The 270 Project: Try To Predict Who Will Win The Election

Donald Trump has staked his brand on winning. "We will have so much winning," he has said in this campaign, "if I get elected, that you may get bored with winning."

But can he win the presidential election? In a country that has changed rapidly demographically, Trump's best shot is to drive up turnout among white voters, especially white men. But how likely is that?

We at NPR Politics wanted a data-driven, quantitative way to answer the fundamental question of whether Trump can win, or if this is Hillary Clinton's race to lose — and give readers the power to test it out themselves. There's perhaps no better way to do that than through demographics. It's by no means a crystal ball, but how we identify is arguably among the best predictors of how we will vote.

So we created The 270 Project, a handy tool where you can adjust voter turnout and margin of victory for five demographic groups — white women, white men, African-Americans, Latinos and others (Asian, Native American, mixed race) — to see what it would take for Trump or Clinton to win. (The project derives its name from the number of electoral votes needed to win the White House — 270.) Give it a shot and see what you come up with — and there's more analysis from us below.

The fascination with whether Trump can win hinges on what an unconventional candidate he is. The New York billionaire is eschewing political norms — spending less money than what the experts say is needed; hiring far less staff than Clinton; traveling to states that aren't traditional battlegrounds; taking positions on things like trade that run counter to the standard GOP line; and continuing to say inflammatory, controversial things that threaten to alienate record numbers of nonwhite voters in an era in U.S. history when demography is rapidly changing.

The U.S. has become less white and more diverse. The lowest share of white voters in history is expected to head to the polls in November, continuing a steady decline over the past three decades that is likely irreversible. (Consider that the current under-age-5 population in the country is now majority minority.)

Demography alone appears to give Clinton a clear advantage in this election (see the latest NPR Battleground Map). But Trump argues the usual rules don't apply to him. So is he right? Can Trump win or is this Clinton's race to lose?

In this analysis, we take a deep dive into several potential scenarios (that should also help you use the tool), including the importance of white men to Trump; why Latinos are a key firewall for Clinton; and how turnout could affect the result. But first a little bit about how to use the tool and the overall demographic and political landscape.

How did we come up with this tool? And how do you use it?

There's more on the methodology of how we built this tool and the data sources behind it at the bottom of this post. But in a nutshell, with the help of William Frey, a demographer at the Brookings Institution, we used 2016 demographics and superimposed that on (1) 2012 exit poll data and (2) 2012 census raw voter turnout statistics. That way we could see how a generic Democrat or Republican would do today, assuming he/she got Mitt Romney/Barack Obama levels of support.

We then looked at the 2012 election map, and created a list of 19 states based on the guidance that NPR political editor Domenico Montanaro outlined earlier this spring — a combination of traditionally accepted battleground states and potentially contested states that lean Republican or Democrat. And for kicks we added New York, because Trump has repeatedly promised to win his home state.

We've built interactive sliders, where you can adjust the turnout and the margin of victory to see what Trump needs to win. When the slider is at 0 percent, the margin of victory is set at 2012 levels and the same for turnout. Our sliders move at 1 percentage point intervals for all the states at a time, and under the hood, they adjust each of the 20 states accordingly.

So, for example, if you want to bump up the white-male margin of victory by 1 point, that means in Georgia, the Republican candidate would win 80 percent of white men, but in Iowa the Republican candidate would win only 56 percent of white men. Even though you can't see it, the data are state-specific. And, as you move the slider one direction over the other, you'll see on the side that the states change color.

2012 vs. 2016: Obama 3.0 would win by even more

Demographics alone clearly favor a Democratic candidate.

President Obama won the 2012 election by 126 electoral votes — 332 to 206. And in the past four years, the population of the country has changed in ways that would favor Obama even more.

Put it this way: If Obama 3.0 were running for president today and performed with the same margins as he did in 2012 with our five demographic groups, our model shows he would win by an even larger margin — 152 electoral votes: 345 to 193. Three states would flip: North Carolina and Georgia would go to Obama — and Ohio to the Republicans.

