Did the Lockdown Cause More Suicides?

A common assertion made by those opposed to public health measures such as lockdowns was that suicides were rising markedly – see e.g., Carney/Washington Examiner via AEI, Arthur Laffer and Stephen Moore, and blog commenters like sammy. The data are in. This article indicates suicides actually went down in 2020, 5.6% relative to 2019.

Source: Ahmed and Anderson (JAMA, 2021).

Figure 1 shows a longer time series – raw statistics (on left log scale) and per million (on right scale).

Figure 1: US Suicides (blue, left log scale), 1999-2016 log linear estimated trend (gray, left log scale), suicides per million (red, right scale).  2020 data is provisional. Source: CDC, Ahmed and Anderson (JAMA, 2021), BEA via FRED (POPTHM) for population, and author’s calculations.

Suicides fall in 2020. Using a recursive 1-step ahead Chow test on growth rate of the suicide rate, there is a statistically significant (at greater than 1% MSL) in 2020.

Interestingly, suicides rise above log linear trend (estimated over 1999-2016) in 2017-18.

39 thoughts on “Did the Lockdown Cause More Suicides?

  1. pgl

    Sammy wrote this flexible forecast:

    “The only solution is a vaccine or cure, but that is 6mos, 1 year or 2 years or never (we haven’t ever been able to develop a vaccine for a coronavirus). You can’t keep the economy closed down that long or there will be nothing left to support anybody, that is certain.”

    His 2 year to never was not exactly spot on. Of course the people at Moderna sort of knew we would have their vaccine well before his 1 year forecast. And yes Sammy – we do have something left. Wear your mask and get vaccinated.

  2. pgl

    “Interestingly, suicides rise above log linear trend (estimated over 1999-2016) in 2017-18.”

    A Trump effect?! Maybe 2020 saw suicides go down because people realized that Trump was about to be booted out of the White House!

    1. Barkley Rosser

      Following up on a post I made last night not yet up here regarding happiness trends across nations, in which I noted young people have been much more unhappy by the pandemic than older ones, I shall list the nations list that went up in happiness between 2019 and those that went down in descending order of their reported happiness in the Britain section of the March 20 The Economist. This is a 10 point scale with in 2020, and I shall list them in descending order of the 2020 number, giviing 2019 number first and 2020 number second.

      Finland 7.8 – 7.9
      Iceland 7.5 – 7.6
      Denmark 7.7 – 7.5
      Germany 7.1 -7.3
      Sweden 7.4 – 7.3
      New Zealand 7.3 – 7.2
      US 7.0 – 7.1
      UK 7.2 – 6.8
      Taiwan 6.4 – 6.7
      Spain 6.4 – 6.5
      Japan 5.7 – 6.2
      Brazil – 6.4 – 6.1
      Mexico 6.5 6.0
      Sourh Korea 5.9 -5.8
      Colombia 6.3 – 5.7
      China 5.5 – 5,7
      Rissia 5.5 – 5.5
      South Africa 4.8 – 4.9
      India 3.6 – 4.2
      Tanzania 3.5 – 3.8
      Zimbabwe 3.3 – 3.1

      for whatever all that is worth.

  3. pgl

    But here’s the complicating factor: Although lockdowns are preventing some deaths, they are undoubtedly increasing deaths by other causes. This virus is killing people not only by infecting millions but also by inducing a policy response that kills people.

    ‘First, amid this policy-induced isolation, we should worry about deaths of despair. People need connection and community. Isolation kills in many ways. We don’t have data that the lockdowns are causing more suicides, but plenty of health experts believe they will.’

    Shorter Carney – I don’t have the data to show this but some health expert who is also is the junior senator from Kentucky tells me I’m right anyway.

    ‘People struggling with drug addiction are finding it harder to get the treatment they need. This could cause relapses and eventually, death.’

    I again have no data here but let me call Rand Paul!

    ‘It’s reasonable to worry about increases in drinking, and thus alcoholism and alcohol-related deaths. (Though, with less driving, we’ll have fewer DWIs.)’

    Since a lot of alcohol related deaths are from DWIs Carney just shot himself in the foot.

    Yes AEI quality research!

  4. macroduck

    The claim that “closing down” the economy (which was never done) led to increased suicides is the sort of pretend empiricism common in politics. Something which might be true is Glenn Becked into truthiness in the absence of evidence. Turns out Glenn Becked truthiness is a pretty good sign something isn’t true at all.

    Here’s what we do know about social conditions and suicide: Unemployment increases the odds of suicide. Poverty increases the odds of suicide. Gun ownership increases the odds of suicide. So anyone who argued to open (sic) the economy to prevent suicide should support fiscal stimulus in times of high unemployment, support anti-poverty initiatives and support restrictions on gun ownership. If, in fact, suicide is the real issue.

