On Moneyball and Romney’s “Momentum”

by John Sides on October 24, 2012 · 14 comments

in Campaigns and elections,Media

The lesson of this debate over Romney’s momentum—see James Fallows—is this: reporters have to ground their stories in data.  It’s fine to quote from campaign officials and gauge their mood.  But rather than use this to conjure up a narrative that implies SOMETHING NEW IS HAPPENING HERE!, reporters can be the watchdogs they typically want to be and use the data to scrutinize the spin.

If they did use the data, what they would see is this.  Romney definitely closed or at least narrowed the gap nationally after the first debate (there is some debate about how much he did so), and the gap tightened as a consequence in key swing states.  That made Obama’s lead in states like Ohio smaller and it arguably gave Romney a narrow lead in FL and a tie in VA.

On the other hand, that debate was almost 3 weeks ago.  Nothing that’s happened since then has helped him gain much, if any, additional ground.  All of the models and polling averages suggest relative stasis since then: 538, Sam Wang, Votamatic, Pollster, RCP.  So why are we talking about Romney’s momentum now?

The consequence of giving in to “narrative” and refusing to play Moneyball is evident at Politico.  Just 2 days ago, Glenn Thrush and Jonathan Martin were on the leading edge of the Romney momentum story:

With a little more than two weeks left until judgment day, Barack Obama’s campaign is embracing a fundamentally defensive strategy centered on winning Ohio at all costs — while unleashing a new barrage of blistering attacks against Mitt Romney aimed at mobilizing a less-than-fired-up Democratic base. A surging Romney is suddenly playing offense all over the map, and the upward movement since the Denver debate gives him the luxury of striking what his advisers — and more than a few Democrats — think is a more positive, presidential, “Morning in America” tone.

The rest of the story goes on to hedge a bit here and there, but the overall message is clear from the lede. And now this morning Mike Allen walks it back quite a bit:

As an antidote to the (perhaps) irrational Republican exuberance that seems to have seized D.C., we pause for the following public-service announcement. To be President, you have to win states, not debates. And Mitt Romney has a problem. Despite a great debate and what The Wall Street Journal’s Neil King Jr. on Sunday called a polling “surge,” Romney has not put away a single one of the must-have states. President Obama remains the favorite because he only needs to win a couple of the toss-ups. Mitt needs to win most of them. A cold shower for the GOP: Most polling shows Romney trailing in Ohio, Wisconsin, Nevada, New Hampshire and Iowa – by MORE than Obama trails in North Carolina. Glenn Thrush and Jonathan Martin reminded of us of the 2008 primary analogy: Whatever else Hillary Clinton had, Barack Obama had the math. And math, not momentum, gets you the big house, the bulletproof car, the cool plane.

This is a problem I’ve noted before regarding Politico in particular, but the problem isn’t limited to them.  When you can’t peg your analysis to any consistent metrics, you end up veering all over the place following the well-known incentive to write interesting stories.  On Monday, OBAMA SURGES.  On Tuesday, ROMNEY SURGES.  Other media repeat the same story because pack journalism lives.    And so on.  In reality, nothing changed and the polling fluctuations are just sampling error.

Again, the lesson is: write the story of what the campaigns are doing, but use the data to bird-dog their spin.


thegalen October 24, 2012 at 10:53 am

Great post. Related piece from Bloomberg:
“And compared with all other animals, we are endowed with remarkable capacities for causal discovery and causal reasoning, the skills that underlie the narrative habit. The divide between human and our nearest primate cousins in causal cognition capacities is as dramatic as our advantage in language use.
The trouble is that narrative thinking often supplants scientific thinking in domains of analysis and policy where we should look for more than a good story. Narrative thinking is easier for the thinker than its less natural analytic alternatives, and it is often persuasive when used to make arguments to others.
For one thing, narratives give us a false sense of understanding and control, when they are really mere redescriptions of selected subparts of the events to which they refer. Once we have a good narrative summary, we have the illusion that we could have intervened and controlled outcomes, or could have predicted what in hindsight seems to be an obvious outcome. But, unlike valid causal explanations that support informative forecasts and suggest ways to change events further down the causal stream, narratives lack these basic properties of true causal explanations.
Narratives also tend to be dominated by a few major actors, and faux explanatory power is derived from simplistic interpretations of those actors’ characters and motives. And the universal human illusion that consciously accessible thoughts are in the driver’s seat and controlling our own actions means that the salient actors in a narrative we want to understand are attributed information and incentives to a greater degree than is warranted.”

