If there is a path to the White House for Donald Trump, it almost certainly will follow roads like Interstate 75, 80, and 94 that rumble through the big battleground states across the Rust Belt.

Large Rust Belt prizes such as Ohio, Iowa, Michigan, Pennsylvania, and Wisconsin would be indispensable for any Trump general-election bid because they offer him the best chance to tip states primarily by widening the GOP advantage among the blue-collar white voters who have keyed his advance in the Republican primaries.

Whites without a college degree exceed their share of the national vote in all of those states except Pennsylvania (where they equal the national share). And in each of them, Democrats have run slightly better among those working-class whites than they have elsewhere—a pattern that helped President Obama carry all five states in both of his victories. If Trump, as the general-election nominee, can merely drive down the Democratic vote among working-class whites in the Rust Belt to the level of Obama’s national performance in 2012, he can threaten to take all of these states, especially Ohio and Iowa.

“I look in Iowa along the Mississippi River, where you have all these blue-collar people, John McCain and Mitt Romney didn’t communicate to those people, they did not relate,” says Craig Robinson, the former political director for the Iowa Republican Party and founder of The Iowa Republican website. “Trump, even with some of the problems his campaign has, I think he speaks to those people and they come into play. With his populist message and getting tough with companies that outsource, he has a market, and it is a market the GOP has struggled with.”

While acknowledging that potential appeal, other analysts in both parties believe Trump would face formidable obstacles in seeking to tip these metal-bending states back toward the GOP. Though the Rust Belt battlegrounds have been less affected by demographic change than more rapidly growing states across the Sunbelt, they have not been immune to it: In all of them, the share of the vote cast by non-college whites has declined over the past quarter century, while the share represented by college-educated whites and minorities has increased. And even if Trump can overcome the consistent tendency of working-class whites in these states to give Democrats slightly more of their votes than they do elsewhere, he faces the risk that the same hard-edged agenda and style that has electrified so many blue-collar white men will drive away other voters. That list could include not only Democratic-leaning groups like Millennials, minorities, and socially-liberal white-collar whites, but also working-class white women, who usually tilt mostly toward the GOP.

“The limits on this strategy include the countervailing effect among college-educated voters, Millennials and minorities,” said the long-time Democratic pollster Stanley B. Greenberg, who first became prominent for his studies of Michigan working-class whites. “But it may be equally important that half of the working class [i.e., women] are put off by him.”

Henry Olsen, a leading conservative electoral analyst and senior fellow at the Ethics and Public Policy Center, has been perhaps the top proponent of a GOP strategy that would try to win back the White House by maximizing blue-collar support across the Rust Belt. Olsen believes Trump is positioned to improve the GOP’s performance among working-class whites in the region. But he also believes “Trump would have a very difficult time” flipping these key Midwest battlegrounds precisely because many of the same ideas drawing blue-collar white men—such as mass deportation for undocumented immigrants and a temporary ban on Muslim immigration—will alienate the white-collar voters already displaying more resistance to him in the primaries.

“What he’s doing is offering a voice to those [white working-class] people’s beliefs and aspirations in a way that no Republican that has run for president has done,” Olsen said. “Which is why I think he could do better with them, but worse in the suburbs and that’s his challenge. If you do better, for instance, in rural Wisconsin but significantly worse in Madison and small suburbs of Green Bay and the exurbs of Milwaukee...you don’t do overall better. That’s the challenge he faces which is the reverse of the typical Republican challenge; they can speak to suburbanites but they can’t speak to these [blue-collar] people.”

Republicans like Olsen focus on the Rust Belt because turning states in this older and mostly white region offers the party a plausible, if narrow, path toward a 270-vote majority in the Electoral College, even without improving its performance among the Millennial and minority voters that the GOP has struggled to attract.

Any Republican calculation for 2016 starts by assuming the party can hold the 206 Electoral College votes that Mitt Romney carried in 2012. (Democrats believe that with Trump as the nominee that isn’t assured in North Carolina, Arizona, and maybe Indiana and Georgia.) But if Republicans can defend that base, they could reach a 270-vote Electoral College majority by adding Ohio (18 Electoral College votes), Wisconsin (10), Pennsylvania (20) and Michigan (16).

Since most GOP analysts acknowledge that Michigan is a very daunting climb for the party—and Pennsylvania is uncertain, too—the more common scenario among conservative strategists is to recapture Florida (whose 29 Electoral College votes Obama won last time by only about 75,000 votes). If the GOP can retake Florida, Olsen noted, it could reach 269 Electoral College votes just by adding Ohio, Wisconsin, and Iowa (6 Electoral College votes). At that point, any other state would give Republicans the victory. Alternately, an Electoral College tie would allow the House—which presumably would remain within GOP control if the party nominee runs that effectively—to select the next president.

Any of these Midwest-first scenarios, though, would run against the current of the region’s recent presidential history. Michigan, Pennsylvania, and Wisconsin are all part of what I termed in 2009 the “blue wall”: They are among the 18 states that have voted Democratic in at least every election since 1992. (That’s the most states Democrats have won that often in the party’s history.) Democrats have carried Ohio in four of the past six presidential elections and Iowa in five of the past six.

Over this period, Ohio has been consistently very close—among nominees from both parties since 1992 only Obama in 2008 exceeded 51 percent of the vote. But since the Blue Wall first emerged in 1992, Democrats have consistently posted bigger margins in Pennsylvania and especially Michigan. Wisconsin and Iowa were close during George W. Bush’s two elections (he narrowly carried Iowa in 2004), but both Bill Clinton and Barack Obama each won them relatively easily twice.

Yet these states constantly intrigue Republican presidential strategists because the Democratic advantage in them depends largely on an act of political levitation: the ability to consistently win a slightly greater share of working-class white voters here than almost anywhere else. Explanations for that Democratic advantage range from a greater union presence, to a smaller number of heavily Republican evangelical Christians, to a more vibrant tradition of class consciousness that has elevated economic loyalties over the conservative cultural affinities that often power the GOP gains among blue-collar whites elsewhere.

Whatever its cause, the effect is a small but significant advantage that has helped Democrats hold these states even as their national showing among working-class whites has cratered. In 2012, exit polls showed that Obama carried 36 percent of whites without a college education nationally but drew 40 percent of them in Ohio, 43 percent in Pennsylvania, 44 percent in Michigan, 45 percent in Wisconsin, and 50 percent in Iowa. In 2008, Obama exceeded his 40 percent national showing among non-college whites by double-digits in Iowa, Michigan, and Wisconsin, and by smaller margins in Ohio (four points) and Pennsylvania (two). In 2004, John Kerry ran at least six points better than his 38 percent national showing among non-college whites in all five states, and in 2000, Al Gore performed at least five points better than his 40 percent national number in each of them except Ohio (where he ran three points better).

That history frames both the opportunity and the challenge that Trump would face in trying to recapture these states by appealing to white working-class voters. Assume that the balance of the electorate in these states—college-educated whites and minority voters—split their votes between the two parties exactly as they did in 2012, and that all three groups comprise the same proportions of the 2016 electorate as they did in 2012. How much would Trump need to improve on Mitt Romney’s vote among working-class whites to flip these states?

In that scenario, Trump would need to reduce the Democratic vote share among working-class whites by nearly three percentage points in Iowa, five points in Wisconsin, six points in Pennsylvania, and about 11 points in Michigan. In Ohio, which Obama won by an eyelash, even a fractional decline among blue-collar whites for Democrats would tip the state.

At times, these states have seen election-to-election swings of roughly that magnitude (except for the level required in Michigan). Just from 2008 to 2012, for instance, Obama’s vote among working-class whites dropped by seven points in both Michigan and Wisconsin, and by four points in Ohio. From 2000 to 2004, the Democratic showing among working-class whites fell five points in Pennsylvania (before dropping another three points from 2004 to 2008).

From another angle, though, the climb looks more daunting: Trump could only capture these states by holding his 2016 Democratic opponent to less of the working-class white vote than any Democrat has won since 1988, according to exit polls. By that standard, Ohio and Iowa look the most achievable for Trump. In Iowa, if all else remained constant, he could win if he held the Democratic nominee to 47 percent or less of the vote among working-class whites, not much different from the party’s recent low point of 48 percent in 2004. In Ohio, Trump could win by pushing the Democrat even slightly below the recent low point of 40 percent that Obama notched in 2012.

In the other states, though, Trump would need to more forcefully shatter historical precedents. In Wisconsin, if all else remained constant, he’d need to hold the Democratic nominee to 40 percent or less among non-college whites; the party’s weakest recent performance was 45 percent in both 2004 and 2012. In Pennsylvania, Trump would need to reduce the Democratic nominee to 37 percent or less, well below its recent low of 42 percent behind Obama in 2008. And in Michigan, Trump would need to hold the Democrat to around 33 percent or less among working-class whites; that’s far below the low of 41 percent that Michael Dukakis hit in 1988.

Even these calculations are probably optimistic, because they assume that working-class whites will remain as large a share of the vote in 2016 across these states as they were in 2012. In fact, while the pace of change has been slower than nationally, working-class whites have also steadily declined within the electorate of these Rust Belt states, while college educated whites and minorities have grown.

