Should Trump lead— La Times be ignored?

I go on sites to check the polls every few days— 538, New York Times, Sam Wang Princeton Consortium, then the Stossel betting odds. I check to see the percentage of chance for Hillary to win— and if, on all those sites, her odds are 65% or higher, I feel a little bit better about the world. And since right now the percentages are higher than 65% I feel much better.

But, then I go to the polls posted by the Los Angeles Times:
And Trump is winning by 2 points on the Los Angeles Time poll :confused:
How much should this worry me?
Do you think there is a chance the La Times polls could be right that Trump has a slight lead. Or should the La Times be ignored.


It is a tracker, not really a poll. Of some value for trending and very little as an absolute value. Reuter/Ipsos is similar. In this case it has moved from Trump +6.7 9/18 to Trump +3.6 now - a move of 3.1 Clintonward, not far off from what the RCP two-way rolling average has moved in that period.

You handle it though as part of how you handle polls and have rules set up to not overweight trackers which report often. 538 has its way to weight it (and does) or you can use Wang’s approach and toss out the top and bottom outliers of your recent data set.

Short version is that the aggregators have considered it as part of how they come up with those odds.

No voted on Brexit was (IIRC) never less than 70% odd on favorite to win, so don’t count you Clinton’s before they are sworn in.

The polls were NOT off on Brexit. Really. Eve of the election the aggregate was within 2% and leave had been ahead in the week before. In the two weeks before more polls had Leave ahead than Remain.

The betting market stunk the place up to be sure. But the polls called it as a toss up. And in toss ups turnout is the key. Younger voters who preferred Remain did not turn out so well; older voters who prefered Leave did. That was enough to tip the close scale.

In my understanding, the LA Times poll is unique in that they poll the same people every week. At the beginning of their poll, they identified 2500 or 3000 people or something to agree to be part of a daily or weekly poll on their preference. If they happened to pick, for some reason, right-leaning voters that first time, it would be reflected for every poll since then. Further, you’ll notice that their sample size is a bit different every week – if the Trump voters are more excited, then they might be more willing to pick up the phone and talk to the voters for a given week than less excited Clinton voters, even if they both will definitely vote.

In short – don’t ignore it, but don’t elevate it above other polls. It contributes in a small way to the larger picture just like other polls. Aggregates, especially from guys with great track records like Nate Silver, are best.

538 counts it, but with a pro-Trump house effect. As with any tracking poll, seeing how it goes up and down is more important than the single day result. In that respect, the poll is doing the same thing other polls are doing, moving a bit towards Clinton. Although the state polls seem to be moving her way more strongly than the national polls.

The LA weekly just published this today:

Continued here:

Good to know!