When I think of primaries that really cost an incumbent president in the general election, I think of Teddy Kennedy in 1980 and Pat Buchanan in 1992. Both of those campaigns served as giant in-party critiques of a sitting president that caused lasting divisions and hard feelings. I don’t think Ronald Reagan’s 1976 run against President Ford helped Ford’s cause, either. In the case of challengers to a president, I do think that the candidacy of Bernie Sanders created divisions that didn’t heal, and these were artfully and criminally amplified by the Russians and WikiLeaks in coordination with the Trump campaign.

General election challenges by third party candidates have a different character. In 1968, most George Wallace voters would have cast a vote for Richard Nixon in a two-way race. Nonetheless, most of the Wallace voters were accustomed up to that time to casting a vote for the Democrats. Wallace served less as a spoiler than a way station on the voyage between being a Democrat and a Republican. Nixon won the election narrowly, but he would have won by more if Wallace had not been an option. Essentially, he helped the incumbent party.

In 1980, the candidacy of moderate Republican John Anderson seems to have done less to divide the GOP than to drain away votes from Jimmy Carter. On balance, it hurt the incumbent party.

There has been a lot of contentious debate about whether H. Ross Perot’s independent bid in 1992 primarily hurt President George H.W. Bush or his support was basically split between the incumbent and challenger Bill Clinton. An exit poll at the time suggested it was more of a split, but Clinton probably carried Georgia and Montana in that election solely due to Perot splitting the right. I don’t think I can identify any states Poppy carried due to Perot splitting the left. On balance, I think it’s safe to say that Perot hurt Bush more if for no other reason than he savagely attacked his presidency and created a two-on-one dynamic.

In 2000, it’s clear that both Pat Buchanan and Ralph Nader hurt Al Gore despite the fact that neither of them got a lot of votes. A flawed ballot design in Palm Beach County, Florida caused a lot of voters, many of whom were Jewish, to erroneously cast their ballot for the anti-Semitic Buchanan. This alone may have changed the outcome of the national election. Likewise, it wouldn’t have taken many Nader voters in Florida casting their ballot for Gore instead to reverse the outcome there. It’s safe to say that in 2000, third-party candidates hurt the incumbent party.

In 2016, it’s arguable that third party candidates (particularly, third party candidate Jill Stein) may have cost Hillary Clinton the election. If true, it was because of lingering divisions in the Democratic primary that were stoked by the Russians. Again, the incumbent party suffered.

The general trend then, is that an incumbent party is hurt by a third party challenge. The effect may be modest but in a close election a modest shift in the vote can be all that is required to change the outcome.

Of course, this is a very simplistic way of looking at history. It ought to matter a great deal where the third party candidate stands on the ideological spectrum. A right-leaning candidate would logically seem more likely to take votes from the Republican Party’s nominee and a left-leaning candidate should do the same for the Democrat’s nominee. But history also shows that things have rarely been that simple.

George Wallace was a Democrat whose vote was historically Democratic, and he split the Democratic vote. The problem with that analysis is that most of those Democrats would have voted for Nixon in 1968 and did vote for him in 1972.  John Anderson was a Republican alarmed at the conservative drift of his party. He served as a way station for more liberal Republicans who were not yet ready to cross-over and vote for a Democratic incumbent they had not supported four years earlier.  Much like Wallace before him, H. Ross Perot had a mixed effect but he also gave people dissatisfied with the president but unwilling to cross over a safe place to land.

Looking at this complex history, I do not agree with the seeming consensus that a third party bid by Starbucks CEO Howard Schultz is likely to split the Democratic/anti-Trump vote and make Trump’s reelection more likely.

My analysis is highly colored by my experience watching my home state of Pennsylvania shockingly and narrowly vote for Trump in 2016.  Trump won here for two reasons. The first is that rural support for the Democrat absolutely cratered. The second is that, although Clinton did about the same as Obama in the cities and outperformed him in the suburbs, her suburban advantage came nowhere near compensating for the loss of rural votes. Simply put, too many suburban voters just could not get past their historic opposition to the Clinton family.  Outward support for Trump in the Philly suburbs was hard to find from any quarter, but Republican support did not completely collapse.

Beginning in that election and then in the 2017 off-year elections and the 2018 midterms, the Democrats made the progress in the Philly suburbs that they’ll need to carry the state in 2020. But that support is newfound and tenuous. A Democratic candidate that is far to the left on economic issues will find a wall of resistance from people in the suburbs who traditionally voted Republican. Whether they left the party in 2008, 2012, 2016 or since seeing firsthand how Trump attempts to govern, they are not yet reliable Democratic votes. With Howard Schultz voicing their concerns about the deficit and the costs of Democratic proposals and tax-the-rich rhetoric, he will provide an alternative to Trump.

An unnamed Schultz adviser told Axios that their research “shows a centrist independent would draw evenly from the Republican and Democratic nominees, and bring Trump down to a ‘statistical floor of 26-27-28 percent.’” I don’t know if that would hold true nationally, but it rings true for places like the Philly suburbs. I suspect we’d see the same effect in key suburbs outside Detroit, Milwaukee, Atlanta, and Charlotte.

Historically speaking, it’s unnatural for the cities and suburbs to vote in concert, and for this reason the Democrats’ new coalition is unstable. Cities and suburbs are drawing close to each other now more for cultural reasons than economic ones, and as the Dems lose rural and small town votes, the urban/suburban coalition becomes critical. A third party candidate who is moderate-to-liberal on cultural issues and fairly fiscally conservative is going to appeal to people in the suburbs. But their main appeal is likely to be not with Democrats who will be desperate to be rid of Trump but with folks who are ready to ditch Trump but are not comfortable endorsing a Democrat, especially one that is far to their left on spending and taxes. These are basically Arlen Specter Democrats, and they’re reachable for the right kind of Democratic candidate. Many of them are not reachable for the kind of Democrat the party is likely to nominate in 2020.

One way of looking at this is that is will cost the Democratic nominee every time someone goes into the booth and casts a vote for someone else, but votes cast for a third party candidate instead of their Republican opponent is a net win for them. The question is whether the third party candidate will bleed them more than the Republican. To answer that question you really need to know how economically populist the Democratic nominee will be. I suspect they’ll set a new standard for economic populism in 2020. I don’t see how they could secure the nomination in any other way.

If the Democrats defy my expectations and nominate an economic moderate, then I’d agree that Schultz’s candidacy would hurt the Dems. More likely, Schultz would serve the same purpose Wallace served in 1968 and Anderson served in 1980. He’d serve as a way station for people who want a change but are too set in their ways to change parties.

In 2020, I see little prospect of any significant number of Democrats going with any third party candidate, but the ones most likely to do so are not to the left but in the middle. Schultz would prevent them from voting Republican and he’d attract Republicans disgusted with Trump but either constitutionally unwilling to support a Democrat or just alarmed at the threat to their economic self-interest.

So, count me as preliminarily unconcerned about the prospect that a Howard Schultz candidacy would reelect Trump. I think the opposite is more likely true.