Skip to main content
ABC News
Aug. 14: The Fog of Polling — and Ryan’s ‘Bounce’ So Far

As I noted on Saturday, past vice-presidential announcements have typically been greeted with a small “bounce” in the polls. Harry J. Enten, at The Guardian, has run the numbers on this for vice-presidential announcements dating back to 1984. He finds a net gain of four percentage points in the polls — and a mean of six — for the party that just named its new running mate.

The Guardian

Mr. Enten, importantly, evaluated polls conducted in the brief window of after the vice-presidential announcement but before the party convention. (He excludes the pick of Dan Quayle in 1988, who was announced during the Republican National Convention itself.) The conventions produce their own bounces, which tend to be both larger and more predictable than the vice-presidential bounce.

An objection can nevertheless be raised to Mr. Enten’s method on the following grounds. Take Al Gore’s announcement of Joseph I. Lieberman as his running mate in 2000. Mr. Enten says that this produced a nine-point bounce for Mr. Gore, in polls conducted after the announcement but just before the Democratic convention.

However, we might have been expecting Mr. Gore to be gaining ground in the polls anyway at this time. Why? Because although the announcement of Mr. Lieberman was made on Aug. 8, just before the Democratic convention, it was also made just after the Republican convention, which concluded on Aug. 3 that year. And a party’s convention bounce typically peaks just a couple of days after its convention ends. In other words, the apparent bounce from Mr. Lieberman may have reflected the recession of the Republican Party’s convention bounce, rather than anything having to do with Mr. Gore’s vice-presidential pick.

So eliminate Mr. Lieberman from the list — along with Sarah Palin, who was named during a similar time frame, in between the party conventions in 2008. Now we’re looking only at the vice-presidential candidates who were named before either party’s convention, as Representative Paul D. Ryan just was.

We see a median bounce of three percentage points, and an average of five, with Mr. Lieberman and Mrs. Palin eliminated. So that’s a little lower than Mr. Enten’s initial estimate, although not much.

What about Mitt Romney and Mr. Ryan? We are starting to see more polls trickle in that were conducted wholly or partly after the announcement.

On another day, these polls would constitute some decent numbers for the Republicans — but so far, also, they suggest a below-average bounce.

The only two polls done entirely after the announcement of Mr. Ryan were conducted by Rasmussen Reports. Their latest survey in Ohio now shows an exactly tied race, whereas their last poll before the announcement had Mr. Obama ahead by two points instead.

There is also Rasmussen’s national tracking poll, which has now rolled over so that all three days of interviews were conducted after the announcement. It shows Mr. Romney ahead of Mr. Obama by three points, versus two points before. So this Rasmussen polling seems to be suggestive of a one- or two-point bounce for Mr. Romney and Mr. Ryan.

The other firm that was prolific on Tuesday was Public Policy Polling. It released data from Ohio, New Hampshire and its national tracking poll. All three polls conducted some of their interviews after the announcement of Mr. Ryan.

In their Ohio poll, Mr. Obama was ahead by three points, the same margin as in the firm’s poll in Ohio in June. The surveys are not strictly comparable, since Public Policy Polling has switched over to using a likely voter model since June. (Previously, it used an ambiguous category called “voters,” which we treat as being somewhere between a registered voter and likely voter poll.) However, ordinarily, a switch to a more restrictive voter universe helps Republican candidates, so that does not necessarily work in Mr. Ryan’s favor.

In New Hampshire, oddly, there were two separate Public Policy Polling surveys out on Tuesday. One poll was conducted Aug. 7 and 8, before the announcement of Mr. Ryan, on behalf of Democracy for America. The other was conducted Thursday through Sunday, partly after the announcement.

But this does give us a good head-to-head comparison, especially as both polls were conducted among likely voters. And the polls showed the same result, both putting Mr. Obama up by six points (although there were fewer undecided voters in the newer survey). So this seems to suggest no bounce at all.

(An earlier Public Policy Polling survey of New Hampshire, conducted among “voters” in May, had given Mr. Obama an 11-point lead there, although it had been a modest outlier at the time.)

Public Policy Polling’s weekly national tracking poll, conducted among registered voters, was also released on Tuesday and gave Mr. Obama a two-point lead — the same as his margin last week.

