Bryan Caplan  

Social Desirability Bias and Abortion

Canadian success... Richard McKenzie on Foreign Ex...
Economists have long argued that we should pay a lot more attention to what people do and a lot less attention to what people say.  But they make little effort to justify their pro-action/anti-talk position.  The strongest support for economists' methodological scruples actually comes from psychology, especially research on Social Desirability Bias.  Earlier today, though, I stumbled on some striking confirmation from a unexpected field: medicine. 

Choi, Riper, and Thoyre, "Decision Making Following a Prenatal Diagnosis of Down Syndrome: An Integrative Review" (Journal of Midwifery and Women's Health, 2012) meta-analyzes research on women's decision to abort a fetus with Down Syndrome.  As their decision becomes less and less hypothetical, women are more likely to opt for abortion.  In other words, Social Desirability Bias declines as the cost of doing what sounds good (accepting a mentally disabled child) rises:
The decision to undergo an induced abortion varied depending on whether participants were prospective parents recruited from the general population (23%-33% would terminate), pregnant women at increased risk for having a child with DS (46%-86% would terminate), or women who received a positive diagnosis of DS during the prenatal period (89%-97% terminated).
These are huge effects.  A supermajority of prospective parents says they would not terminate the pregnancy - but an even more lop-sided supermajority of parents who actually face this tragic choice takes the opposite stance. 

My question: Why do so many economists rely on half-baked philosophy of science to justify their pro-action/anti-talk position when fully-baked supportive empirical research is only a Google Scholar search away?

COMMENTS (21 to date)
david writes:

"talk is always cheap"/"talk is never cheap" is rarely the dichotomy being advanced

(whoops, wrong email address earlier)

Kevin Erdmann writes:

I suspect it's because speaking of it in the abstract is safe, but, by definition, to defend revealed preferences in these matters is to defend something that is very socially undesirable, which is reputationally dangerous.

Tracy W writes:

In related news, my physics teacher never bothered to point out fully-baked supportive empirical research on the existence of gravity or friction.

Jameson writes:

Let's do a different experiment. Ask people whether they would ever steal to eat. Then look at people who are actually hungry and have an opportunity to steal and possibly get away with it.

Is this an example of "social desirability bias"? Or is it just possible that doing the right thing can be hard?

Mack S writes:

Intuition. Anyone who's ever been a child under the power of adults (I think that's most of us) learns quite forcefully the cheapness of talk and the costliness of action.

Articulated or not, applied or ignored, this knowledge is part of the consciousness of every intelligent person.

The problem is it's socially undesirable to say so, because that would mean calling a lot of other people out on their cheap talk.

Thus does a very common bit of knowledge come to seem like the eccentric creed of a few econ professors.

RPLong writes:

Prof. Caplan, does this mean you categorically disagree with Situationism?

Ivan writes:

I think in this situation it may not necessarily be social desirability bias but shallow analysis on the part of prospective parents.

There is a HUGE difference between hypothetical scenarios for prospective parents and the hard reality for those who are at increased risk or received the DS diagnosis. Those in the later group have an enormous incentive to think through their choices, while those in the former probably gave it only a minute thought.

Lack of incentive for hard thinking could lead to various biases, including social desirability, but in a lot of cases it is just results in sloppy analysis.

Luke Muehlhauser writes:

Good find. Somewhat similarly, citizen support for IVF flipped as soon as IVF became actually feasible:

gwern writes:

[Comment removed for policy violation. Email the to request restoring your comment privileges. A valid email address is required to post comments on EconLog and EconTalk.--Econlib Ed.]

Stan Greer writes:

Has it not occurred to Dr. Caplan or to any of his commenters that prospective mothers and fathers who oppose aborting their unborn children as a result of a DS diagnosis won't have their children tested in the first place?

If that is largely true, then the results to which Caplan points are basically meaningless, and his whole post is hash.

awp writes:

Exposing social desirability bias is against our social desirability bias.

awp writes:


Population one and two have not tested. Both samples are from general populations. For your complaint to be substantive you would need to posit a positive correlation between beliefs on abortion and risk for down's uncorrelated to actual exposure to down's.

Going from one and two to three on the otherhand, I do agree with your complaint. If I know I am not going to abort is there any other reason for me to get this test?

But, the difference between one populations one and two is still pretty significant. Which I believe makes this post not hash.