They go on to talk about Bayesian methods which, you guessed it, Silver employs in his modelling.


Gregg Murray October 24, 2012 at 11:05 am

Let’s not forget the commercial bias of the media–they are businesses that have to make money to remain viable. “Nothing changed and the polling fluctuations are just sampling error” does not draw eyeballs (and sell advertising). The “race is a nail biter” narrative is not surprising.

Adam Schaeffer October 24, 2012 at 11:41 am

John . . . no doubt the media is over-extending the surge/momentum narrative, but I think some of the data do help to explain the narrative that’s developed.

The most important is probably the pretty remarkable swing in the national poll averages, from 4-points down to a tie/one point up for Romney post-debate that’s held. The debate completely reconstructed the campaign narratives.

The other part, which has fed into the poll swing and the “momentum” narrative, is the very real boost in enthusiasm among Romney supporters. It is very high in terms of the poll numbers and in more qualitative terms like the size and energy of Romney crowds, such as in Red Rock, CO. The Republican base might be far mor optimistic than the swing-state numbers warrant, but the enthusiasm seems to be real and to have built since the first debate.

Since we’re dealing with “likely voter” numbers now, this enthusiasm and engagement translates into much bigger margins or at least a closer race for Romney in the reported numbers. Take a look at the trends in the NBC/WSJ survey, which show remarkable stability among registered voters, but a substantial shift among those counted as “likely voters.” Other polling trends show more movement, but there is always more movement among “likely voters.” In other words, the debate didn’t seem to necessarily shift vote preferences overall very much, but it did shift people into and out of the likely electorate. See here:

I think that the media is being driven in part by these real changes in the state of the race, even if they haven’t translated into increasingly better numbers for Romney in all of the crucial swing states.

John Sides October 24, 2012 at 12:19 pm

Adam: But has GOP enthusiasm increased notably in the past several days or the past week? If so, that might justify a “momentum” narrative for Romney — although no article that I’ve seen is citing data on this point, nor do I know that this is what the data would suggest. If the RV/LV gap has grown over time, then maybe something real is happening. If not, then the narrative is just that. (I’d rather not rely on qualitative impressions from rallies.)

Adam Schaeffer October 24, 2012 at 3:32 pm

Hi John . . . I don’t disagree with any of your points, and enthusiasm doesn’t appear to have grown for Romney after the first rise . . . everything seems to have settled into a “new normal.”

I just think that the abrupt shift to a closer race, and even a lead in the national polls, in conjunction with the big boost in enthusiasm and optimism among Republicans has had a lingering effect on everyone, including journalists, who also often focus on individual new polls rather than averages . . . the narrative began with surprising new data, and I think journalists kept looking for new evidence of the surge. There was momentum, so to speak, in the new perception of the race . . .

John Sides October 24, 2012 at 7:22 pm

So, put differently, a narrative had momentum even if the race didn’t. I’d say that’s precisely the problem!

Andrew Gelman October 24, 2012 at 12:14 pm


Can you clarify the “Moneyball” analogy? Is the implication that a Billy Beane-like media entrepreneur could succeed by hiring lower-cost journalists rather than paying for expensive polls? Or are you saying that the Moneyball story is happening at the level of the candidates or the parties? I ask because I, like you, get frustrated by pointless horserace coverage, and if there is a way to convince candidates or media organizations that they can save money by avoiding the horse-race trap, that would be great.

John Sides October 24, 2012 at 7:28 pm

Andy: I mean “Moneyball” only in the sense that people’s instincts or impressions of something should be checked against data when possible. Just like Beane tried to build the A’s around on-base percentages rather than whether scouts like the way the player “looked” or thought his girlfriend was pretty. I guess the analogy breaks down in terms of money, however. I’m not sure whether it’s cheaper to hire journalists or run polls, but I do know that it can take more effort — is more “expensive” in the sense of time — to look hard at the available data and build it into a story.

Maggie October 24, 2012 at 6:25 pm

Good post. The reset in the race after the first debate changed a great deal. Before the debate Romney would have been lucky to win NC. Now that’s pretty well in the bag. Florida is more likely to go his way than not. VA and CO are probably marginally in his favor, etc. The electoral advantage Obama enjoyed isn’t as formidable as it was a month ago.

