Hacker News Reader

This study seems to take that into account[1]:

> Restrictions on the import and sales of WHO Class I toxicity pesticides in 1995 and endosulfan in 1998, coincided with reductions in suicide in both men and women of all ages. 19,800 fewer suicides occurred in 1996-2005 compared to 1986 – 1995. Secular trends in unemployment, alcohol misuse, divorce, pesticide use and the years associated with Sri Lanka’s Civil war did not appear to be associated with these declines.

Here's another study[2]:

> Overall suicide mortality dropped by 21% between 2011 and 2015, from 18.3 to 14.3 per 100,000. The decline in pesticide suicides during this same period was larger than for overall suicides: from 8.5 to 4.2 per 100,000, a 50% reduction. This was accompanied by a smaller concurrent rise in non-pesticide suicide mortality with a 2% increase (9.9 to 10.1 per 100,000).

[1] https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3154644/ [2] http://journals.plos.org/plosone/article?id=10.1371/journal....

Pretty much every time someone complains about the whole correlation != causation thing, it's a huge red flag that they didn't read the actual research or have no experience in academic research. It's easily accounted for and is the lowest hanging methodological fruit, and pretty much is always addressed when actually appropriate. It's an undergrad stats thing, and I know for my degree I had to have at least a year of graduate level stats, and the same for undergrad.

On the other hand, epidemiology is fundamentally incapable of proving positive causation. It can prove that something is not a cause by showing that there is no correlation, but that’s it.

the kind of conclusion as "all of 70% decrease was caused by this single factor" (as the title implies) is incredibly intelectually dishonest to ever make. So I'm not looking at the data from the perspective "I bet they were too stupid/uneducated to think about correlation != causation", but rather from the "looks like they're trying to find data to substantiate their political agenda" perspective.

Edit: having said that, yeah, I don't have any actual research experience, and the data does look pretty convincing - as in, banning pesticides probably does contribute to a reduction in suicides a non-trivial amount.

You're not looking at the data at all, and making assumptions about the scientists who did the research without even bothering to read beyond a podcast headline.

It's an interesting reaction. "I didn't bother looking at the studies, but I bet the doctors who studied this didn't realize that correlation != causation, didn't realize that Sri Lanka had a civil war, didn't realize Sri Lanka's economy was growing, etc."

It's worth keeping in mind when people wonder how there could be so many people who reject certain scientific findings. A lot of the time it's a similar attitude - "I'm not going to bother looking at the studies, but I'm guessing the people who studied it didn't think of the problem that immediately cam to my mind, and as such their results are probably wrong."

I've always wondered if it was a defense mechanism of some people with certain attitudes about their own intelligence or achievement.

Outside of the slightly distasteful attitude ("I didn't bother" and "I bet" are slightly abrasive sounding), isn't the attitude one of the skeptic? It jibes with my understanding of how to make better decisions: don't start with the assumption something is true, question information before you store it in your brain.

Annie Duke talks about this in her latest book.

No, that is being contrarian for its own sake. OP specifically said that they don't bother to look at or understand the actual research - that's not what a skeptic does.

I suppose I believe it's less important whether or not we collectively determine the correctness of the other person, and more important that I use their contrarianism to question my own knowledge. I'm less interested in judging their argument and more interested in using their comments to evaluate my own. Kind of a "less wrong" approach, but focused on my own model of reality.

Whoa, that seems like a very strong claim claim and is contrary to my understanding. I know that instrumental variables can be a powerful approach, but it's not always easy to find a good instrument. And you can try to control for confounders, but it's hard to convince yourself you've gotten all of them. See, for example, https://arxiv.org/abs/1706.04692 where standard controls were of limited efficacy and it's only when they include an additional 3,700 controls that things start looking up.

Then there are things like https://en.wikipedia.org/wiki/Parameter_identification_probl... and whole books like https://en.wikipedia.org/wiki/Causality_(book) examining the problem and its' difficulty.

Can you explain what you mean?

Easily accounted for doesn't mean just proving absolute causality - you can account for the issue by either certain statistical tests giving evidence for causation, but more easily you write about associations, what gives you evidence for the reduction of confounds (strengthening your evidence, as they did in this paper), and, of course, papers are written to an audience that understands correlation != causation, and are written as such. People pointing out such an obvious thing that is always accounted for in various ways in research shows their ignorance of the specific article as well as the philosophy of science in general.

It's just a reflexive response for some people at this point.

Their "taking into account" seems like mere handwaving away. How would the ending of a civil war not be associated with an increase in mental health?

>The decline in pesticide suicides during this same period was larger than for overall suicides: from 8.5 to 4.2 per 100,000, a 50% reduction

The way they picked these statistics just seems self serving to their thesis. Another way of looking at the same numbers is that the drop in pesticide suicides accounted for only 10.75% of the total drop in suicides.

