Abortion restrictions for the fetus-indifferent

October 21, 2005 | 21 comments

A recent study concludes that mandatory waiting periods lead to fewer women killing themselves.

Another study by the same people suggests that parental notification laws reduce venereal disease among teens. This is ostensibly because teens behave like rational actors, but that can’t be right.

21 Responses to Abortion restrictions for the fetus-indifferent

  1. ed on October 21, 2005 at 7:32 pm


    Have you read these studies? Do you find them convincing?

  2. Seth Rogers on October 21, 2005 at 10:19 pm

    You’re right Adam. That can’t be right.

    The economists are going to be so disappointed.

  3. Boris Max on October 22, 2005 at 5:26 am

    Ed, you could also ask Adam if he’s carefully read the abstracts of these articles or looked closely at where they came from. For example, the article on abortion waiting periods is clearly labled as a “working paper” by the Social Sciences Research Network website. This strongly suggests that it hasn’t been peer reviewed yet, which makes its conclusions tenuous at best. So when Adam begins his post by announcing that “a recent study suggests that mandatory waiting periods lead to fewer women killing themselves,” he is participating in the overstatment common to popularizers of academic studies.

    Furthermore, the study itself seems to be guilty of the post hoc, ergo propter hoc logical fallacy. Let’s take a look at the abstract:

    Proponents of laws requiring a waiting period before a woman can receive an abortion argue that these cooling off periods protect against rash decisions on the part of women in the event of unplanned pregnancies. Opponents claim, at best, waiting periods have no effect on decision-making and, at worst, they subject women to additional mental anguish and stress. In this article, I examine these competing claims using adult female suicide rates at the state level as a proxy for mental health. Panel data analyses suggest that the adoption of mandatory waiting periods reduce suicide rates by about 10 percent, and this effect is statistically significant. The result is robust to various attempts to control for unobservable heterogeneity and simultaneity.

    So, if a state has a mandatory waiting period, there are fewer female suicides. Because the abstract does not specify suicide reduction among those with unwanted pregnancy–this is something a peer reviewer would catch–it makes it sound like all women in a given state are less likely to commit suicide. Really? The 85-year-old widow who is lonely and in financial difficulty would not do herself in because her state makes women wait to have an abortion? I find that difficult to believe.

  4. Mark B. on October 22, 2005 at 11:13 am

    Can anybody who writes a sentence like “The result is robust to various attempts to control for . . .” be taken seriously?

    The abstract, of course, is just a grim foretaste of the sociolgo-babble that fills the article itself. It’s not clear to me whether the author is simply incapable of writing clearly (which would suggest that he is similarly incapable of thinking clearly) or if he is intentionally shielding the weakness of his arguments behind the gobbledygook smokescreen.

  5. Frank McIntyre on October 22, 2005 at 5:04 pm


    Your logic works fine for lawyers, who write and think in words. But one can do quite well in economics with math as one’s thinking language. So I’ve known some very bright thinkers in econ who were not very good writers. Depressing but true.

    Ed and Boris,

    The methodology is a standard dif-in-dif. Which means that one uses changes in state laws and then assumes that states that do not change laws give you a counterfactual as to what the changing states would have looked like. The estimate is unaffected (is robust to) whatever covariates are thrown in, including the male suicide rate and the unemployment rate. Unobserved (but constant) differences across states are also accounted for. Standard errors are estimated allowing for unobserved autocorrelation. All in all, the method is used with all the appropriate bells and whistles, but still requires one to believe that the control and treatment design works (which is not that states are all the same, but rather that they trend the same, after accounting for trends in male suicide).

    Also Boris, the experiment is looking at suicide for adult women less than 65, but of course, the treatment should only affect those who abort. In which case the 10% number is right for the state, but would be much higher risk increase for those actually having an abortion.

    Do I find it convincing? If I had to guess, I’d say they probably found something, but no guarantees that the assumptions hold. Controlling for male suicide is perhaps a nice touch to account for unseen third causes.