The question is whether this has now stopped or if it will continue to inch forward. Reasons to see momentum on the side of Republicans (aside from the indisputable fact that there was a big swing in Romney’s direction that changed the landscape) include the fact that the internals all now have Romney as an acceptable alternative. His favorables are better than Obama’s. He’s winning independents. He’s narrowed Obama’s margin with women. Yes, that’s in some sense baked into the topline numbers on polls — but it changes the picture about how that last swing of the few undecideds will go. Silver has a long post trying to debunk the idea that late deciders break to the challenger — but on my reading, his last table undermines the claim. In elections involving a sitting president, the polls in October understate the incumbent’s vote by 1.5% and the challenger’s vote by nearly 5%. Romney is now in the position to be the acceptable alternative for late deciders who really do want a change.

Silver tries to wave that discordant fact away by talking about mean reversion, but it’s not clear to me how that measures a phenomenon different from the “myth” that voters break for challengers at the end. In any case, the better way to resist the point that it’s bad for an incumbent to be tied (and mired at 47-48%) at this point is to say that we’re in a new world where there are fewer undecideds, and the technology shifts on GOTV make the past a poor guide to the future. And I think that’s why you can have a world where both camps plausibly claim to be winning. If challengers get the break they get historically, Romney is exactly where he needs to be. Tied or nearly so, and more importantly widely viewed as a plausible alternative to pick up the votes of folks who would have signed on with Obama by now if they were willing to sit with the status quo. If by contrast we’re in a world where it’s all about turnout, the Democrats feel good about their widely touted GOTV efforts. Pretty much: is this 1980 or is it 2004?

Probably it will turn out to be 2012, which means it’s hard to call. I tend to think that the small ball posture of Obama over the last few weeks will hurt him and shift this a bit more to a 1980 sort of model. But I don’t know if that’s enough to overcome the countervailing GOTV model.

Be that as it may, it’s not narrative vs. moneyball. It’s moneyball using one model about the end game vs. moneyball using another model about the end game.

John Sides October 24, 2012 at 7:33 pm

Maggie: Lots of good points here. I certainly don’t think that Obama is in any way guaranteed to win, for the reasons you point out. “Tied or nearly so” describes the race pretty well: tied at the national level, nearly so in key battleground states.

But all of these points have mostly been in place for 3 weeks. There’s little if any systematic data that suggests that suddenly, as of Monday or so, Romney has “momentum.” That could change between now and Election Day. It just hadn’t changed when these stories started emerging in the news media.

Maggie October 24, 2012 at 11:47 pm

I agree the basic shift has been in place for a few weeks. I think what triggered the momentum narrative in the last few days largely drafted off the perception that Obama debated as though he were behind. He was the one on the attack, bringing up some rather small ball critiques at odd moments and so on. When you’re making noise about lipstick on a pig or women in binders it’s usually a mark that you are losing. I think also the fact that Romney slogged through that third debate without relinquishing the air of being a plausible alternative is in the mix.

I agree that impression got seamlessly blended with the impression from the polling data that had been set in for a little while at least, to give a sense of momentum!now! in the polls that is a false impression of what the data are actually doing.

But to be fair the the narrative tribe — I do think there’s a point to be made that if you are using the “challenger gets the late swing” model in combination with Romney successfully passing the commander in chief test in combination with Obama’s thus far desperate seeming close — it’s reasonable to get the sense that Romney’s chances are better than the static nature of the polls over the last few days would suggest. A few more pieces fell into place that shift the picture towards Romney if you are using that model.

Of course, we’ve a few days yet before we know if that third debate did anything. Who knows? Maybe Romney’s sweat put an end to the race and we don’t even know it yet!

Peter Principle October 24, 2012 at 9:17 pm

There were a few data points late last week suggesting that Romney was grinding out some additional gains — not so much in the top line numbers (although if you were willing to suspend critical thinking a bit and take the robopollers at face value, Ohio did appear to be tightening) but in the internals: favorability and who-do-you-trust-on-the economy in particular). Add a bit of attempted market manipulation in the betting shops, and voila! momentum.

But I think today’s polling haul — especially the Gallup correction and Time’s Ohio poll — kinda torpedoed that “argument” (such as it was) at least for the moment.

Crucible Guy October 24, 2012 at 9:17 pm

Great post. But the Atlantic piece is actually written by Charlie Cook, not James Fallows.

John Sides October 24, 2012 at 9:50 pm

Thanks. I’ve fixed the link.

Comments on this entry are closed.

Previous post:

Next post: