Feb 19 2014

Bad Science: No, Atheism Does Not Cause Suicide

More innumeracy for today: this religious apologist is claiming atheism causes suicide, and he cites a study that supposedly proves this, but both s/he and the study’s authors suck at numeracy and basic logic. I warned about this before (Innumeracy: A Fault to Fix). This is another example of that.

Just excerpting from the study citation and abstract as reported by this author:

METHOD: Depressed inpatients (N=371) who reported belonging to one specific religion or described themselves as having no religious affiliation were compared in terms of their demographic and clinical characteristics.

RESULTS: Religiously unaffiliated subjects had significantly more lifetime suicide attempts and more first-degree relatives who committed suicide than subjects who endorsed a religious affiliation. Unaffiliated subjects were [also] younger, less often married, less often had children, and had less contact with family members. Furthermore, subjects with no religious affiliation perceived fewer reasons for living, particularly fewer moral objections to suicide. In terms of clinical characteristics, religiously unaffiliated subjects [also] had more lifetime impulsivity, aggression, and past substance use disorder. No differences in the level of subjective and objective depression, hopelessness, or stressful life events were found.

CONCLUSIONS: Religious affiliation is associated with less suicidal behavior in depressed inpatients. After other factors were controlled, it was found that greater moral objections to suicide and lower aggression level in religiously affiliated subjects may function as protective factors against suicide attempts. Further study about the influence of religious affiliation on aggressive behavior and how moral objections can reduce the probability of acting on suicidal thoughts may offer new therapeutic strategies in suicide prevention.

From: Kanita Dervic M.D. et al., “Religious Affiliation and Suicide Attempt,” American Journal of Psychiatry 2004.

There are a number of things wrong with both these scientists’ stated conclusions (and study design) and this religious apologist’s use of it to argue atheism causes suicide. I’ll just focus on a few:

(1) The study did not even count atheists. The only nonreligious category was “religiously unaffiliated,” which will include many believers who just don’t affiliate with a sect or church. It is amusing to see the religious author completely overlooking this, and the fact that it completely destroys the logic of his/her argument.

(2) The study’s confounding variables (as even the abstract reports, the “unaffiliated” were very different demographically) suggests the same defect found in other similar studies: they fail to distinguish between having a religion and the mere fact of having a social network (and identifying with any worldview, religious or not). I’ve discussed this before (Atheism Doesn’t Suck: How Science Does Not Prove Atheists Are Less Happy, Healthy, and Sane). This is bad study design. Perniciously bad, in fact, since it leads the study’s author to make a poor recommendation for treatment. This is just like ancient witch doctors concluding shaking a wand at someone makes them feel better, therefore wands should be used more in treatment. In actual fact, just the attention and human company and the belief that they should feel better when wands are shaken at them (i.e. the placebo effect) is what is producing the effect. The wand is irrelevant–and can safely be discarded, for something less expensive and less superstitious (see my analysis in Sense and Goodness without God, IV.2.2.4, pp. 270-72).

Even worse than needlessly costly wands, this analogy calls attention to this study’s authors’ failure to weigh the costs of religious affiliation before recommending it (a fault that would nix this from any peer reviewed journal had they been similarly recommending a drug that likewise they were completely disregarding the negative side-effects of). Those range from adopting a system of false beliefs and the defective epistemology needed to sustain them, to the harm the resulting superstitions will have on the community, society, and polity as a whole (e.g. endorsing sexism, suppressing the rights of homosexuals), as well as to the individual, who will often in result be adopting self-damaging false beliefs about morality (e.g. that their doubting and asking questions makes them evil, or that homosexuality is evil and therefore if they are gay something is wrong with them).

Several psychologists have explored these and many other damaging costs of religious affiliation, which likely far outweigh, for both the individual and society, any of the benefits, e.g. Marlene Winell (Leaving the Fold: A Guide for Former Fundamentalists and Others Leaving their Religion), Valerie Tarico (The Dark Side: How Evangelical Teachings Corrupt Love and Truth), and Billy Wheaton (Hooks and Ladders: A Journey on a Bridge to Nowhere with American Evangelical Christians). See also the findings of journalist Janet Heimlich (Breaking Their Will: Shedding Light on Religious Child Maltreatment), historian Hector Avalos (Fighting Words: The Origins of Religious Violence), and my remarks (and cited scholarship) in The End of Christianity (pp. 338-39, with notes on p. 422). And lest one think one can solve this problem by recommending liberal rather than conservative religious affiliation, see also What’s the Harm? and chapters 4 through 7 of Greta Christina’s Why Are You Atheists So Angry.