    1. pgl

      And Princeton Stevie is now claiming that old age increases the chances of suicide. His evidence predates the pandemic so his point is ???? Maybe he will school us later on his Fox and Friends version of statistical analysis!

  5. SecondLook

    Why social analysis gets complicated:

    According to the National Safety Council, quoting:
    “For the first time since 2007, preliminary data from shows that as many as 42,060 people are estimated to have died in motor vehicle crashes in 2020. That marks an 8% increase over 2019 in a year where people drove significantly less frequently because of the pandemic. The preliminary estimated rate of death on the roads last year spiked 24% over the previous 12-month period, despite miles driven dropping 13%. The increase in the rate of death is the highest estimated year-over-year jump that NSC has calculated since 1924 – 96 years.”

    Now, factor in the percentage of those deaths that are due to vehicular suicide, or what is termed willful disregard for personal safety including alcohol impairment.

    To quote a literature and meta-data review, circa 2003,:
    Literature shows suicides by motor vehicle crash to be a small proportion of the total road toll with estimates of between 1% and 7% of all motor vehicle crash fatalities. Official statistics suggest a lower proportion (0.8% of motor vehicle fatalities (Canada, Australia, Sweden) and 0.5% of suicides (Australia)). Suicide by motor vehicle crash typically involves either a single vehicle, single occupant crashes, or a head-on collision of a single occupant vehicle with a heavy goods truck. Most cases of suicide through a motor vehicle crash are male (approximately 90%). The majority of victims are in the 25-34 age group and 24% of all victims have attempted suicide on at least one previous occasion. Approximately half the traffic suicide victims have a psychological state ranging from “mental disturbance” to being “depressed”.
    That higher estimate – of about 5-7% – does fit what is a widespread understanding in the insurance industry.
    Granted, the overall numbers aren’t large, but it is a good example of how messy working with this kind of data can be.

    1. Menzie Chinn Post author

      SecondLook: The 2020 increase in vehicular deaths is approx. 2953, the decrease in suicides is 2677. High end ratio of 7% attributable to vehicular suicide implies 208 deaths are misattributed to accidents rather than suicide. This implies the correct decrease in suicide count would be 2469 instead of 2677.

      On the other hand, WSJ suggests that the typical ratios might not apply.

      1. SecondLook

        I did stipulate the numbers weren’t large but was offering vehicular suicide as an example of how the official suicide numbers are soft data, not hard.
        The accepted definition of suicide is “the act or an instance of taking one’s own life voluntarily. and intentionally.”
        Suicidal behavior doesn’t qualify – the teenager who gets drunk, doesn’t wear his seatbelt and drives onto the highway at high speed, and dies, is not counted as a suicide. After all, behavior isn’t always an indication of intent, right? Well, not conscious intent, and that is all that matters, right?
        Similarly, overdosing on opiates, alcohol poisoning, etc, are very nearly always counted as accidental deaths.
        And so on.
        Given our society’s distaste for the act, it’s not surprising that obtaining more accurate data is difficult.

        Hmmm, the relationship between less traffic due to the pandemic and higher fatality rates could be cross-compared to prior years ratios of those two events among the various states. Say, do states with high average percentages of 70 mph drivetime have greater than average vehicular mortality – that speed implies little average traffic (I think Wyoming is on top) – also have higher than average mortality rates? By the way, 70 mph is considered the death speed odds of survival from an accident at that speed, is vanishing point low.

  6. Barkley Rosser

    Last week’s The Economist had an article in the section on Britain that actually covered the world about changes in happiness levels from 2019 to 2020. In Britain happiness levels went down, but the striking story was that they massively dropped for younger people while only doing so mildly for older people. This differential age-related result is thought to hold more widely around the world, younger people more negatively impacted in terms of happiness.

    The article than provided a fascinating figure showing changes in overall happiness for many nations. This was quite curious in that there was no general pattern with it going up in about as many nations as it went down, the overall level of measured happiness. There was not a clear pattern and definitely not one clearly tied to whether there were lockdowns or not.

    Among those with the most sharply declining happiness levels were Mexico and Brazil, both places with weak lockdowns, especially the latter. Brazil might be suffering a Bolsonaro Effect similar to a Trump Effect in the US, with almost complete avoidance of any clear actions against a raging coronavirus. Two nations at the top of the world happiness scale saw further increases: Finland and Iceland, with the former having a fairly strong lockdown. Weak lockdown Sweden saw a decline in happiness.