The fact that this study focuses only on the immediate tool used to kill oneself rather than the plethora of long term contributing social factors that lead up to a suicide shows a willful ignorance of the root problems.

Exactly I would bet the ending of the civil war is the real cause here, major improvement in the quality of life.

> Their "taking into account" seems like mere handwaving away. How would the ending of a civil war not be associated with an increase in mental health?

From the paper:

> The temporal changes in suicide do not appear to correspond with the period of civil war (1983-2002).[27] By the time of the outbreak of the civil war in 1983, Sri Lankan suicide rates were already extremely high – 37.6 per 100,000 in 1982. In subsequent years, suicide rates stabilised. The civil war was still in progress at the time suicide rates began declining in 1995. By the time of the ceasefire in 2002 rates had already halved, and suicides rates were relatively stable in the 3 years after the ceasefire.

They also link to studies showing that wars tend to be associated with decreased the rates of suicide.

Considering the fluid nature of social dynamics, there still seems to be a relation there.

>By the time of the outbreak of the civil war in 1983, Sri Lankan suicide rates were already extremely high

Civil wars don't just spontaneously start for no reason. They start because people have been unhappy for a long time.

Notice also that the suicide rate began peaking during the Elam War II phase of the civil war, a time of "unprecedented brutality", and then began declining at the time of the ceasefire.

https://en.m.wikipedia.org/wiki/Sri_Lankan_Civil_War

Dismissing the study on the grounds that "How would the ending of a civil war not be associated with an increase in mental health?" looks a lot like hand-waving to me.

Is mass murder a positive manifestation of mental health?

The standards sociological finding is that suicide is more likely to decline than rise during a war:

"Suicide rates are thought to fall during wartime. Durkheim [1] suggested this was due to increased social cohesion, while others have suggested that competing outcomes may be important, with individuals who might otherwise have died by suicide being more likely to die of other causes [2]. Neeleman [3] cites this as an example of contextual effect modification, with suicide risk modified by the likelihood of dying of other causes.

Examples of reductions in suicide rates in times of war have been described for several time periods and cultures [4-9]. Marshall [10] argued that the reduction in suicide rates reflected a reduction in unemployment, a recognised risk factor for suicide [11], rather than a direct effect of war, an argument offered some support by Lester and Yang [12]. Stack [13] suggested that the suicide rate decreased as the proportion of the population employed in military roles increased."

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1526726/

Did you read the final conclusion of your study?

>Conclusion: In contrast to the established view, our findings suggest that compared to both the pre-War and post-War periods there was an increased likelihood of suicide during the Second World War if the overall declining trend is taken into account.

Moreover, the argument of "increased social cohesion" against a foreign enemy is not applicable to a civil war, where social cohesion has completely broken down.

That paper disagrees with the established view, but helpfully lays out what the established view is. Scanning papers on the topic it seems like there's mixed results war to war, and it depends what other things you try to control for - I think you could get to saying it's neutral but not that there should be a strong presumption that a war will suicide.

"the argument of "increased social cohesion" against a foreign enemy is not applicable to a civil war"

You'd get increased social cohesion among each of the ethnic groups which is in conflict with the other.

In any case this focus on the Sri Lankan civil war specifically isn't that interesting - I posted a review article above which includes controls and studies a wide range of cases across different countries, and concludes pesticide control has large effects.

"Prevention of suicide with regulations aimed at restricting access to highly hazardous pesticides: a systematic review of the international evidence:

FINDINGS: We identified 27 studies undertaken in 16 countries-five low-income or middle-income countries (Bangladesh, Colombia, India, Jordan and Sri Lanka), and 11 high-income countries (Denmark, Finland, Germany, Greece, Hungary, Ireland, Japan, South Korea, Taiwan, UK, and USA). Assessments largely focused on national bans of specific pesticides (12 studies of bans in six countries-Jordan, Sri Lanka, Bangladesh, Greece [Crete], South Korea, and Taiwan) or sales restrictions (eight studies of restrictions in five countries- India, Denmark, Ireland, the UK and the USA). Only five studies used optimum analytical methods. National bans on commonly ingested pesticides in five of the six countries studied, including four studies using optimum analytical methods, were followed by reductions in pesticide suicides and, in three of these countries, falls in overall suicide mortality. Greece was the only country studied that did not show a decrease in pesticide suicide following a ban. There were no high-quality studies of restricting sales to people for occupational uses; four of the seven studies (in three of the five countries studied-India, Denmark, and the USA) showed sales restrictions were followed by decreases in pesticide suicides; one of the two studies investigating trends in overall suicide mortality reported a fall in deaths in Denmark, but there were also decreases in suicide deaths from other methods.

INTERPRETATION: National bans on highly hazardous pesticides, which are commonly ingested in acts of self-poisoning, seem to be effective in reducing pesticide-specific and overall suicide rates. Evidence is less consistent for sales restrictions. A worldwide ban on the use of highly hazardous pesticides is likely to prevent tens of thousands of deaths every year."

https://www.ncbi.nlm.nih.gov/pubmed/28807587

Ah, there's a transcript "below the cut". So, scanning it, it seems they draw conclusions based on the strong temporal correlation between pesticide use and suicide.

