Questioned by debate moderators on Thursday night about his loyalty to the Democratic Party, Bernie Sanders offered his go-to rejoinder.

"[I]n virtually all of the general election match-up polls between [Donald] Trump and Secretary [Hillary] Clinton, and Trump and Bernie Sanders, in almost all of those polls, I do better than Secretary Clinton," he said. "You know why? Because in fact a whole lot of people — this may be a shock to the secretary, but there [are] a whole lot of independents in this country. ... [W]e are not going to win the White House based on just long-term Democratic votes."

That latter contention is literally true but figuratively questionable. If a Democrat receives only Democratic votes, he or she will indeed lose. But in two of the last four presidential elections, the Democrat has lost the independent vote, according to exit polls, including a five-point loss among independents by Barack Obama in 2012.

But let's set that aside. Instead, let's look at the argument that Sanders has made many times before: that he's the more electable candidate in November.

A few weeks ago, we looked at the long-term trends of each of the five remaining candidates against one another. Sanders is correct that general election polling routinely shows him outperforming Clinton. On the Republican side, interestingly, the best candidate isn't Trump, who, by Sanders's logic, might be expected to be the top contender given that he, too, is the outsider candidate. Instead, the top-performing Republican is John Kasich.

General election polling at this point doesn't tell us much about the actual outcome in November, but it can tell us something about the dynamics of the race and the interplay between the candidates.

To assess the question of why Kasich and Sanders fare the best, we grabbed data from a Quinnipiac University poll conducted a few weeks ago. This poll had two things we were looking for: Head-to-head match-ups between all of the candidates and favorability ratings for all of them. So here's a complicated graph with the data from that poll that I will then explain.

Here's what we're looking at. Top row is all voters; bottom row is independents. The three graphs compare, in order:

  • The margin of support between the candidates vs. the difference in net favorability between them,
  • The vote margin vs. the difference in how well known the candidates are, and
  • The difference in how well known the candidates are vs. the difference in net favorability.

Positive values on each are higher values for the Democrat — including higher values of being unknown.

The important thing to notice is that the six marks on each graph generally travel from the lower left to upper right. The figure in the bottom-right corner of each, the r-squared value, shows how strongly correlated the two things we're looking at happen to be. An r-squared closer to 1 means a strong correlation.

So the first thing we notice on the graph at the upper left is that there is a very strong correlation between the margin of support between two candidates and the favorability between them. The less a candidate is liked relative to his or her opponent, the less well that candidate fares in a head-to-head match-up.

In other words, this graph suggests that there's a link between Clinton being less well-liked than Sanders and her doing worse against the Republicans than him. That's an important correlation. It also suggests why Kasich does better than his opponents. His overall net favorability is plus-23, vs. Trump's minus-28.

The same pattern holds for independent voters separately. They like Sanders and Kasich better, too.

But notice that one box in the lower-left graph that's outlined with a dashed line. That's Sanders vs. Kasich, and in the head-to-head match-up in this poll, Kasich does better. And this is with independents.

Which brings us to the other two correlations. The third graph in each row shows that there's a correlation between how unknown a candidate is relative to his or her opponent and how favorably he or she is viewed. And the middle graph shows that there is a correlation between how unknown a candidate is vs. his or her opponent and how much better he or she does in polling.

Kasich is less well-known than everyone — and does better than them in polling.

This makes sense. If I say to you, "Who would you rather vote for, someone you hate or someone you've never heard of," who would you pick? A lot of people would pick the person they'd never heard of. The defining characteristic of the remaining candidates is that the front-runners are deeply unpopular. Pitting a guy you think maybe you've sort of heard of up against them means those other people will probably do alright.

This doesn't mean that as Sanders (still unknown by 12 percent of the electorate) gets better known he will necessarily do worse in polling. For one thing, correlation isn't causation. For another, he likely won't reach Clintonian levels of unfavorability anytime soon. Instead, it suggests that the link between Sanders's general election polling success and his status as an independent isn't as clear-cut as he portrays it.