    The US saw a modest increase in happiness, which would consistent with a decline in the suicide rate. Article noted polarized intensity in US and wide variety of responses within the country in terms of policies across the states. The article seemed sort of mystified by the US result, but that is what was reported.

  7. ltr


    February 23, 2021

    As the pandemic took hold, suicide rose among Japanese women.
    By Motoko Rich and Hikari Hida

    While the pandemic has been difficult for many in Japan, the pressures have been compounded for women. As in many countries, more women have lost their jobs. In Tokyo, the country’s largest metropolis, about one in five women live alone, and the exhortations to stay home and avoid visiting relatives have exacerbated feelings of isolation.

    Other women have struggled with the deep disparities in the division of housework and child care during the work-from-home era, or suffered from increases in domestic violence and sexual assault.

    The rising psychological and physical toll of the pandemic has been accompanied by a worrisome spike in suicide among women. In Japan, 6,976 women died by suicide last year, nearly 15 percent more than in 2019. It was the first year-over-year increase in more than a decade.

    Each suicide — and suicide attempt — represents an individual tragedy rooted in a complex constellation of reasons. But the increase among women, which extended across seven straight months last year, has concerned government officials and mental health experts who have worked to reduce what had been among the highest rates of suicide in the world. (While more men than women died by suicide last year, fewer men did so than in 2019. Over all, suicides increased slightly less than 4 percent.)

    The situation has reinforced longstanding challenges for Japan. Talking about mental health issues, or seeking help, is still difficult in a society that emphasizes stoicism….

  8. nothanks

    This is a weak empirical analysis. By this reasoning, FOMC rate cuts hurt the economy because GDP goes down as they cut. There’s a lot of cross-sectional and panel variation in lockdowns, so even without an identification strategy, you might start there for descriptive analysis. I’d leave this one to the applied micro folks, your analysis here doesn’t cut it.

    1. pgl

      Identification may be an important issue in econometrics but your point sort of alludes to the idea that suicides may cause lockdowns. Pray tell us what empirical evidence gives you that idea???

      This is like when my Econometrics professor used to tell his joke about the effect of the number of fish in Brazil in some macroeconometric model for the US. I know it was an odd joke but trust me – it cracked us up every time.

      1. nothanks

        i did nothing of the sort. there may be any number of third factors affecting both lockdowns and suicides simultaneously, resulting in biases in both directions, large or small. i merely point out that one cannot draw any conclusions from the data Menzie presented

        1. baffling

          “i merely point out that one cannot draw any conclusions from the data Menzie presented”
          that is not an accurate statement at all. and if you believe something is incorrect, you need to be specific about what it is that is wrong and how it is to be corrected.
          you are playing the game of trying to introduce uncertainty into a situation to increase confusion when the results are against your beliefs. if you think prof chinn is wrong, then provide us with a better analysis and interpretation of what happened over the past year? or chastise those who argued that lockdowns increased suicides to begin with, since apparently their analysis is even less convincing.

          1. nothanks

            I have no basis for saying it’s correct or incorrect, only for saying that you cannot draw a conclusion from the data presented. We do not know as much as we think we know, and drawing a casual conclusion requires really strong arguments about how random your variation is. There’s no attempt at that here.

            If I thought the guys he is criticizing cared, I’d chastise them too! Menzie at least claims to care about academic rigor, I don’t think they do. He should be receptive to these observations, they’ve already told us they’re not.

          2. baffling

            “and drawing a casual conclusion requires really strong arguments about how random your variation is. There’s no attempt at that here.”
            you are implying an argument about how random the variation is, without evidence. that is my point.

            i will also agree that further data would be helpful. that said, the world is full of inconclusive data sets, and if you wait until you get perfect data to prove your point, the time is past. that is a luxury most people do not have-its part of the ivory tower issue thrown at many academics. prof chinn’s analysis may not be complete, but it is an attempt and spurring the discussion. you seem to be knowledgable about the issue nothanks. are you now delving deeper to provide a more robust and timely analysis for us? i would rather have prof chinn’s contribution now than a perfect analysis 5 years from now. you seem to prefer he remain silent on a relevant and current issue. i think we can all agree that does not help society at all.

          3. Baffling

            “If I thought the guys he is criticizing cared, I’d chastise them too! Menzie at least claims to care about academic rigor, I don’t think they do. He should be receptive to these observations, they’ve already told us they’re not.“
            If you hold one side accountable, but ignore the other side because they will not listen, you are implicitly supporting the other side by not criticizing. At least that is the message sent to readers on this blog. Readers only hear what you say, and not what you don’t say.