Of the two data sources that measure the electorate’s composition, the Census Bureau’s election surveys consistently show non-college whites comprising a larger share of the vote than the exit polls conducted by a consortium of media organizations. But both sources show the same pattern of decline. Census Bureau figures indicate that from 1992 to 2012, the share of the vote cast by non-college whites fell by about one-eighth in Ohio, one-seventh in Iowa, one-sixth in Michigan, one-fifth in Pennsylvania, and fully one-fourth in Wisconsin. College-educated whites and minorities—each of which have shown much more skepticism about Trump in polls than working-class whites—have made up the difference. If this shift in the electorate’s composition continues, the hill for Trump as the nominee would grow more steep because he would need even bigger margins among non-college whites to flip these states, absent gains with the other two groups.

Using a slightly different technique that tries to harmonize the difference between the exit poll and Census estimates of blue-collar turnout, Ruy Teixeira, a senior fellow at the Center for American Progress, has produced for The Plum Line blog largely similar estimates of the white working-class gains that Trump would need to win these states. Generally the approach used by Teixeira, co-director of the States of Change project, requires slightly greater gains for Trump with working-class whites than the above calculations, which rely on the exit polls alone. But Teixeira’s simulation also shows Iowa and Ohio considerably more within reach for Trump than Wisconsin, Pennsylvania, or especially Michigan.

Trump’s supporters believe he could boost turnout among working-class whites, and his prospects would brighten if they increased their vote share from 2012. But given their steady pattern of decline, a successful mobilization might only result in their vote share remaining constant from 2012. Trump also faces the risk worrying Henry Olsen: that he will drive away college-educated white voters. In 2012, Romney carried college-educated whites by double-digits in Ohio, Michigan, and Pennsylvania, and lost them narrowly in Iowa and Wisconsin. Yet several recent national polls have shown likely Democratic nominee Hillary Clinton leading Trump among college-educated whites—a group no Democratic presidential nominee has carried in national polling dating back to 1952. If the price of advance among blue-collar whites is retreat among those white-collar voters, Olsen notes, Trump can’t close the gaps in these states.

“He badly needs to stem defections and the defections aren’t going to come from the hard-core right, they are going to come from...the business conservatives,” Olsen said. “The people who find his call for a [border] wall and ban on Muslims anathema. He needs to find educated suburban and high-income urban Republicans and bring them back to the party.”

Likewise, Teixeira argued, “Every action has an equal and opposite reaction. The concept that Trump could run as aggressive a campaign as possible to increase his white non-college share of the vote and not have a reaction among white college [is unrealistic]. You lose on one end what you gain on the other end.”

Greenberg, the Democratic pollster, sees that same risk for Trump, but also highlights two others. One is that Trump’s history of personal attacks on President Obama will inspire a large African American turnout. The other is the potential that Trump could underperform among working-class white women. In all five of these states, non-college white women constituted a larger share of the vote in 2012 than their male counterparts. And in the latest CNN/ORC national survey, while nearly three-fifths of non-college white men said they view Trump favorably, nearly three-fifths of non-college white women expressed unfavorable views about him.

“On the one hand, we have an unusual position where a Republican is willing to talk about trade and immigration and attack corporations that are not putting America first-that gives him an entry,” said Greenberg, who recently conducted a series of focus groups in Cleveland and Akron for the liberal group Women’s Voices Women Vote. “[But] it’s still a limited playing field. He’s turned off the women. He is driving white working-class women away because he is so sexist and chauvinist, and they view him as very insensitive to the struggles of working women, and Hillary [Clinton] does have an appeal there that plays out. You can’t do this on half the playing field.”

Robinson doesn’t dismiss these concerns. But he also believes Trump has the opportunity to connect with voters who believe the political and economic system is failing them at a visceral level—especially white working-class voters like those in blue-collar towns along the Mississippi River in eastern Iowa. “I grew up in eastern Iowa; I’ve done poll-watching in eastern Iowa,” Robinson said. “I remember looking at everyone who came in the door after 5 p.m. at night [after the work day ended] and saying ‘there’s no way in hell that person voted for my candidate.’ I think Trump gets these people interested and engaged and considering [to] vote Republican because he offers them the change in Washington that they so desire and will advocate on their behalf.”

Robinson, though, puts a big caveat on that possibility: It would require a focused, presidential Donald Trump. To attract those disaffected voters without suffering the significant white-collar defections that polls show him now facing, Robinson said, Trump must avoid the perpetual personal conflicts that have mostly characterized his campaign, and focus instead on a message of domestic economic renewal. (“If we are going to spend our time in the gutter he is going to turn those more reasonable people off,” Robinson said.) But if Trump can maintain “message discipline” around an agenda to “produce things in America with American workers,” Robinson predicts, “he can put things in play.”

If Trump wins the nomination, these burly Rust Belt battlegrounds represent his best chance to shake up the electoral map as Robinson hopes. If he can’t unbolt them from the Democratic Electoral College majority, it’s likely that his path to the presidency will quickly reach a dead end.