Currently, there are around 30 Democrats vying for the 2020 presidential nomination, and Democratic voters are starting to think about their options. In a sense, the choice is easy: Pick someone based both on the issues and their ability to defeat Donald Trump. But that second quality—electability—is tough to determine.
The 2016 election challenged a lot of assumptions about what electability means. Trump was, according to the polling, an utterly unelectable candidate—until he beat Hillary Clinton. Now, following a 2016 election that saw white voters flock to Trump, some leading Democrats seem to have adopted a rather cautious election strategy: Just nominate a white man. While it would be amazing to elect a minority candidate, the stakes of the election are simply too high, so the thinking goes, to risk nominating a candidate who presents more of a long shot to the White House.
There's one problem with that line of logic: The evidence to suggest a non-white male candidate would be unelectable is, in fact, quite weak.
There is some research on the electability of African-American candidates, and the results are quite mixed. Some white voters are reticent to support an African-American candidate, but it can be hard to determine whether this is because of intrinsic prejudice or because they generally perceive African-American candidates to be more liberal than others. Meanwhile, a black candidate seems to boost turnout among African-American voters. It's not clear whether these two trends cancel each other out.
More generally, while there is evidence around African-American candidates running for Congress and state legislature, there's very little data at the presidential level. But here it's worth remembering that Barack Obama, an African-American man, won the popular vote nationwide less than seven years ago. Since the end of World War II, only three presidents have won a majority of the popular vote twice; one of them was black. To think an African-American candidate is unelectable nationwide is to ignore some very recent history.
What about a female candidate? The bulk of studies on this topic largely suggest that women perform about as strongly as men. There are some important differences, such as the fact that women are generally more hesitant than men to run (and tend to have more impressive credentials when they do run), but there's not much to suggest a woman would underperform based on her gender. And again, let's recall the most recent presidential election, where the female candidate received some three million more votes than her opponent, and actually over-performed expectations for a nominee hailing from the incumbent party (even if, yes, she failed to take the Electoral College).
As for LGBT candidates, acceptance of gay and lesbian candidates has increased substantially in recent decades. In fact, political science professor Don Haider-Markel's writes in his book, Out and Running, that "sexual orientation does not have a negative effect on electoral support, and in some cases may enhance electoral chances."
What the bulk of the research suggests is that partisanship and campaign fundamentals (the conditions of the economy, the popularity of the incumbent, conditions of war or peace, and so on) have far greater impact on the vote that any intrinsic demographic qualities of the candidates themselves.
Of course, equally important is how an individual candidate behaves once they're actually elected to office. Pretty much any Democratic nominee would broadly support the range of issues within the party's platform. There would, however, likely be great variance in which issues that person would actually prioritize.
Perhaps, then, the real question shouldn't be which candidate could defeat Trump, but rather which candidate would perform best in the White House.