538 Election Update: Clinton’s Lead Is Clear And Steady
Im going through this somewhat tedious explanation because its basically how FiveThirtyEights forecast models calculate their trend line adjustment, which works by comparing polls against previous editions of the same surveys. The idea here is simple. If Trump is gaining ground on Clinton overall, we ought to see him gaining ground in individual polls (including state polls, which are also used to calculate the trend line adjustment). But he isnt gaining on Clinton, mostly. In polls that had previously surveyed the race since the conventions, Trump is up a point or so in some polls and down a point or so in others the overall trend is about flat. And in polls that are taking a post-convention snapshot for the first time, Clinton is generally polling better than she did before the conventions. Thus, our polls-only model continues to show Clinton ahead by around 8 percentage points. And it gives her an 89 percent chance of winning, close to where shes been over the past week.
Still, Im not quite ready to declare that Clintons lead is completely convention bounce free and that weve reached a new equilibrium in the race. Thats because Id like to see another round of high-quality, traditional telephone polls for instance, the NBC News/Wall Street Journal poll or the ABC News/Washington Post poll. Those polls generally showed excellent results for Clinton just after the conventions, with a larger bounce than some of the tracking polls had. But none of them have surveyed the race twice since the conventions, to give us a sense for whether Clintons bounce is holding. This more conservative attitude is closer to the one our polls-plus model takes, which still assumes the race is more likely than not to tighten. It shows mildly brighter prospects for Trump, giving him a 22 percent chance of winning.
http://fivethirtyeight.com/features/election-update-clintons-lead-is-clear-and-steady/