  6. Frank McIntyre on October 22, 2005 at 5:06 pm

    By the way, I am just talking about the first paper, not the second. The first paper also had some evidence that eliminating public funding for abortions raised suicide, althout this was not as strong an effect nor as robust(?).

  7. Karl D. on October 22, 2005 at 6:53 pm


    The second paper also uses dif-in-dif, and standard errors that account for clustering at the state level. The empirical design is very similar to the first paper. For example, in the second paper they use the gonorrhea rate of adults to try to account for unseen causes.

  8. Julie in Austin on October 22, 2005 at 7:31 pm

    “Which means that one uses changes in state laws and then assumes that states that do not change laws give you a counterfactual as to what the changing states would have looked like.”

    I don’t know how anyone can approach this with a straight face. States don’t randomly end up with different laws; it should be obvious that a state that chooses to change a law has a different political/social/religious/whatever make-up from the state that doesn’t. In other words, the law is the result of an underlying difference that exists between the states, and it is not the law that causes the difference.

  9. Frank McIntyre on October 23, 2005 at 9:47 am


    Far be it from me to discourage skepticism, but the situation is not quite as bad as it appears. The method used accounts for unobserved (but constant) differences across states. It assumes that the difference between, for esample, Utah and Massachusetts is roughly constant over time, not that the two states are the same. One fact worth noting is that the results do not vary at all with whether or not puts in these controls for “unobserved differences”. This suggests that the two groups do not have some huge baggage of unobserved differences that are messing up the estimation.

    Also, anything one can observe is not going to be a problem because you can control for it. Thus religious differences are controlled for in the regression (although they don’t matter much). Also, any change in the female suicide rate _difference_ that also affects male suicide rates is accounted for, since those rates are controlled for. Regional differences (that are constant) are also accounted for, since each state is allowed its own unobserved set of causes. And lastly, violations of the assumptions only matter if the violations are correlated to the outcome you are interested in. So if Utah and Mass do not have a constant difference, but deviations from that assumption are unrelated to the law change, the estimation is still fine.

    None of this is perfect, so feel free to doubt away, as many economists do. The method in some cases is used wildly inappropriately. But the method is more plausible than simply a raw comparison across states or a “before and after” of one state.

  10. Julie in Austin on October 23, 2005 at 10:44 am

    Thanks for the info, Frank. You’ve taken me down from eye rolling to just scoffing, so I suppose that’s progress.

  11. Wilfried on October 23, 2005 at 11:57 am

    Interesting topic, but I had the same reaction as the skeptics, whether justified or not. In my own field (effects of variables in approaches to learn languages) we have been bombarded for decades with “a recent study suggests …” Then thorough analysis of the study reveals various flaws and neglected variables “suggesting” that actually nothing was proven. Meanwhile we have to live with “the overstatement common to popularizers of academic studies” (Boris).

    I also had the same reaction as Julie: “States don’t randomly end up with different laws; it should be obvious that a state that chooses to change a law has a different political/social/religious/whatever make-up from the state that doesn’t.” I applied the remark to the international scene and thought about the make-up of e.g. countries like the Netherlands & Belgium, with its “liberal” regulations in matters of e.g. abortion, but also with its developed protective and informative facilities. Independent of the “morality or not” of these regulations and facilities, I wondered to what extent such changes in law would have similar effects in those countries.

    By the same token, what would be the value of this recent study giving indications that “religious belief can cause damage to a society, contributing towards high murder rates, abortion, sexual promiscuity and suicide”? To quote: “In general, higher rates of belief in and worship of a creator correlate with higher rates of homicide, juvenile and early adult mortality, STD infection rates, teen pregnancy and abortion in the prosperous democracies… The United States is almost always the most dysfunctional of the developing democracies, sometimes spectacularly so.”

    I did not say so, the study did. It was published in the Journal of Religion and Society, a refereed U.S. academic journal.

