Ever since Barack Obama failed to win the New Hampshire primary (for reasons probably having little to do with people lying to pollsters), the media just cannot get enough of the Bradley effect. (For a cogent explanation of why the Bradley effect has been on the endangered species list since about 1991, and why it probably didn’t even happen to Tom Bradley himself, in the mayoral race of Los Angeles for 1982, see here.)

The Bradley effect is an attempt to measure the existence of hidden racism among the electorate. People are racists but cloak their views before a judgmental pollster, goes the theory. It’s worth pointing out that the phenomenon itself requires a special combination of circumstances. If you make a line chart of “racism in society over time,” where it starts out at 100% (everyone is always racist) and it slopes diagonally downward to 0% (nobody knows what racism is), the Bradley effect would only obtain when you have a bunch of racists but the racists aren’t really in charge of the discourse. In other words, too much racism in the society and nobody’s embarrassed about expressing it; too little racism and it doesn’t get expressed. You have to have a whole bunch of racists who are feeling a bit sheepish. In a way, it’s not surprising that the window for the Bradley effect is always a fleeting one.

My opinion is that for a subject of a poll misrepresent what candidate he or she supports to an anonymous pollster who possesses no power to alter the subject’s life … well, you have got to be talking about some serious shame/embarrassment. In other words, not wanting to vote for the black guy isn’t a potent enough cocktail of shame and embarrassment to induce the lie. You have to be supporting … pretty much a Klansman or a Nazi to elicit it.

At this point, I’d like to bring in two men, David Duke and Jörg Haider.

I hear a lot about the Bradley effect, but I rarely hear anyone mention David Duke. David Duke was a former Grand Wizard of the Ku Klux Klan who participated in a runoff election in the Republican primary for the gubernatorial race in Louisiana in 1991. A man named Edwin Edwards beat Duke pretty solidly, it turned out, but there were a few weeks there in which that outcome did not seem ensured, and in that period you heard a lot about white racists lying to pollsters. To be frank, it’s the last time (barring the possible exception of Harold Ford’s 2006 Senate run) that people talked about this subject at all in the United States. Here’s The New York Times, after Duke got beat:

Although the prospect of a large “hidden vote” for David Duke received a lot of speculation from poll takers and commentators in the weeks before Louisiana’s runoff election for governor, a hidden vote did not materialize in Edwin W. Edwards’s victory over Mr. Duke.

A hidden vote could have occurred if some voters were not willing to disclose their preferences to poll takers. In Mr. Duke’s past attempts at public office, his support was stronger than some polls had predicted, making some poll takers wary about simply using their standard methods in the runoff.

They seemed to take the effect pretty darn seriously, even if it didn’t manifest.

Between 1991 and 2008, you didn’t hear much about the Bradley effect in the United States. But you did hear about it (albeit not by that name) quite a bit in Austria, a country that featured the most successful radical right-wing politician in Europe: Jörg Haider.

By chance, Haider died in car crash a couple of weeks ago. In the 1990s, he led the Freedom Party of Austria to a series of very successful showings, finally entering a coalition government in 2000. Before Haider, the FPÖ was kind of a forgotten little right-wing party where the former Nazis would hang out; not a big deal. Haider changed all that, gradually building it up to nearly 30% of the vote and generally freaking a lot of liberals out, both inside and outside of Austria. Also, Haider would occasionally say flattering things about the Third Reich, which would get him into trouble.

And in Austria, you heard constantly about how polls were underrepresenting his support. I’m not an expert, but well-informed Austrians assure me that the electoral tallies tended to outstrip his support in polls.

In Austria, support for the Nazis is a crime punishable by incarceration; it’s a serious business, and the social sanctions against it are high—maybe not as high as here, but still very high. As with the KKK, perhaps, you don’t just casually admit to any anonymous caller that you are into supporting crypto-Nazis in Austria. Haider was deft enough a politician to blur his own Nazi ties (I myself think they were somewhat overstated)—but the whiff of social sanction was never far from him.

So that’s my thesis. if it’s just mild distaste for the black candidate, you’re not going to go and change the candidate you support to a pollster—that’s the threshold we’re discussing here. You might not admit the racism on the phone, but you’ll say you support the other guy because of his tax policies. Only for a candidate who is synonymous with evil are you going to cloak your views.

Personally, I think that the Duke and Haider cases constitute almost a death blow to the Bradley effect if you think through their implications; in one of the two cases, it didn’t even exist! The media want to keep interest in the race high, so they have incentives to dismiss countervailing examples like David Duke. But that doesn’t mean we should believe them.