Want to know why Republicans sign Grover Norquist’s anti-tax pledge? Because it helps them win elections. Want to know why the pledge isn’t broken more often? Because breaking it makes them likelier to lose elections — even among voters who support tax increases.

Grover Norquist’s pledge works. Here’s how. (Joshua Roberts/Reuters)

The answer, in most cases, is the candidate who took the anti-tax pledge. Start with the respondents who wanted to balance the budget through spending cuts — which is to say, the respondents who are, themselves, against tax increases. For this group, taking the pledge was a slam dunk. If candidate A and candidate B both opposed taxes but only candidate A had taken the pledge, he got 74 percent of the vote. Call that the “Republican primary” scenario: If you’re running in a Republican primary and you refuse to take the pledge, you’re probably going to lose.

The important part is what happens next: Voters — no matter their opinion on taxes vs. spending cuts — hate politicians who break pledges. If a candidate took the pledge and then broke it, voters who wanted more spending cuts turned on him, but so did voters who wanted more tax hikes. They saw it, the authors hypothesize, as a character issue. You can see this on the table below, where I’ve highlighted the scenarios in which a “pledged” politician flipped.

But here’s the real issue: Even if that politician is facing an electorate that wants tax hikes, he’ll lose to a challenger who wants tax hikes and never broke the pledge, and he’ll lose some of his appeal against candidates who want only spending cuts. “Voters who disagree with a pledge are nevertheless willing to enforce it,” write the authors.

So it’s very hard for a politician to find a political upside in breaking the pledge: He loses too much among voters who hate taxes without gaining enough among voters who support taxes. The bottom line, say the authors, is “in the most likely general election scenario, pitting a pledged Republican against an unpledged Democrat, breaking the pledge would hurt the Republican’s electoral prospects unless nearly all voters (98%) wanted higher taxes.”

So that’s the trick: Norquist gets politicians to sign the pledge because it makes them more popular in primaries. And then he keeps them from breaking the pledge because breaking a promise makes them less popular with everybody. And you can see an example of today’s Politico.

Sen. Tom Coburn took Norquist’s pledge. And then he broke Norquist’s pledge. Now, Coburn is a conservative’s conservative, and his argument is that the debt is too significant a threat to spend time mucking about with pledges. But notice how Norquist attacks him.

In an interview, Norquist shot back, saying that Coburn had “lied” about his group’s position opposing ethanol tax breaks and that Coburn was trying to blame “someone other than the American people for their unwillingness to raise taxes...He took the pledge — not to me, to the voters.”

He’s not just saying Coburn now wants to raise taxes. He’s saying Coburn “lied.” That he can’t be trusted. That his reversal proved something about his character rather than just his provisions. That’s exactly the dynamic Tomz and van Houweling identify. And they fret that it could lead to “non-representative outcomes,” where “interest groups can use pledges to lock-in policies that the majority would not freely choose.”

But Tomz and van Houweling are presenting a world in which voters know nothing about the candidate save for the pledge. In the real world, voters know quite a lot about candidates — particularly incumbents — and so they base their votes on a multitude of factors. Someone like Coburn, who has a strong personal connection with his constituents, might be able to survive breaking the pledge. But weaker incumbents probably don’t want to risk it. And unknown candidates who run in primaries can’t resist it. And so the pledge endures.

The question this study leaves open is what happens if the entire Republican Party — or at least quite a lot of it — breaks the pledge all at once, as part of some kind of grand bargain.