  12. Frank McIntyre on October 24, 2005 at 11:46 am

    Wilfried and Julie,

    Your response is perfectly reasonable one– faced with a study the methodology of which you suspect is not sound but have no way to evaluate. Since I do this stuff for a living I feel more confident in evaluating the good from the bad.

    For example, Wilfried, your alternative study suffers from three severe problems not found in the other one. There is a severe reverse causality issue that is not so obvious a problem in the first, countries are far more varied in their unobservables than U.S. states, and the study probably uses a cross section of countries rather than following states over time and controlling for corss sectional differences. The abortion study controls for (and therefore is not affected by) all fixed differences across states in a way that makes the results at least possible. A cross sectional study can’t do this.

  13. Wilfried on October 24, 2005 at 12:05 pm

    Indeed, Frank. The weaknesses of the study I mentioned are obvious. Point is that it too will be cited as “a recent study shows …” And the ideas contained in that study will of course overjoy some and be misused… I must say I was suprised that the Journal of Religion and Society would publish such a piece, which is “tentative” but still suggests gross generalizations.

    Thanks for your helpful insights!

  14. ed on October 24, 2005 at 12:40 pm

    I, of course, agree with everything Frank said, but I think there is yet another reason to be skeptical of the findings.

    If you wanted to show that waiting period laws are a good idea, you could just look at a long list of social pathologies, and just by chance there would likely be one or two that appeared to be trending better in states that adopted parental consent laws. You could then publish your “statistically significant” findings as evidence that parental consent laws are a good thing.

    I don’t know if this is what happened here, but since I’ve never before heard of a strong connection between suicide and abortion laws, I’m suspicious. I’d expect a priori any such connection to be small and probably empirically undetectable. I’d be more convinced if a similar relationship could be found for several other measures of mental health, or if the connection could be made more explicitly to those who have considered abortions. (Disclaimer: I haven’t read the papers.)

  15. Frank McIntyre on October 24, 2005 at 1:03 pm


    I agree that that is a likely problem. Although the suicide rate is going to be higher for those that have abortions, it still seems a stretch that it would be so important to cause this big of a rise. And abortion impacts are certainly the sort of thing to suffer from opportunistic statistical inference. It is much more compelling when the causal chain is laid out explicitly and the links are obvious and confirmable.

  16. Adam Greenwood on October 25, 2005 at 11:41 am
  17. aaron on October 27, 2005 at 11:14 pm

    BBC is reporting on a counter study from the British Medical Journal to the first one Adam lists.

  18. aaron on October 27, 2005 at 11:17 pm

    Sorry, link didn’t work. Maybe just cut and paste.


  19. Jon Klick on November 1, 2005 at 4:27 pm

    Thanks for all the helpful comments. Regarding whether states randomly adopt waiting periods, one would think that they do not and the paper does investigate this a bit through instrumental variables techniques. Interestingly, the result appears to be larger when this self-selection is accounted for. As for why I focused on suicide, I started with a British Medical Journal article that had found (in a dataset of Finnish women) women who aborted were more likely to commit suicide than either those who miscarried or those who delivered their babies. Some pro-life groups picked up on this study, but it struck me as a case where self-selection may be important (as the BMJ authors suggest) as some of the same forces that induce some women to choose abortion may also be related to mental health. To avoid this possibility, the natural experiment (or differences-in-differences methodology seemed to make sense). Perhaps one could find a similar effect in other indicators of mental health, though data problems become more acute. I am currently examining the funding restrictions and suicide result more carefully, and I think there is some connection there too.

  20. Adam Greenwood on December 3, 2005 at 4:25 pm

    Thanks for your reply, Mr. Klick. We will follows your future research with much interest.

  21. El Jefe on December 3, 2005 at 9:49 pm

    Reading the jargon in the some of these studies reminds me of Winston Churchill’s comment: “What if I had said–instead of ‘We shall fight them on the beaches’–’Hostilities will be engaged with our adversary on the coastal perimeter.’”


Times and Seasons is a place to gather and discuss ideas of interest to faithful Latter-day Saints.