> the suicide rate had increased dramatically after the introduction of highly hazardous pesticides into the agriculture

> in ’94, the most toxic insecticides [...] were banned [...] another very toxic pesticide [...] was banned in ’98, and further class two pesticides were banned in 2008 to 2011

So I guess it could be true but it's still making huge leaps of faith from the data.

Now, my moral value system doesn't classify suicide as negative, so I can't really agree with the premise of the article, yet I definitely support limitations on the availability and use of pesticides, e.g. to prevent children accidentially poisoning themselves etc.

> Now, my moral value system doesn't classify suicide as negative, so I can't really agree with the premise of the article, yet I definitely support limitations on the availability and use of pesticides, e.g. to prevent children accidentially poisoning themselves etc.

I have a similar painpoint with this: even if their attempt at reducing suicide succeeded, they didn't help people much.

Assuming (as they have to) that the situation in Sri Lanka is the same or similar as with toxic pesticides, the only thing they achieved is make escaping that shitty situation harder, not making the lives better, by effectively reducing suicidal thoughts. The point they are trying to make that suicide is a spontaneuous/irrational decision hardly counts in a statistic like that.

>Assuming (as they have to) that the situation in Sri Lanka is the same or similar as with toxic pesticides, the only thing they achieved is make escaping that shitty situation harder, not making the lives better, by effectively reducing suicidal thoughts. The point they are trying to make that suicide is a spontaneuous/irrational decision hardly counts in a statistic like that.

How does it not??? If the suicide is a decision made under the influence of a 30-minute-long depressive episode, and outside that episode the person would prefer to live, how is preventing the suicidies by making them harder to perform spontaneously not making people's lives better? Why does it 'hardly count'?

Because this is a strucural thing.

If we are to believe that not much changed, there are still ~57/1000 people who have an episode so severe as to make them consider suicide, of which 40 are stopped by the mere abscence of a simple method, there are still 57/1000 people whose lifes are so bad they honestly consider suicide.

Additionally, the researches do not say the people are not depressed before committing suicide, the research says they consider suicide with that specific method for just 30 minutes prior to killing themselves, their lifes and state of mind outside that 30 minute timeframe is not considered. And my assumption is that their lives had been bad before considering suicide, but something pushed them over the edge.

Of course their lives are bad! No one claims this solves the entire problem. But claiming that it is not an improvement is falling for the same fallacy as to when people were complaining that peta was giving water to people for turning vegetarian. You don't have to fix everything - in fact no one ever can - to be helpful

Yes, that looked like the relationship between number of pirate ships and global warming to me.

The gdp per person increased by 5x between 1995 and 2016 [1]. That makes be believe pesticides were only a small part of the change.

[1] https://www.google.com/search?q=sri+lanka+gdp+per+person

They attribute the drop to lack of pesticides because according to the researcher the drops coincided with changes in policy, drops were also seen in other countries that banned pesticides, and simular situations were seen when access to common forms of suicide were restricted in other countries also saw a drop in rates (i.e. gun restrictions led to fewer suicides). It's hard to run controlled trials at a macro scale so it's unlikely that the effect was only pesticides but it sounds like historically these sorts of interventions do work.

> ...therefore I judge the title as "click-bait-y"

I think you're out-of-line here.

It's fine to be suspicious of relative percentages, but that just means you need to dig into the numbers to understand it.

In this case underlying numbers are huge... With a population of 22 million, this is a drop from about 12,500 suicides to around 3,700 per year. And, the suicide rates themselves are high as your own source points out, so a large proportionate change is quite significant from that perspective as well. Your argument would make sense for a similar proportionate change for a country near the bottom of of the wikipedia list, but not the top.

To your second point, yes, it's hard to have complete certainty (there's no way to ethically or morally test something like this in a scientifically rigorous way) but they do attempt to account for other factors.

It's not like people shouldn't try to reduce suicide rates and attempt to measure the results of those efforts the best they can, simply because they can't have rigorous certainty in those measurements.

Also Sri Lanka had something resembling a civil war from 1983 to 2009.

Not "something resembling" it was an actual civil war

https://en.wikipedia.org/wiki/Sri_Lankan_Civil_War

Those are pretty much all valid points, however at first glance a drop from 57 incidents to 17 is at least statistically significant. The standard deviation can be assumed to be at most around sqrt(57) = ~7.5, so a difference of 40, which would be over 5 standard deviations, shouldn't normally happen.

You're right that relative percentages are sometimes a bit dubious (and even in this case you'd fluctuations on the order of 1/sqrt(57) = ~13 percentage points) but 57 is high enough that it actually has some meaning.

Regression to the mean?