(3) Another study (see my discussion here) shows fewer atheists get depressed. So even if this study counted the rate of suicide among depressed atheists (and it didn’t), that would not accurately reflect the effect atheism has on suicide overall. Because this study only looked at depressed patients (and patients with other mental illnesses known to be associated with suicide). It did not compare atheists with theists in the general population. It is instead already looking at a known cause of suicide: mental illness. That is the principal causal factor. Not religious belief (or lack thereof).

(4) That other study also found differences in extremity of depression for extremity of religiosity: extremely religious people experience more and worse depression than both atheists and less religious people do. Yet this current study makes no distinction for extremity of religious belief. If suicide rate correlates with extremity of depression (and that sure sounds likely), then since the study this apologist is citing did not tease out those two populations (extremely religious from less religious), it may have concealed a higher rate of suicide in the extremely religious. Which makes the study author’s recommendations even more pernicious: as it does not distinguish the greater increased risk of suicide resulting from encouraging (or leaving untreated) extreme religiosity, and thus essentially recommends a dangerous drug without concern for dosage. Just as it also did not consider “wand-free” alternatives to this drug altogether (e.g. a stronger attachment to atheist community, identity, and worldview).

(5) This study actually did not even test religious affiliation’s effect on suicide. Instead, it only found a correlation between “moral objections to suicide” and reduced suicide risk, and then simply inferred that because “moral objections to suicide” correlate with religious affiliation, that religious affiliation is the causal factor, when in fact any worldview (including a nonreligious worldview) that adduces “moral objections to suicide” would have the exact same effect. In their words:

First, religious affiliation was significantly associated with moral objections to suicide. Second, moral objections to suicide was significantly associated with suicide attempt when religious affiliation was statistically controlled. Third, the significant bivariate association between religious affiliation and suicide attempt did not remain significant when moral objections to suicide were controlled statistically. [Likewise] … low moral objections to suicide, and younger age were significantly and independently associated with suicidal ideation. Religious affiliation and responsibility to family were not.

So, religious affiliation actually was shown to have no effect. Only “moral objections to suicide” had an effect. So why, then, does the study recommend increased religious affiliation? No reason is stated. This is why you don’t just read the abstract. Abstracts often simply lie about what the study described actually found.

The take-away for atheists, who generally tend to support moral suicide (euthanasia to escape destruction of the mind or intolerable incurable pain, for example, or self-sacrifice, e.g. jumping on a grenade to save someone else’s life), is that we should consider more openly engaging in philosophical demarcation between moral and immoral suicide (or if it is easier to conceptualize, call it prudent and imprudent suicide).

This is self-evident, but the self-evident is often overlooked: if you are suffering from a mental illness (like depression), your reasons for considering suicide may be irrational (and you may not realize that), and you should have a strong moral objection to taking a life for irrational reasons (including your own); conversely, if you are considering suicide, this may be because you are depressed, and not for any rationally (i.e. morally) acceptable reason on any atheist worldview, consequently you have a strong moral obligation to find out (i.e. see a therapist to determine if you diagnose as depressed, or bipolar, or any other mental illness correlated with irrational suicide). A strong atheistic reason-and-evidence-based worldview is therefore a viable (and much less dangerous) vehicle for producing the effect this study claims to have found for religious affiliation (but only actually found for “having moral objections to suicide”).

(6) Finally, to revisit a point I have made before: as a rule, always look at the effect size and compare it to the margin of error. This study in fact found a correlation between suicidal acts and thoughts and “moral objections to suicide” of just 0.89. Meaning barely a 10% difference (which this study in effect claims is the benefit of religious affiliation for preventing suicide). The error size was around 3 to 4 points, so in fact the effect could have been as low as just 7% (at a p value of .001). That’s weak tea. The number of suicides thus prevented by having “moral objections to suicide” is pretty small.

Sometimes when a study claims to find a significant difference, it’s barely significant at all, and this is one of those times. And when you see that, you have to ask why it is so small. For example, if atheism causes suicide just 7% more often (after subtracting the margin of error from the effect size), one has to ask why its effect is so small. Often that would indicate a confounding factor other than atheism is responsible. Like, oh, let’s say, people driven to atheism by depression (and not reason and evidence). That is already obviously likely, and yet a higher rate of suicide among depressed atheists in that case is entirely to be expected yet has nothing to do with the effect of atheism.

