Dispatches from the Creation Wars

Carpenter vs Blankenhorn, Round 3

The debate between Dale Carpenter and Michael Blankenhorn continues at the Volokh Conspiracy. As usual, Carpenter is simply pummeling Blankenhorn, who has now retreated from his already silly claims to endorsing the monumentally idiotic claims of Stanley Kurtz. It looks as though Carpenter’s polite praise of Blankenhorn’s work was exaggerated and that he is, in fact, little more than your standard issue gay marriage opponent using the same bad arguments we’ve heard a thousand times before.

But as usual, Carpenter’s reply is unfailingly logical and unassailable. I’ll post the primary argument that Blankenhorn makes, followed by Carpenter’s response. The basic argument is over correlation vs causation:

Just because we cannot scientifically demonstrate beyond any doubt that A is causally related to B, that does not mean that it is impermissable or a violation of good scholarship or good judgement to make reasonable (if qualified, and modest) inferences about a likely causal relationship between A and B! Apparently, according to Carpenter, at least when the topic is gay marriage, once we concede the truism that correlation does not demonstrate causation, we must fall into complete silence or profess only studied agnosticism about the entire issue of what seems to be causing what. But that is obviously rediculous. By that standard, no one could ever again say that anything is causally related to anything!

Want to see how this works? Here is a causal assertion: Cigarette smoking causes lung cancer. But wait a minute! Do all cigarette smokers get lung cancer? Is everyone who gets lung cancer a smoker? Of course not. So all we have is a correlation. There is no beyond-any-doubt proof of causation. Therefore, it is illegitimate for anyone to suggest that smoking causes lung cancer. See how easy it is? The tobacco industry made this exact argument for many decades, and some in the industry still do make this ludicrous claim.

It is ludicrous because our common sense observations in many societies over many decades, backed up by a great number of careful studies, have convinced almost everyone by now that the demonstrated correlations between smoking and lung cancer are not spurious or merely coincidental, but in fact are causal. So today, some tobacco lobbyist piously telling us (quite accurately!) that “correlation does not imply causation” does not stop us from concluding (quite reasonably!) that smoking lots of cigarettes increases the risk of lung cancer.

Carpenter absolutely shreds this argument:

What do we make of Blankenhorn’s use of correlations? I don’t think correlations are useless. They might indicate something important is going on. By itself, a correlation could be a starting point for further investigation. It’s a clue that two seemingly unrelated phenomena may be related. But it might also seriously mislead us unless we’re very careful.

Consider the case of smoking as a cause of cancer, which Blankenhorn uses to show that correlations can be valuable because they can help show causation. Yes, there’s a correlation between smoking and cancer. But we know smoking causes cancer not simply because of this simple correlation. Instead, we know smoking causes cancer because decades of careful, replicated, peer-reviewed, and methodologically sound medical research has revealed (1) a correlation (2) that sequentially matches the harm (e.g., lung cancer often follows smoking), (3) we’ve controlled for confounding variables and (4) ruled out multiple other plausible causes of the harm (e.g., auto exhaust or coal-fired plants), (5) and we’ve identified the agent or mechanism (over 70 chemicals in tobacco) that (6) causes a harmful result (tobacco carcinogens damage DNA inside lung cells).

When it comes to gay marriage “causing” harm by leading to non-traditional attitudes about marriage, Blankenhorn gives us only the first of these six. He has only correlation. And even this, it turns out, is suspect.

I’m not just playing with words here and I’m not requiring “scientific proof” analogous to demonstrating pathological processes in the body. I’m asking for a standard degree of reliability in inferences and an accounting when the correlations seem explicable by numerous other factors and are sequentially all wrong (more on that below). There’s good reason to be suspicious of an argument that a correlation allows us to infer a causal relationship. There’s a correlation between people who buy ashtrays and people who get lung cancer, but this hardly proves that buying ashtrays causes lung cancer. If we relied on correlation, we’d think all sorts of crazy things were causally related.

Consider what can be done with a correlation used to “infer” a “likely causal relation.” People in countries without same-sex marriage are more likely to believe women should stay at home and not work, that men should be masters of their households, that there should be no separation of church and state, that people should not use contraception when they have sex, that divorce should never be permitted, and that sodomy should be criminalized. If these correlations exist, have I demonstrated the existence of a “cluster of beliefs” that reinforce one another and “go together,” undermining the arguments against SSM?

Or consider the more sympathetic correlations to SSM that Blankenhorn ignores. Countries with SSM are richer, healthier, more democratic, more educated, more liberal, have more egalitarian attitudes about women, etc. Have I shown that the absence of SSM is likely causing harm in those unfortunate backward countries that refuse to recognize it?

Here’s another correlation helpful to the conservative case for SSM: countries with SSM are enjoying higher marriage rates since they recognized it. Have I shown that SSM likely caused this?

Even Blankenhorn’s correlation is suspect, in a way very similar to Kurtz’s. Non-traditional attitudes about marriage in countries with SSM preceded the recognition of SSM, just as signals of marital decline in Europe preceded SSM. Though I haven’t gone back and checked the previous international surveys from the 1980s and 1990s, I’ll bet my mulberry tree they show that. Besides, even the survey data Blankenhorn relies on show that he’s got a problem. In one survey, the data comes from 1999-2001, before any country had full SSM. In the other survey, the data comes from 2002, when only one country (the Netherlands) had full SSM.

How could SSM have caused a decline in traditional marital attitudes before it even existed? Of course, Blankenhorn is still free to argue that non-traditional attitudes greased the way for SSM, but this doesn’t show that SSM caused or even reinforced non-traditional attitudes. What Blankenhorn needs, even as a starting point, is some evidence that non-traditionalist views rose after SSM. He doesn’t have that.

Of course, even if he had the sequence right, he’d still have the problem of trying to deal with the existence of multiple other factors that have plausibly fueled non-traditionalist attitudes. Here, too, Blankenhorn has the same problem as Kurtz. Just as we can plausibly surmise that factors like increased income, longer life spans, more education, and women’s equality – rather than SSM – have caused actual marital decline, so we can plausibly surmise that factors like these have caused a rise in non-traditionalist attitudes about marriage. And even if the data showed a rise in non-traditional attitudes after SSM, that might well only be a continuation of pre-existing trends. Kurtz has that problem, too, when he tries to show marital decline.

If this was a real boxing match, it would have been a TKO in the first round. Carpenter is working Blankenhorn like a speed bag. This is Tyson-Berbick II.


  1. #1 Lynn
    April 9, 2007

    Ah, correlation! Dangerous ground.

    You know, the coefficient of correlation is close to 1.0 between the ability of children between 6 and 12 years of age to solve math problems and those children’s shoe sizes.

    This clearly shows that math ability is located in the feet, and that the bigger your feet are, the more room you have for the “right math stuff.” It would also explain why boys do better at math than girls, because everyone knows boys have bigger feet than girls do.

    Ah, correlation ;^)


  2. #2 James
    April 10, 2007

    The other thing in social research is that even sequence can be tricky.

    Example: Standard analysis might lead one to conclude that Christmas cards cause Christmas.

    Humans have this annoying habit of anticipating the future that makes the causal arrow of time somewhat more negotiable than in the physical sciences.

  3. #3 Alan Kellogg
    April 10, 2007

    It’s really very simply. Those places that allow SSms have an increase in marriages because now more couple can get married. They’ll also experience an increase in divorces because now more couples will have to divorce in order to separate.