what the Democrats are saying.

[Spag]

They aren't?

"I want to reform the tax code so that it's simple, fair, and asks the wealthiest households to pay higher taxes on incomes over $250,000 - the same rate we had when Bill Clinton was president; the same rate we had when our economy created nearly 23 million new jobs, the biggest surplus in history, and a lot of millionaires to boot." - Barack Obama, 9/7/2012

[David Wharton]

OK, Andrew, so I just did 5 coin tosses. This is the only series of coin tosses I did -- no cherry-picking from a larger series. Here are the results:

tails, tails, tails, heads, tails.

Weird, huh? The probability of that happening is the same as your Moody model getting 4 out of 5 (15.6%). And yet it happened, because a 15.6% probability isn't really all that low.

[Dale]

There is only 1 chance in 32 of getting that pattern.
Times 5 ways to get 4 tails and 1 head.

[Roch]

DW is right that the chance of five coin tosses resulting in 4 out of 5 tails (or heads) is indeed 15.6% (5 possible combinations out of 32 permutations).

And Dale is right, the chance of getting them in a particular order is 1 out of 32.

For the sake of comparing the accuracy of a coin flip to the Moody's model, we are only interested in the probability of getting any four out of five, not the four out of five that exactly matches Moody's sequence. So DW's 15.6% is the proper coin-flip probability to use.

Therefore, a coin flip would have a worse record than the Moody's model 84.4% of the time.*

* Over a series of five picks. The odds of a coin flip being accurate for any given election is much better, 1 out of 2.

[Roch]

I'm reminded that the results of the Gore/Bush election was less like a heads or tails and more like the coin landing on its edge.

[Cunningham]

That's why they play the game.

[David Wharton]

Actually, even taking into account the ordering, your chance of getting just one wrong is still 16.5%. Say the actual sequence of wins was

D R D D D

Here are the combinations that would miss by just one in the ordering:

R R D D D

D D D D D

D R R D D

D R D R D

D R D D R

There are five ways to miss just one in the correct order: 5 / 32 = 15.6%

[Ed Cone]

Not strictly relevant to the statistical analysis, but I believe in politics the formulation is heads I win tails you lose.

[justcorbly]

Spag, see Frog's numbers, then subtract Republicans who are not evangelicals or social conservatives.

I think "about 30 percent" jives with estimates I've across.

My larger point, though, is that fear of conservative social policies becoming enshrined in law does drive some unknown number of people to vote for Democrats when, in the absence of those policies, they might not. That holds if my number is wrong or right.

Obviously, I'm an example of that. Dems and Obama disappoint me on a number of fronts. But, until conservatives, for starters, stop trying to make crimes out of the evangelical roster of sins, I won't even consider voting for any Republican. Most of the people I associate with in my little self-selecting circle agree.

We are so divided in this country, so polarized, and so able and so likely to go about our daily lives with little or no out-of-the-workplace contact with people who hold conflicting social views that it isn't surprising that both camps misunderstand the demographic position of each other.

[Roch]

Yes, DW, I was agreeing with you, and the odds of a sequence of flips that exactly matches the Moody's sequence is 1 out of 32.

[Spag]

Corbs, that doesn't answer the question. What policies are you referring to, and what data do you have to illustrate the popular support of them across the country?

It sounds to me like you need to broaden your social circle to include more diversity of opinion before sounding off on what others actually believe.

[Andrew Brod]

" I just did 5 coin tosses. This is the only series of coin tosses I did -- no cherry-picking from a larger series."

I'm a fan of experiments, but just to be clear: When it comes to probability theory, the only way not to cherry-pick is to do a longer series of coin flips. Flip your coin 5 times and record your results, and repeat the 5-flip sequence 100 times or so. Then you'll have a feel for the probabilities. Doing it once proves nothing, I'm afraid.

[David Wharton]

No, doing just one series is not cherry-picking: there's no other data to pick from. Of course neither is it dispositive (and I didn't intend it as such), no argument using probability statistics is capable of "proving" anything.

But you're right that a single somewhat correct series doesn't demonstrate much in the way of predictive accuracy. Apply that reasoning to your Moody's series, and you'll grasp my point.

[polifrog]

Interesting. Which one of those 100 or so groups of 5 flips would you choose?

At least in statistics a sample size of 30-some is required before some minimum degree of accuracy is met.

[Andrew Brod]

I take your point, DW. However, I just flipped 5 coins and got 2 tails. Anyone can play this game, and if enough people did, we'd see slightly under 1 in 6 of them getting 4 tails and 1 head.

In contrast, Moody's model can only be done one way, and it was right 4 of 5 times. The model's prediction is a function of other data (e.g. the unemployment rate), whereas flipping a coin is conditional on nothing else.

Frog, one would choose none of the 5-flip sequences. That exercise isn't to come up with a prediction; it's to confirm the statistical properties of the predictor. Flipping a coin 30 times would be enough to confirm that a fair coin has a 50% chance of showing tails, but I was thinking that a few more 5-flip sequences might be necessary given the asymmetry of the distribution we're actually talking about. N=30 isn't an ironclad rule in statistics; it depends on the nature of the distribution.

[Andrew Brod]

But again, whether the Moody's model is right or wrong is only part of what's interesting about it. What's most interesting to me is what it tells us about the economic dimension of each presidential election. I don't think it was completely wrong in 2000, because it told us that if people were going to vote their pocketbooks, Gore would win easily. Moody's knew it would be a close election when they released the model's results, but they also knew that a lot more than economics matters in presidential elections. It was obvious to everyone that pocketbook issues weren't front and center in that election.