In other words, depression causes people to lose faith in things (like religion), so depressed people will exhibit nonbelief more commonly than the general population, but not for valid reasons (they did not arrive at an atheistic worldview philosophically, much less rationally or empirically), and when someone is so depressed they even lose faith in their God, this is already likely to correlate with a higher risk of suicide, but not because of losing faith in God, but because their depression was that bad.

This means recommending to someone who loses faith in God because they are depressed that they should have more faith in God is stupid. It’s exactly like recommending to someone who is sneezing because they have the flu that they should sneeze less. The actual problem (which is actually causing their risk of suicide) is causing their reduced faith in God, so the only way to treat the faith in God thing is to treat the depression. But that means you should just treat the depression and forget this unrelated faith-in-God business, which is just a symptom of the depression, not its cause. So once again, this study’s conclusions and recommendations are lost in the logic of witch doctors, rather than sound reasoning.

Of course, actually, we aren’t even talking about faith in God (which this study did not measure), but in organized religion, because that is in fact all this study claimed to measure, and failed to even then–unless one pretends that didn’t happen and assumes “religiously unaffiliated” means atheist–even though it doesn’t–and that finding no correlation with “religious affiliation” counts as finding a correlation with religious affiliation, by assuming only religious affiliation produces “moral objections to suicide,” even though that assumption is false. Yeah. Talk about irrational.

This study is pretty useless. Badly designed, with conclusions, recommendations, and even its abstract hosed in almost every way. And the religious apologist’s use of it only betrays their inability to notice any of this. And wasn’t valid anyway.

Logic and numeracy. Learn it. Live it.


Skip to comment form

  1. 1
    Alison Parker

    “CONCLUSIONS: Religious affiliation is associated with less suicidal behavior in depressed inpatients. After other factors were controlled, it was found that greater moral objections to suicide and lower aggression level in religiously affiliated subjects may function as protective factors against suicide attempts. Further study about the influence of religious affiliation on aggressive behavior and how moral objections can reduce the probability of acting on suicidal thoughts may offer new therapeutic strategies in suicide prevention.”

    I’m a little confused, where in the conclusion does it claim that atheism cause suicide?

    1. 1.1
      Richard Carrier

      That claim was made by the apologist citing that article. I never said the article said that. Read the first paragraph of the post you are commenting on. You seem to have glazed over what I was writing about.

  2. 2

    Thanks for writing such a thoughtful post on this article. A couple of years ago at my Jesuit affiliated medical school, I had an ethics course which included this American Journal of Psychiatry article, much to my dismay. I got halfway through it, decided it was horse-shit, and tossed it aside. I wish I had read this post prior to our journal club, as I would have been better able to articulate its uselessness.

  3. 3
    Phillip Hallam-Baker

    We do however have a demonstrated case of ‘suicide by snake handling’ after that Kentucky pastor killed himself with a copperhead. OK so he wasn’t actually trying to kill himself but he wasn’t doing much to avoid it either.

    And the kicker is that the bit of Mark that he was following is an interpolation someone tacked on to the end of Mark.

    Though why would someone write such a piece unless waving snakes about and drinking poison etc weren’t part of the schtick for at least some in the early church? Why would people believe the claim of meeting Christ on earth if not backed up by at least some evidence? In an era when some priests were cutting off their genitals, snake waving could be seen a positively mild.

  4. 4
    David Olsen

    Yeah, it’s even more weak when you consider the combination of generalized views of religious affiliation with depression. The study pretty much says “withdrawn people are more likely to be depressed.”

    Apologists go: Let’s call them atheists, screw atheists.

  5. 5
    kelly stephens

    As a person who cleans-up suicide events, I can tell you we see family members/victims of suicide from all religions. Any study can be biased based upon the outcome they want to convey with the findings.

  6. 6
    No Way

    This might not be obvious to someone who grew up without religious indoctrination, but a likely cause of higher suicide rates among atheists, assuming the study has any validity at all, is the conflict between the atheist and their entire emotional support structure. People who grew up in religious families have a tremendous amount of pressure to say they believe in god. I would imagine that the stress is even greater with someone who had faith and then loses it in this environment.

    1. 6.1
      Richard Carrier

      That’s a good point to add. (It reminds me of stats similarly skewed for homosexuality: in low-support/hostile communities, rise in suicide rate correlates with homosexuality, but one cannot claim homosexuality causes suicide, that would be a perverse elision of the actual cause.)

