538 Model Megathread

[Gustaf]

Serious question - why does the Nowcast do a trendline? That doesn't make sense to me. If Clinton leads in Nevada now (as she does per their poll model) why would the Nowcast not predict her as ahead?

[Gustaf]

Serious question - why does the Nowcast do a trendline? That doesn't make sense to me. If Clinton leads in Nevada now (as she does per their poll model) why would the Nowcast not predict her as ahead?

Maybe I'm misunderstanding something, but the latest 5 it's showing have Trump up in 3 and Clinton up in 1. The most recent is Clinton +4, but it gets adjusted to a tie based on other factors (e.g. national polling, covariance with other states). So I guess it doesn't surprise me.

If you look at the Nevada Nowcast it has a polling average of Clinton +1.5%.

Adjustment for some of their stuff brings it to Clinton +1.2%.

THEN they adjust for "trend" which swings it by 4.5% in Trump's favour and gives it as Trump +3.4%.

I don't get what trend they're adjusting for if it's "election held today". The net swing is roughly the same as in the forecasts for November 8th.

[Gustaf]

Serious question - why does the Nowcast do a trendline? That doesn't make sense to me. If Clinton leads in Nevada now (as she does per their poll model) why would the Nowcast not predict her as ahead?

Maybe I'm misunderstanding something, but the latest 5 it's showing have Trump up in 3 and Clinton up in 1. The most recent is Clinton +4, but it gets adjusted to a tie based on other factors (e.g. national polling, covariance with other states). So I guess it doesn't surprise me.

If you look at the Nevada Nowcast it has a polling average of Clinton +1.5%.

Adjustment for some of their stuff brings it to Clinton +1.2%.

THEN they adjust for "trend" which swings it by 4.5% in Trump's favour and gives it as Trump +3.4%.

I don't get what trend they're adjusting for if it's "election held today". The net swing is roughly the same as in the forecasts for November 8th.

They are looking at how the national polls have changed since those Nevada state polls were conducted, and assuming Nevada has gone to the right at the same pace since then as the rest of the country (even though we don't have recent Nevada polls to confirm this).

Yeah, now-cast doesn't mean you take all the polls as is at the time they were taken.  It means you take the polls and adjust them for any trends in national polls that have happened since they were taken, to see what they would look like if taken right now.

I thought about that but then why are the trends not a lot stronger for the November forecasts? It's cancelled out by the polls themselves being valued less?

[Gustaf]

I think they forgot to update their nowcast. Pennsylvania should be flipped.

[Gustaf]

What model did Silver use in 2012? Polls-Plus, Polls-Only, Or Now Cast? I thought he only used one model in 2012.

I think all the models will tend to converge as the election approaches.

[Gustaf]

It's stunning that anyone could still believe Trump is gonna be surrounded by good advisors able to rein him in.

[Gustaf]

The model now weights down the 50 state polls.

[Gustaf]

With the latest PPP poll dumn Nowcast is now up to 68% Clinton and she is breaking 60% in all 3.

[Gustaf]

With the latest national polls added in Clinton went up further:

Now-cast: 76.3% (Obama2012 - OH, +NC)
Polls-only: 68.7%(Obama2012 - OH, IA, +NC)
Polls-Plus: 65.2% (Obama2012 - OH, IA)

[Gustaf]

...and there Clinton broke 90% in the now-cast!

[Gustaf]

Yeah I think it's mainly a polling issue. Hopefully we'll soon get more polls out of Georgia, South Carolina and Texas.

[Gustaf]

One thing that always stood out to me as weird in the 538 is their scenarios stuff at the end. Currently in the Nowcast the models says:

Clinton wins at least one state Mitt Romney won in 2012           86.4%
Trump wins at least one state President Obama won in 2012   54.1%

This seems really weird to me. It implies that they assign at least a 40% probability of Trump winning an Obama state while Clinton wins a Romney state. Generally these numbers have always struck me as too high throughout the campaign.

[Gustaf]

One thing that always stood out to me as weird in the 538 is their scenarios stuff at the end. Currently in the Nowcast the models says:

Clinton wins at least one state Mitt Romney won in 2012           86.4%
Trump wins at least one state President Obama won in 2012   54.1%

This seems really weird to me. It implies that they assign at least a 40% probability of Trump winning an Obama state while Clinton wins a Romney state. Generally these numbers have always struck me as too high throughout the campaign.

Well, they're not independent events, so the odds of both happening are not quite that straightforward.  But it does seem like something that could easily happen, e.g. Trump winning Iowa while Clinton wins NC.

Right, but them not being independent events makes the situation worse, because it makes it less likely Trump outperforms the polls to win Iowa while not at the same time maintaining North Carolina. 

My point was that it's at least 40% given those numbers. Which sounds high to me.

When the race was closer this was even more glaring.

So, to Zombie Spenstar - while states don't correlate perfectly, Clinton's win prob in NC is only marginally higher than that in IA and OH. It seems to me that there should be a fairly tiny slice of the prob dist in which she retains NC but loses one of the other ones.

I guess they have a lot of idiosyncratic uncertainty but these numbers indicate to me that maybe that uncertainty is too high. Is my point I suppose.

[Gustaf]

So, to Zombie Spenstar - while states don't correlate perfectly, Clinton's win prob in NC is only marginally higher than that in IA and OH. It seems to me that there should be a fairly tiny slice of the prob dist in which she retains NC but loses one of the other ones.

I guess they have a lot of idiosyncratic uncertainty but these numbers indicate to me that maybe that uncertainty is too high. Is my point I suppose.

The states are also different in their elasticity and demographics. Should the race tighten again, Iowa will probably go atlas-blue before NC. And if Clinton's support falls among white voters without a college degree but remains steady with college-educated whites and nonwhites, that would be terrible for her in Iowa but wouldn't change too much of the math in NC.

Yeah, I'm aware of those aspects of the model. I guess I'm just skeptical of it being true to such an extent.

[Gustaf]

With the Bloomberg Clinton +9, the PRRI Clinton +15 and the Arizona Clinton +4 poll, 538 will most likely have Clinton at or slightly above 50% of the popular vote in at least one of their models, possibly two (polls-only and Nowcast)

This actually resulted in Clinton's win % dropping by 0.1, huh?

They haven't posted the PRRI yet, ftr. And the -0.1 effect comes from lumping in the other polls with the LA tracker etc.

[Gustaf]

Most of the adjustments do come from trend lines. That isn't unreasonable in itself but I think the problem is that these trend lines are largely derived from all the dumb trackers. And that's where they gain undue influence. The last few days the model has been insane, IMO. Most of the win probability adjustments have been going the wrong way.