We should note a big caveat with Georgia. "If Georgia came out and voted Democratic this time it would be the surprise of the election," said demographer Frey with a laugh. Frey, like most political analysts, thinks of Georgia as Republican state, though Clinton and Trump are polling fairly close — and Obama lost the state by just 5 percentage points in 2008 and 8 points in 2012.

Georgia's demography has changed so quickly that Frey acknowledges "it's not totally impossible" for Clinton to win. The white population has gone from 70 percent of the state in 1990 to just 54 percent, as of last year. Since 2004, the eligible black voting population there has increased six times faster than whites, according to census figures.

"Georgia is a big gainer of black migrants from all across the country," Frey noted. "The trick in Georgia for Democrats is to make sure the black turnout ... is so dramatic that it overtakes what is typically strong [white] support for a Republican candidate."

And that's not easy to do. In fact, if you boost white male turnout and margins of victory in both of these states, they turn red again.

White men and their importance to Trump

This November, the electorate is expected be the most racially diverse in history. But still an estimated 72 percent of voters are expected to be white, according to a Brookings Institution March analysis of Current Population Survey numbers from the Census Bureau.

Those white voters hold tremendous political power, and Trump is trying to boost their power even further. His basic demographic premise is twofold:

1. Boost white male turnout, and

2. Win them by more than Romney in 2012, for example.

So, we tested those ideas separately and then together.

First, we increased white male turnout in each of our 20 "battleground" states by 2 percentage points over 2012. And while that helps Trump, he's still far from the finish line. Our analysis shows he would win Ohio and Pennsylvania, but still lose the election in a landslide 112 electoral votes, 325 to 213.

We then tested the second idea — that Trump could win white men by 4 points more than Romney in 2012, but maintain their level of turnout. We held all other demographic groups constant. If Trump could do that, he would significantly narrow the Democratic path to the White House, but he would still come up just short — by six electoral votes, 266 to 272.

Now, if he manages to simultaneously do both — boost turnout among white men (by 2 points) and also win them more convincingly (4 points more than Romney) ...

We found that with this strategy Trump could theoretically flip Colorado, Florida and Pennsylvania and win the election, 282 to 256.

That's still a narrow win — and is contingent on Trump's getting the same levels of support with white women, Latinos, Asians and African-Americans that Romney did.

Florida, Florida, Florida — and the potential Clinton Latino firewall there

In the simulations above, we keep Trump's support with nonwhite voters constant. But that's probably not likely — given how controversial Trump has been, particularly with Latinos.

That could insulate Clinton even if Trump is able to pull off wins in places like Ohio and Pennsylvania on the backs of blue-collar white voters. Our demographic analysis shows that an estimated 19 percent of Florida voters this November are likely to be Latino. The Hispanic community in the state is changing — and becoming more liberal. Second-generation Latinos and an increase of Puerto Ricans who have moved to the state are challenging the traditional power structure in the state and the dominance of older, conservative Cuban-Americans.

For Trump to win Florida, Hispanics would have to support him at the same levels they voted for Romney. But that could be tough, not just because the makeup of the Latino community is changing but because Romney won 39 percent of Hispanic voters there in 2012. If Trump can't do that, it becomes very difficult for him to take the key state.

Nationally, Romney got 27 percent of the Latino vote, a record low for Republicans. What did him in was his comment about "self-deportation," the idea that immigrants in the U.S. illegally would leave on their own because the environment would be so unwelcoming toward them in the U.S.

But that pales in comparison with the way Trump has talked about Latinos — in his announcement speech, he equated Mexican immigrants in the U.S. illegally to "rapists," and he has accused a judge of being biased against him in a Trump University fraud case because he's "Mexican." The judge, born in Indiana, is of Mexican heritage.

