One of the maxims of past nomination campaigns is that a candidate should expect to receive a bounce in the polls after winning a major primary or caucus — particularly if he or she beat the expectations established by the polls and the news media while doing so.
Often, this bounce would persist until the next state voted, allowing the candidate to perform strongly there as well. Potentially, this could create a chain reaction in which the candidate received a fresh dose of momentum every week or two.
So far in 2012, however, the bounces that candidates have received from primary and caucus wins have been minimal or have faded much too quickly to help the candidate in the next state.
Let’s first look at the trajectory of Mitt Romney’s polling in New Hampshire. Mr. Romney, it turns out, did not win the Iowa caucuses, but he was declared the winner by most news organizations and by the Iowa Republican party on caucus night, and the outcome was not reversed until after New Hampshire voted.
Mr. Romney did not receive any sort of bounce at all in New Hampshire after his apparent Iowa win. Instead, his polling was in a slow but steady decline between Iowa and New Hampshire:
Mr. Romney did slightly outperform his polling on primary night in New Hampshire (this is represented by the rightmost data point in the chart). However, that was mostly because he earned his share of undecided voters rather than there being any last-minute rebound toward him. He wound up with 39.3 percent of the vote — worse than the last several polls before Iowa, which had him at 41 percent to 46 percent.
What about Rick Santorum? Even though Mr. Santorum was not believed to have won the Iowa caucuses at the time New Hampshire voted, he did significantly outperform his polls in Iowa, something typically associated with a large bounce.
But Mr. Santorum wound up with 9.4 percent of the vote in New Hampshire. This was slightly better than where the polls had him before Iowa. But it was slightly worse than the polling conducted immediately after Iowa, which had Mr. Santorum between 10 percent and 14 percent. Mr. Santorum did get a bounce, but it was modest, and about half of it had evaporated by election night.
In the next state, South Carolina, Mr. Romney received a bounce of about 7 points after winning New Hampshire.
The bounce peaked about six days after New Hampshire, however — much too early for Mr. Romney. Instead, he lost 10 points in the polls over the next several days and finished with 27.8 percent of the vote in South Carolina — roughly where the polls had pegged him prior to his win in New Hampshire.
Finally, we can look at Newt Gingrich, who has been subject to a similar boom-and-bust pattern in Florida. The difference is that Mr. Gingrich’s bounce was even more fleeting. He held the lead in four surveys conducted within 48 hours of his South Carolina win — and he hasn’t led another one since. Instead, it appears as though Mr. Gingrich will lose Florida by 10 to 20 points — about what would have been predicted before his South Carolina surge.
In this cycle, the primary bounce has been behaving more like the bounce following a party convention — something that fades quickly and predictably and that does a candidate no real good in the end.
In 2008, I had built a “convention bounce adjustment” into the FiveThirtyEight presidential forecasting model but then removed it — most of the regular readers in the FiveThirtyEight community thought it smacked too much of fudging the numbers, and I concurred.
That was a huge mistake. The polls moved in the direction of Barack Obama immediately after the Democratic convention, very quickly reversed themselves and moved toward John McCain after the Republican convention, then settled back to show a slight lead for Mr. Obama — pretty much what the convention bounce adjustment had predicted. The forecasts would have been much more stable and much more accurate had I left the adjustment in — or for that matter, simply ignored the polling conducted during this period entirely.
The convention bounce adjustment will likely be reinstated in some form once we roll out the general election forecasts later this year. In contrast, I am not inclined to tinker with the primary model, which is deliberately designed to apply a “keep it simple” approach and to provide a benchmark that should be considered along with other salient information.
But I’m wondering if the same general principle holds — that it’s best to pretty much ignore any polling conducted in the two or three days after a major primary or caucus. The bounce from winning a primary seems to be fading more and more quickly — to the point that it may throw the forecast off the scent, representing a false dawn rather than real momentum.
It is tempting to say that this has something to do with the peculiarities of the Republican nomination process this year. And perhaps there is something to that: the polling has been extraordinary volatile in the Republican race all year to a degree that is unprecedented in past nomination cycles.
However, there was also some manifestation of this in 2008. Mike Huckabee got almost no bounce out of winning Iowa. Mr. Obama did get a bounce, but it faded so quickly that he lost New Hampshire five days later. Hillary Rodham Clinton won New Hampshire and Nevada — but then lost South Carolina by a margin much greater than the polls had predicted.
With the news cycle progressing as fast as it now does, winning a crucial primary or caucus may only be enough to ensure a day or two of strong headlines.
However, a win in a primary or caucus can produce the burden of higher expectations, one that is compounded when the candidate initially surges in the polls. (It may be significant that pollsters are now turning out surveys almost overnight, which contributes to the acceleration of the political news cycle.)
Then when the bounce evaporates, the candidate is perceived as losing momentum. This may not be entirely right; it may be better to think of this process as reversion to the mean, as in the case of a convention bounce.
Nevertheless, the perception that the candidate is failing to live up to expectations can have real consequences as negative news stories circulate about the candidate. The phantom loss of momentum turns into a real one, and the candidate may wind up in no better a position than where he or she started — or perhaps even be worse off.