By Nate Silver – – – – –
The way some Democrats like to tell the story, Nevada went from red to blue without ever really stopping at purple. With its rapidly growing Latino and nonwhite populations, Nevada has demographics that increasingly look like California’s. And President Obama not only won the state twice, but did so by more than his national margins of victory, claiming it by 12.5 percentage points in 2008 and 6.7 percentage points in 2012.
|YEAR||NEVADA RESULT||NATIONAL RESULT||NEVADA RESULT RELATIVE TO NATIONAL RESULT|
A generation ago, the idea of Democrats relying on Nevada would have seemed ludicrous. Ronald Reagan won the state by 35.6 percentage points in 1980, even as he beat Jimmy Carter by “only” 9.7 percentage points nationally. Bill Clinton won Nevada in 1992 and 1996, but both years were close calls compared with his relatively emphatic national margins. Democrats have come a long way there.
So far, though, polls have shown a relatively close race in Nevada, even as Democrat Hillary Clinton has a fairly clear lead on Republican Donald Trump nationally. Granted, there hasn’t been a lot of data. On Monday, Monmouth University published a poll showing Clinton leading Trump by 4 percentage points in Nevada. You can nitpick at the small sample size — 408 likely voters — but that’s an improvement over the table scraps we had to work with previously. Before Monmouth, the last publicly available, Nevada-specific poll had been way back in November. There were also subsamples from polls such as this one that surveyed several swing states, although they also had small sample sizes.
Those earlier polls had shown a tied race or Trump narrowly ahead. Our “polls-only” model gives the most weight to the highly rated Monmouth poll, but it considers those other surveys also. Thus, its adjusted polling average in Nevada puts Clinton ahead by only 1.6 percentage points.
The polls-only model doesn’t take that for a final answer, however. Instead, it blends that polling average with what we call a “demographic regression.” Since Nevada doesn’t have much polling, the regression gets a fair amount of weight.
Wait a second … why are we calling it a “polls-only” model if it looks at demographics also? Because all the inferences it makes from the regression are based on polls of other states.1 If you tried to infer what was going on in Nevada based on polls of other states, you’d assume that Clinton was relatively far ahead there. In general, Clinton is polling quite well in states with large Latino populations, such as Florida, Arizona, California and even Texas (she’s losing in Texas, but not by as much as Democrats usually do). She’s doing fine out West, even in Utah, although Colorado, where there also hasn’t been a lot of polling, looks closer than expected. Her numbers in other swing states aren’t extraordinary, but they’re pretty good, basically matching her national numbers.
Therefore, the regression estimate thinks Clinton “should” be winning Nevada by 10.2 percentage points, based on the polling it sees elsewhere. But Clinton’s polls from Nevada itself are mediocre, so the model compromises and projects her to win the state by 5.2 percentage points. Notably, that’s slightly worse for Democrats than the country as a whole. (The polls-only model has Clinton ahead by 6.2 percentage pointsnationally.) If Clinton maintains her current national lead over Trump, she’ll probably win Nevada — but if the race tightens, it’s a long way from a safe state.
As far as our model is concerned, the analysis basically stops there. It sees a gap between polls and demographics in Nevada, and it compromises. The terms of the compromise are dictated by the volume of polling and how long we have to go until the election.2
But as a sanity check, it’s worth thinking more about Nevada. How plausible is it that Nevada behaves idiosyncratically — that it becomes redder even as seemingly similar states are becoming bluer?
In my view, it’s plausible — Nevada is an idiosyncratic state (I mean that in the nicest possible way). Let us count the ways:
- Nevada has something of an anti-establishment, anti-incumbent streak. Obama’s margin fell quite a bit from 2008 to 2012. Reagan won Nevada by less in 1984 than he had in 1980, even though his margin of victory increased nationally. The same held for Bill Clinton in 1996 and George W. Bush in 2004.
- Nevada suffered greatly during the housing crisis, and its recovery is a long way from complete. Its unemployment rate is 6.1 percent, tied for the sixth-highest in the nation.
- The gambling (ahem, “gaming”) industry, which is heavily associated with Trump, is a huge source of employment in Nevada, although ironically the Trump Hotel in Las Vegas has no casino floor.
- Nevada has an option to vote for “none of these candidates.” It’s not clear whether that benefits Trump or Clinton, but it’s certainly different. It also might explain why Libertarian Gary Johnson is polling relatively poorly there, since “none of these” gives voters an alternative way to cast a protest vote.
- For a variety of reasons, turnout tends to be low in Nevada. In 2012, there were barely more than 1 million votes there. By contrast, there were almost 1.6 million votes in Iowa, whose population is only slightly larger than Nevada’s.
All that said, there are reasons to be skeptical that Nevada will wind up being as tight as polls have it now. Polls there have tended to underrate Democrats — most notoriously in 2010, when Harry Reid won re-election to the Senate despite public polls that showed him behind Republican Sharron Angle. Democrats have a large and expanding voter-registration advantage. To the extent that it’s a low-turnout state, that probably puts more emphasis on the ground game, which presumably works to Clinton’s benefit.
It’s even plausible that the polls are right and Clinton is only a couple of points ahead right now — but that she’ll be up 8 or 10 points by November, after the campaign has kicked in and pollsters have a better idea of who will turn out to vote. But few advantages are permanent in Nevada, and Clinton, who is advertising heavily there, can’t take a lot for granted.
- And to a lesser extent, based on national polls. ^
- It’s pretty early, and there isn’t much polling in Nevada, so the demographics get a fair amount of weight — about 40 percent, whereas polls get 60 percent. By November, demographics might get only 5 percent or 10 percent of the weight, assuming there’s a lot more polling in Nevada between now and then. ^
Nate Silver is the founder and editor in chief of FiveThirtyEight.