Recent national polls show Trump capturing closer to 21 percent of the Latino vote. That kind of figure would imperil Trump's chances in Florida, provide a cushion for Clinton in places like New Mexico and Nevada, give Clinton a better chance in Colorado and possibly even provide hope for her in a place like Arizona. There are also smaller, but growing Latino populations in North Carolina, Virginia and elsewhere.

But Florida is the battleground state with the most electoral votes where Latinos would matter. And if we take what we gave Trump with white men and adjust Trump's level of Latino support to his current polling average of 21 percent, Florida would flip to blue, and Clinton would win the White House, 285 to 253.

The Midwest wild cards — Wisconsin and Michigan

If Trump loses Florida, he'll need to find another state (or two) to make up for those 29 electoral votes. Given his message on trade and how he has flouted GOP dogma in an attempt to appeal to blue-collar workers, that means looking toward the Rust Belt — and trying to win Wisconsin or Michigan.

So how would it work?

Wisconsin: Assuming Trump increases turnout among white men by 2 points and everything else remains at 2012 levels, he would need to boost his margin of victory by a whopping 8 points over what Romney did. In other words, Trump would need to win 64 percent of white men in Wisconsin compared with Romney's 56 percent.

That's very tough to do, and Clinton maintains the advantage there. But for argument's sake, even if you gave Trump Wisconsin, unfortunately for him it wouldn't be enough.

Michigan: Trump also has Michigan in his sights — and he might need it. "I'm going to win places like Michigan that the Republicans can't even think of," Trump has said.

Demographically, it's plausible. Michigan is a state Obama won twice, but between 2008 and 2012, his support with white voters dropped there significantly. Obama won whites in Michigan, 51 percent to 47 percent, in 2008. But, in 2012, he lost them 55-44, a net 15-point drop.

"It's really the home of the traditional Reagan Democrats," Frey said, "and that's a prime group for Donald Trump."

Assuming Trump increases white male turnout in Michigan by 2 points, and everything else remains constant, he would need 63 percent of white men, 5 points better than Romney.

A tie!? If everything else holds, and Trump were to win Michigan, but not Wisconsin, that could present a potential historic problem — an Electoral College tie. It's something that's happened just once in U.S. history, in 1800, when Thomas Jefferson and Aaron Burr both got 73 electoral votes each. (Jefferson became president after it was settled by the U.S. House.) The House would again break the tie with each state receiving one vote — and that would favor Republicans, given the current makeup of the House.

The best-case scenario for Trump is winning both states, which would give him a 279-259 win. But Frey cautions that Michigan is a "long shot" because of the state's strong left-leaning union history.

Still, Trump might need the state. His most practical road to the White House goes through picking up the leaning-Democratic states of Michigan and Wisconsin, plus the tossups of Ohio, Pennsylvania, Colorado and North Carolina — along with holding the leaning- or likely Republican states of Arizona, Georgia, Indiana and Missouri.

If he can squeeze through all of those states, he would be the next president.

"I've always said that this idea that Donald Trump could win by turning out very large numbers of ... white men is what I would call an 'inside straight' strategy if you were a poker player," Frey explained. "It means it's a very rare kind of thing, but if it happens you get to win and pick up all the chips."

It all comes down to turnout? Maybe

All of these scenarios so far have assumed that turnout would remain at 2012 levels for everyone except white men. But that's unlikely.

Most analysts predict Trump may spark an increase in turnout among Latinos and Asians. On the other hand, Clinton may have challenges with black voters in ways that could benefit Trump.

Let's explain: African-Americans came out to the polls at historic levels in 2012. That year, the raw national black turnout rate exceeded that of whites. It's possible, if not likely, though, that black voters will not show up in quite the same numbers this November with Obama not on the ticket.

So what if black turnout reverts to 2004 levels when John Kerry was the Democratic standard-bearer? In 2004, an estimated 60 percent of eligible black voters participated in the election. That's 6 points less than black turnout in 2012.

Increased white male and Latino turnout plus decreased black turnout: OK, so give Trump a 2 percentage point turnout increase with white men; Clinton, a 2-point turnout increase with Latinos and a 6-point drop with black voters.