  7. 7

    A good rule-of-thumb for odds ratios is that you want them to be not just statistically significant, but large (I’ve seen recommendations of >2/<0.5). So the idea that "age (odds ratio=0.97, 95% confidence interval [CI]=0.95 to 0.99; Wald χ2=7.84, p=0.005)…and responsibility to family (odds ratio=0.93, 95% CI=0.91 to 0.97; Wald χ2=17.99, p<0.001) were significantly associated with suicide attempt status" is ridiculous.

    I'm willing to bet that they treated age as a continuous variable as well; rookie mistake.

    They performed so many statistical tests on 300 observations, I wonder if they did a proper post-hoc.

  8. 8
    Carlos Cabanita

    Some time ago I was severely depressed and beginning to atempt suicide and I know it is a very strange singularity point where rationality is very difficult. The depression, not a rational force, battles the survival instinct, not a rational force either. I survived seeking external help, which is a form of emergency rational rescue. But amidst all my turmoil, I of corse had my moral objections to suicide, in my ethical atheist worldview. My main objection to suicide was this: should I kill my daughter’s father? My mother was 94 and delusional, but should I kill her son? Should I kill my grandchildren’s grandfather? Or my friends’ friend?
    Suicide may be valid under other extreme circunstances, but I’m happy thay my survival instinct and my moral objections to suicide kept me alive for enough time to seek help.

    1. 8.1
      Richard Carrier

      Thank you for that comment. It means a lot, and is good to hear about.

  9. 9
    Geoff Barrett

    The author refers to a directional cause and effect relationship between variables that is not supported in the analysis:

    “Therefore, religious affiliation may affect suicidal behavior both by lowering aggression levels and independently through moral objections to suicide.”

    That there is a correlation between religious affiliation and lower aggression levels does not necessarily mean that religious affiliation causes lower aggression levels. It could mean that people with lower aggression levels are more apt to affiliate with a community of like-minded believers. And, as you pointed out, “moral objections to suicide” and “religious affiliation” could be controlled by social networks.

    Here’s another interesting train of logic:

    “Responsibility to family was inversely related to acting on suicidal thoughts. Most religions stress the importance and value of family. Thus, consistent with previous reports (5, 6, 30), a commitment to a set of personal religious beliefs appears to be a more important factor against suicidal behavior than social cohesiveness per se.”

  10. 10

    Richard wrote:

    [R]eligious affiliation actually was shown to have no effect. Only “moral objections to suicide” had an effect.

    The authors claim that they hypothesized that religious affiliation would reduce suicide risk by increasing moral objections to suicide, and they claim that the results support their hypothesis:

    Moral objections to suicide mediated the association between religious affiliation and suicide attempt as all three stipulated conditions were met (28). First, religious affiliation was significantly associated with moral objections to suicide. Second, moral objections to suicide was significantly associated with suicide attempt when religious affiliation was statistically controlled. Third, the significant bivariate association between religious affiliation and suicide attempt did not remain significant when moral objections to suicide were controlled statistically (F1).

    They are correct in a sense. If their hypothesis is correct, and furthermore, if moral objections to suicide is the only mediator of the effect of religion on risk of suicide, then they should get the results that they did; and, I suppose, that if they really formed their hypothesis before looking at the data, then their results actually support their hypothesis. Support, but not prove, since I doubt that they have a firm enough theory to justify such a strong assumption. Additionally, social scientists often practice HARKing—Hypothesizing After the Results are Known—in which case their “hypothesis” could simply be a post hoc rationalization of unexpected results.

    1. 10.1
      Richard Carrier

      You do realize that would almost be a tautology, right?

      “People who have reasons not to commit suicide, commit suicide less.”


      Can you imagine a scientific study proving “People who have reasons not to buy hamburgers buy hamburgers less”?

      And then acting like you’ve discovered something revolutionary?

    2. 10.2

      Richard, is the above comment actually a response to my comment? It doesn’t seem to have anything to do with what I wrote. I said nothing about reasons not to commit suicide. My post is about the authors’ claim that their results show that religion decreases risk of suicide by increasing moral objections to suicide.

    3. Richard Carrier

      Yes, and (among other things) my article is about how religion isn’t the only thing that does that.

      So either (A) they erred in assuming it was or (B) they were making the tautological discovery I just described.

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite="" class=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>

%d bloggers like this: