A supporter of former Democratic presidential candidate Bernie Sanders and of Green Party candidate Jill Stein holds a sign at a rally at City Hall in Philadelphia on July 25 as the Democratic National Convention formally selects Hillary Clinton as the party’s presidential nominee. (Nicholas Kamm/AFP/Getty Images)

Jill Stein’s Green Party is aggressively courting Bernie Sanders supporters and others from the Democratic Party’s far left. Gary Johnson of the Libertarian Party seems to be doing the same, while presenting himself as an alternative to both Clinton and Donald Trump.

Meanwhile, on social media, we are seeing the 2016 iteration of that quadrennial debate between purity and pragmatism. Is it ethical to “vote one’s conscience” (i.e., rejecting the U.S.’s two main candidates and voting for a third party)? Or are those who support Stein or Johnson “wasting their votes” — or, worse, by withholding a vote from whomever they perceive as the lesser of two evils, effectively casting one for the candidate they most dislike?

But why must this be the debate? Surely, in some alternate universe, lesser U.S. parties could influence the election significantly. So let’s ask a different question: Do third-party candidates have a chance in 2016?

Political science says no.

Why? The answer lies in what is known among political scientists as Duverger’s Law.

The U.S. picks its elected officials in a way that creates a two-party system

Duverger’s Law says that the way a country’s electoral system is structured usually determines how many competitive parties that country will have.

Here’s how it works. First, when each district gets only one legislative seat (known as a single-member district, which we have in the United States) and, second, when the election’s winner takes that seat, then the system tends to have two dominant parties.

In such a system, all a party needs to win is more votes than the other side. That winner-takes-all nature of single-member districts encourages broad coalitions to form before elections. The odds of a party winning such elections are much higher if only two parties exist, enabling each side to work to bring as many people to its side as possible.

In the United States, that’s writ large in presidential races, because the Electoral College is itself a winner-take-all system: Within each state, the candidate who wins more votes takes all that state’s electoral votes. Even in Nebraska and Maine, where electoral votes are allocated by congressional districts, each individual district is winner-take-all.

So what’s the alternative? In many countries, each district gets many seats, and they are allocated in a way that proportionally matches the votes each party receives. In those systems, there will be many parties. That’s because a party has a much lower threshold for gaining a voice in government. In a proportional representation system, if a third party wins 10 percent of a district’s votes, it gets 10 percent of that district’s legislative seats. (That’s not true in a single-member district system; there, if a third party wins that same 10 percent, it gets nothing.)

As a result, parties don’t need to build broad coalitions before the election. They don’t need to beat everyone else; they just need to attract enough voters to be included. They build coalitions after they’ve gotten into government. Thus, parties in proportional representation systems can be much more specific in their platforms and can afford to appeal to a narrower range of voters. Without that possibility of getting into power on a narrow platform, political actors tend to consolidate their efforts until they end up in one or another of two dominant blocs.

Here’s the data that supports Duverger’s Law

In 1992, Republicans split, with some backing Ross Perot — and partly as a result, Democrat Bill Clinton won the presidency. In 2000 Democrats split, with some voting for Ralph Nader, and partly as a result, Republican George W. Bush won. Those notable U.S. elections reinforce an existing body of data that has tested this law repeatedly since it was first articulated in 1951.

There are a few exceptions. Canada, India and, more recently, Britain have winner-take-all, single-member-district systems — and yet they have many legislative parties. Why?

Canada has strong third parties at both the provincial and national level because the partisan makeup of the provinces tend to be more geographically concentrated and significantly different from each other.

Research on Britain shows that it has been accommodating third parties for the past decade, with the largest victory coming in 2010 when the Liberal Democrats won 57 seats. Yet the Labour and Conservative parties remain dominant and in the 2015 elections support for the Liberal Democrats dropped dramatically (resulting in only eight seats), while support for the main two parties grew, leading some observers to frame it as a “two-and-a-half” party system.

India, meanwhile, appears to defy the law completely both locally and nationally for unknown reasons.

But in the United States, the law holds.

The U.S. system is an outlier

Most of the world’s democracies have proportional representation. That’s why voters in places such as Uruguay, Israel or Italy can choose from a wide range of parties. In proportional representation systems, citizens vote for a party. The party then seats candidates from a list it had already drawn up. If the party wins 10 percent of the vote, then the 10 percent of the candidates highest on the list win seats. If it wins 15 percent, then the next 5 percent take seats in the legislature.

There’s another effect from the United States’ single-member district, winner-takes all structure: Candidates are elevated over parties. A person, not a party, wins a seat.

Some observers have called the U.S. approach “an accident of history,” commenting that, “Nothing in the Constitution sanctions a two-party system.”

That’s not entirely correct. The U.S. Constitution allocates House of Representative seats (Article 1, Section 2) by population. Under current practice, one representative is elected per district; each state gets two seats in the Senate (Article 1, Section 3). While this doesn’t explicitly mandate a two-party system, it does steer us in that direction. How congressional seats are allocated, combined with the winner-take-all elections by which those seats are assigned, keeps us in a two-party system, especially nationally.

So how could a U.S. third party succeed?

A critical mass of people would have to defect from one party to the new party, essentially causing the old party to crumble and the new (formerly third) party to take its place.

That’s what happened the last time a third party won the U.S. presidency in 1860, putting Abraham Lincoln in the seat. The insurgent Republican Party replaced its predecessor, the Whig Party, after the Whigs unraveled over slavery during the 1850s.

Encouraging more third parties would require revising our voting systems dramatically. One alternative is called Approval Voting. In this system, rather than voting for one and only one candidate, you can vote for as many candidates as you wish — for example, you could vote for Clinton, Stein and Johnson or for Trump and Johnson). The candidate with the most votes would still win the election. But that way, voting for an outsider candidate wouldn’t unintentionally reward the “opposition” candidate. Third parties could get on record as significant and could build competitive momentum.

Amanda C. Skuldt, PhD, is a political scientist and senior project manager for Wikistrat, an online global consulting network.

Note: An earlier version of this post mistakenly implied that the Constitution mandates single-member districts. It does not.