In that scenario, Trump would lose the election by 50 electoral votes, 294 to 244.

That, though, doesn't take into consideration the margin with black voters. Clinton is pulling Obama-like margins currently. The latest NBC/Wall Street Journal poll, for example, has Trump at just 5 percent with black voters. (In 2012, Romney got 7 percent.)

And, keep in mind, we haven't adjusted turnout for white women at all. Most current polling shows the Republican advantage with white women to be less than 2012 (when Romney won them by 14 points).

"It could be in these states," Frey noted, "where the white males are voting more strongly for Donald Trump, white women, their wives, may either stay home or maybe even vote Democratic to a greater degree than they did before."

Things that make you go hmm ...

Here are three key takeaways from all this, aside from just how narrow Trump's path is:

1. Virginia, which has long been considered a traditional battleground state, seems fairly convincingly Democratic this election cycle based on the demography. It remained loyally blue in all of the hypothetical scenarios we outlined above.

2. Ohio and Pennsylvania both seem feasibly winnable for Trump if he can boost white male turnout by a few points. Ohio, in particular, is possible, which is why you're starting to see the AFL-CIO and labor leaders, who back Clinton, begin to push back against Trump's trade message:

Just a 1.5 percentage point shift in how white women vote could flip Ohio.

In Pennsylvania, we found that just a 1-point change in the white-male margin could give the state to Trump. "If Donald Trump was able to bolster up the turnout, bolster up the enthusiasm, especially in Southeastern Pennsylvania ... to counter that kind of East Coastal population you see in suburban Philadelphia," Frey said, "he would have a shot at winning. I think Pennsylvania is winnable for Donald Trump, but everything would have to go his way."

3. New York: "I will win New York against Hillary Clinton," Trump promised at a campaign stop this spring. It's a claim he's fond of reiterating, and since he has articulated a specific desire to win New York, we wanted to see what it would take for him to turn his home state red. Assuming all other demographic groups vote exactly as they did in 2012, and assuming turnout also remains constant, Trump would need to win 97 percent of white men in New York. 97 percent.

Methodology

With the guidance of Brookings Institution demographer William Frey, we broke the current electorate into five categories: white men, white women, black, Latino and other.

We used 2012 turnout rates as compiled by the census for individual states and multiplied that by the current number of 2016 eligible voters, as indicated by the March Current Population Survey as citizens above the age of 18.

That gave us an estimated voter population for each category. We then multiplied that by the support for Romney and Obama, according to 2012 exit polls for each group, to achieve a baseline estimated level of support for Trump and Clinton in 2016. In some cases, there were no state-specific exit poll data for some groups, and in those instances, we substituted national exit poll numbers.

For our "other" demographic, which includes Asians, we averaged the 2012 exit poll data for "Asian" and "other" populations in most states. A few exceptions: In North Carolina, we averaged the national Asian exit poll results with the statewide "other" exit polls. And in Virginia, since there is a sizable Asian population, we used the 2012 Asian exit poll data for our "other" category.

A note on Georgia: Since it was excluded from the 2012 exit poll surveys, we averaged exit poll data from two very different election cycles — the 2008 presidential election (which was a high watermark for Democrats) with the 2014 midterm election (a stronger year for Republicans). It's by no means perfect, but part of the reason we averaged the numbers is that white and black voters in Georgia seem to vote along deeply segregated lines with white voters overwhelmingly voting Republican and black voters consistently voting Democratic in presidential election years.

We did not account for a viable third-party candidate, partly because we were modeling our calculations on the 2012 election. So in all of our "margin" scenarios, a subtraction from the Republican candidate is an addition to the Democratic candidate and vice versa. But in reality it's plausible these votes could go to a libertarian or Green Party candidate instead.

For those of you who have reached this far, you deserve an Easter egg. Take a look at our number-crunching spreadsheet.

NPR's Domenico Montanaro contributed to this report.

Copyright 2016 NPR. To see more, visit NPR.