Many Lie About Their Support Of Gay Marriage: Study

Thanks to New York Times’s (yes) Russ Douthat who alerted us to the paper “Findings from a Decade of Polling on Ballot Measures Regarding the Legal Status of Same­ Sex Couples” by Patrick Egan (NYU).

Statisticians have long known that people when answering political questions lie like a rug, like their pants are on fire, like they are in a mighty hurry to be somewhere else. Not always of course, but especially over “controversial” topics. For example, I am a (part time) academic, a milieu where it is customary and expected to voice support “for” progressive causes. I cannot recall a single soul among my local associates who said at the time they were going to vote for George Bush.

I often use the tale that when Bush was battling Kerry, New York City polls had Kerry besting Bush by a multiple of three to five. The actually result was far smaller. You don’t tell the truth because you don’t know who could be listening. Just like on campus you musn’t let it be discovered that you are against gay “marriage” or that you are a climate skeptic (as to that, see “part time” above).

So when the pollster calls, people lie. The real question is: how much? Egan thinks he has an answer for gay “marriage.” This is:

1. “The share of voters in pre‐election surveys saying they will vote to ban same‐sex marriage is typically seven percentage points lower than the actual vote on election day.”
2. “survey estimates of the proportion of voters intending to vote against same‐sex marriage bans tend to be relatively accurate predictors of the ultimate share of ‘no’ votes.”

I find Egan’s wording confusing (he changes for and against in the sentences), so I’ve re-written his conclusion:

1. Votes to ban same-sex marriage are on average seven percentage points higher than polls indicated. So that if polls found (say) 45% will vote to ban SSM the actual vote will be 52% (on average).
2. Votes for SSM, i.e. votes to ban the SSM bans, match poll estimates on average. So that if polls found 55% against an SSM ban the actual vote will be 55% (on average).

These numbers aren’t far off actual polls and votes. Problem is, they don’t add up, and won’t unless in real cases there are large numbers of undecideds. So is must be that there is lying on both sides, with more coming from those who say they favor SSM. Egan says there is no “immediate evidence” in his data that people are lying to pollsters. But there’s plenty of experiential evidence. Certainly the scenarios I mentioned above are well known. And you yourself will know if you dare to voice opposition to SSM.

This is Egan’s Fig. 2. Each dot is a separate poll, taken over various states. This seems to me pretty good immediate evidence that many people, if they weren’t lying to pollsters, underwent an Obama-like evolution once they stepped behind the curtain. Or it could be that people all told the truth (in a way) but that supporters of SSM much more often stayed home on election day.

Egan has another intriguing result (his Fig. 3). Each of various states had the percent of gay and lesbian population estimated. Surely that is fraught with error, but never mind that. He then plots the average gap between poll-projected support and the actual vote to ban SSM. Regardless of the gays and lesbian estimate, this gap averages about 4%.

This, and a similar result found for automated versus human-contact polls, is the evidence Egan uses to say that people don’t lie to pollsters because of the subject matter. But I don’t buy it. Who trusts the computer which calls your house? Who trusts a pollster? Many people just don’t like being put on the record. Right, Mr Obama?

I like this kind of research and hope we can see many more papers who examine the outcome of actual elections versus polls. This will allow us to put real, not abstract mathematical, plus or minus bounds when giving out a poll result. When you hear a poll if you listen carefully you catch something like, “The margin of error is plus of minus four percent.” But that number is a theoretical calculation based on at least the assumption that everybody is telling the truth.

Because people lie, we need real margins of error discovered from real data. This would be an excellent masters of dissertation topic.

Also see this article from the Washington Post; via HotAir.