About 30 percent of the interviews in Public Policy Polling’s national poll were conducted after the announcement of Mr. Ryan. If you dig into the cross-tabs of the poll, you can find a break-out of the results they received on each day of the survey. Although the results were highly variable because of the small sample sizes — Mr. Romney and Mr. Ryan had a very poor day in their poll on Saturday, but a very good one on Sunday — in the two days combined of post-announcement interviewing, the Republicans trailed Mr. Obama by one percentage point, as compared with two points in the poll overall.

Lastly, three of the seven days of Gallup’s national tracking poll now post-date the announcement of Mr. Ryan. Gallup now shows a slight gain for Mr. Romney — he holds a lead of two percentage points in their poll, versus an exact tie before.

So there’s something of a consensus in the polls, showing a net gain for Mr. Romney of between zero and two percentage points since the announcement of Mr. Ryan.

The consensus is not quite as robust as it looks — we’re mostly relying on data from just two polling firms — and whatever signal there might be is competing against statistical noise.

I can’t rule out the possibility that, a few days from now, we’ll be talking about a signif
icant bounce for Mr. Romney and Mr. Ryan — or about how there seems to be none at all.

But this seems to be our best guess: that they’ve gained a percentage point, or perhaps two. If so, it would constitute a below-average bounce for the Republican candidates.

It would also not be that much of a surprise: there are several reasons to suspect that Mr. Ryan might get a below-par bounce. Mr. Ryan’s favorability ratings are fairly low by the standard of recent vice-presidential nominees. The announcement was made on a Saturday — a slow point in the news cycle to begin with, and especially when the announcement was competing against the final days of the Summer Olympics.

And the polls have been very stubborn this year, rarely moving in a consistent direction for any reason at all.

While I haven’t looked into the vice-presidential bounce as deeply as Mr. Enten has, I have looked at the convention bounce, and my research suggests that the volatility of polls before the party conventions is correlated with the magnitude of the convention bounces. Volatile polling years seem to predict larger convention bounces, but they are smaller when the polling has been more stable heading into the conventions.

Since the polls have been especially steady this year, we should probably expect below-average convention bounces: perhaps more like four percentage points rather than the long-term average of around seven points.

Our forecast model, which does not adjust for potential vice-presidential bounces, did have Mr. Romney and Mr. Ryan making some modest gains — largely because of the Ohio polls, which show somewhat better numbers for them than other recent polls. The model estimates their chances of winning the Electoral College to be 29.3 percent, up from 28.6 percent in Monday’s forecast.

But you should probably get accustomed to reading the polls with a more of a jaundiced eye during the next several weeks. The reason we call a “bounce” a “bounce” is because, more often than not, it fades, and polling conducted in the midst of it is not very predictive of the eventual result.

The model will make an effort to adjust for the convention bounces. For instance, if Mr. Romney and Mr. Ryan are only polling in a tie with Mr. Obama in the days between the Republican and Democratic national conventions, this will be a bearish sign for them, since this is normally a high-water mark for the opposition-party candidates. But if the polls show a tie a few days after the Democratic convention, that will be a bullish sign for the Republicans.

I think it’s better to make an effort for adjust for the convention bounces than not; in theory, our forecasts will be more stable than those at some competing polling Web sites. Nevertheless, the conventions simply introduce a lot of noise into the system, and the adjustment could easily overcompensate, or undercompensate, for their effects. This year, also — as in 2008 — the conventions will nearly overlap with each other: there may still be some after-effects from the Republican convention in Tampa, Fla., after the Democrats hold theirs in Charlotte, N.C.

A cardinal rule of polls is that they become more accurate as you get closer to the election: gradually, at first, then more vigorously over the final 45 days of the campaign.

But this is the part of the cycle when that rule is violated, because of the vice-presidential announcements and the conventions. We are now entering a foggy period in the polling.

Keep in mind that, before any of this, Mr. Obama appeared to hold a lead of two or three points in the national race overall. While I don’t take the position that there’s no new information to be gleaned from the new polls that will come in — we’ll keep updating the model and wrestling with the polls as best we can — you might want to hold that number in your head as a sanity-check.

Suddenly, in mid-September, the conventions will be far enough behind us, and the fog will clear up. And the election itself will be just a few weeks away.

Nate Silver founded and was the editor in chief of FiveThirtyEight.