(Not That) Bill O'Reilly writes:


It's not immediately apparent to me why those "who oppose aborting their unborn children as a result of a DS diagnosis won't have their children tested in the first place." There are many perfectly good reasons to test for developmental abnormalities while the fetus is in utero, if only to be able to start taking proactive preparatory measures to ensure that you will be able to effectively tend to the child's needs.

Stan Greer writes:

I haven't read the study, but I'm not at all sure from Caplan's description of it that group number two is taken from a random sampling of all pregnant women, as opposed to women who chose to be tested after they were pregnant to find out if they are at high risk of having a DS baby.

And my experience is that many pro-life women don't think medical science currently has much to offer them, while they are pregnant, to address the problems of a DS child after birth, other than to give them the opportunity to have an abortion. That's why, e.g., my wife was not tested for this while she was pregnant with any of our six children, including one who was born when she was 42.

So I think my point stands.

Michael S writes:

agree with Stan above — possibly large selection bias. We did no pre-natal testing with any of our three children, nor will we with future progeny. What would be the point unless you'd at least consider killing your child based on an undesirable test outcome?

Tom West writes:

One thing this ignores is the importance of "talk", especially when the "do"-ing is relatively rare.

In many of these cases, the overall social consequences of talk may be far more important in day-to-day lives than the more severe, but seldom realized consequences of "do".

In this case, I think society and those with Down's syndrome are far better off with the "talk" making clear the value of all members to society, even if it's not matched by our actual actions in a crisis.

Such "talk" does, however, make a small difference in an unmeasured way each and every other day.

Economists have long argued that we should pay a lot more attention to what people do and a lot less attention to what people say.

And they wonder why people dread economists giving policy advice... The unmeasured matter to most people's life at least as much as the measured, but economists feel free to ignore it, and yet expect to be taken *more* seriously than the man on the street who understands this implicitly.

Ed Hanson writes:

It would seem to me a better survey of actual DS abortion rate could be derived from a statistical measure of Down's babies born as a ratio of all live births before prenatal testing was available compared to those today with such testing. It should be a matter of the change in ratio to get a reasonable answer.


Tracy W writes:

Tom West:
Firstly there's a big difference between saying "we should pay less attention to xx" and saying "we should ignore xx."

Secondly, it's crazy to accuse economists of ignoring the unmeasured. Economics is built around the unmeasured. Do you think that no economist talks about supply/demand curves until they've actually varied the price and measured the supply and demand curves? Do you think economists have comprehensive theories about why people have the preferences they do? Do you think economists have solved the problem of inter-personal comparisons of utility?

Perhaps you can find some unreformed Marxist economist somewhere who is still determinedly clinging to the labour theory of value, but every economist I can think of is firmly in the "unmeasured things can be darn important" camp.

I'm sure that plenty of people that would say they would never help the Nazis kill Jews, but would do so if they were in the situation of German soldiers.

Michael Crone writes:

Thanks to both Stan Greer and Jameson for pointing out major flaws in Bryan's conclusions here, before I even noticed the post.

Ed Hanson, what you propose would be good as balance, but it has the opposite bias. Mothers who had no idea of their child's condition would be counted as mothers who chose not to abort.

I think Jameson, "Or is it just possible that doing the right thing can be hard?", is on to a bigger problem with the whole idea. Even if an unbiased study showed the same disconnect, it doesn't argue "pro-action/anti-talk", it argues that talk and action don't agree.

In this case, I think the most plausible answer would be "anti-action/pro-talk" in the sense that the true preferences of people are better revealed in the survey, but that decisions made under fear and stress can fall short. Most people believe this already. (And this should humble those of us who haven't been in that situation, but that's beside the point.)

Furthermore, I see a tone of regret in enough stories from women who have had abortions that I have met and those that have been recorded, even by Exhale, which leans pro-choice. (There is also Silent No More, an explicitly pro-life organization that records abortion stories.)

I know that, in theory, expressed regret could be just another example of selection bias or Social Desirability Bias, but I recommend that anyone making that argument read the stories or explore on their own first. I find that too many of them are too raw for that to be the case, and I haven't seen anything to refute this more systematically.

Ed Hanson writes:

Michael Crone

It has been thirty to twenty years since our last children's births and we were offered the test. I find it hard to believe that the test is not universally offered when indicated as well to all on request. I would expect any bias would be much smaller than other errors that occur and, most importantly, much smaller than the referenced survey.


Comments for this entry have been closed
Return to top