          4. baffling

            “We do not know as much as we think we know, and drawing a casual conclusion requires really strong arguments about how random your variation is.”
            nothanks, i think you may not be paying attention to what prof. chinn has argued. he did not say that lockdowns reduced the number of suicides. what he said was
            “A common assertion made by those opposed to public health measures such as lockdowns was that suicides were rising markedly” was not supported by the data. and i think this case can be made with the data we have. he was NOT arguing that lockdowns reduce suicides. he did argue that suicides went down in 2020, but that is factually correct.

        2. pgl

          I guess Identification means something more to some of us than you think it means. Now if you are saying that the virus is causing lockdowns and the virus is causing more suicides, then I have no problem with that. But what you are saying is contrary to the right wing fluff that Menzie is critiquing.

          1. nothanks

            I’m not drawing any conclusions here. But Menzie’s arguments are no better than the ones he’s criticizing, this data just doesn’t support drawing casual conclusions. There are entire fields full of economists who debate this stuff day in and day out, and this argument doesn’t pass the faintest muster.

            Moreover, as secondlook pointed out above, dynamics matter. The literature on the cyclicality of health has really rich dynamic empirical results; the health consequences of economic downturns aren’t purely coincident. If Menzie were an applied microeconomist, he’d probably be familiar with that literature. He might be familiar with that literature, and decided not to mention it for some reason, but it seems pretty relevant to the question.

            This is just not his wheelhouse, and this data doesn’t support any conclusion whatsoever. “Lockdowns cause suicides” may be true at the same time that “recessions reduce cyclicality,” which is actually a classic result. It may not.

          2. nothanks

            Apologies. I meant “Causal” and not “casual” (autocorrect?) and “recessions reduce mortality” (my error)

  9. F

    Not to give the critics too much credit, but Unintentional Injuries did go up by 20000 (11%), and this is the category that includes drug overdoses. Preliminary CDC data indicates that this is mostly due to drug overdoses, which increased from an annual rate of 75000 to 88000 by August 2020. If that trend continued through the rest of 2020, almost all of the 20000 additional deaths from Unintentional Injuries were likely due to drug overdoses, and this represents a 20-25% increase over the previous year.

    1. pgl

      In my city (NYC) we have a lot of homeless doing a lot of drugs. You are not telling us that they are obeying lock downs I trust.

  10. Steven Kopits

    Those over 85 have the highest suicide rate. They are also the ones most likely to die of covid.


    “Statistics from the National Council on Aging state that those 85 years and older have the highest suicide rate of any age group. It is over four times higher than the nation’s overall rate of suicide[1]. As with most age groups, the majority of elders who kill themselves are male. As high as those figures are, the American Association for Marriage and Family Therapy (AAMFT) suggests that the rates for elder suicide are under-reported by 40 percent or more due to what they refer to as “silent suicides.” These would include overdoses, self-imposed starvation, and dehydration, as well as accidents. The rate of suicide completion in this age group is also high due to the lethality of methods used.”

    1. pgl

      I would suspect they truly miss their families, which is a real shame.

      Of course we should all sit back and wait for Bruce Hall to go his usual pathetic and dismiss once again the importance of losing grandma.

    2. pgl

      Wait a second here. Just noted the date of your article:

      Posted Jan 28, 2020

      If you noted this before you posted, then you knew you were posting something unrelated to the news developments from this pandemic.

      Come on Stevie – we know you bilk your clients with your dishonesty buty damn it – stop lying here. Geesh!

      1. Steven Kopits

        I wasn’t attempting to comment on suicide rates during the pandemic with the article. Rather, I was trying to determine the a priori tendency of the very elderly to commit suicide, since this overlapped with the primary covid mortality demographic. If we allow that suicides were 5,000 below expected and that deaths were 455,000 over expected, then those covid deaths could have hidden those who would have committed suicide otherwise. That is, if 1% of those who died of covid would have committed suicide under other circumstances, then the underlying suicide rate would have been unchanged compared to trend-adjusted 2019.

        To get a better apples-to-apples comparison, we would want to look at suicide rates from those not materially at risk from covid mortality, say, 18-24 year olds and 25-54 year olds.

        1. pgl

          Apologies for the misunderstanding now that you have finally cleared up why you brought this up. I will defer to our host, however, as to what he thinks of what you just wrote.

  11. Steve

    The reduction in suicides is even more impressive given the reduction in employment. Loss of jobs markedly increases the likelihood of suicides.

  12. Steven Kopits

    Deaths in 2020 appear 455,000 above trend, which I think may be reasonably attributed to the covid pandemic. Suicides were only 5,000 below trend, and we know that the very elderly are 4 times as likely as the general population to commit suicide and that these same elderly were the ones most likely to die of the coronavirus. As a result, I don’t think these figures by themselves are sufficient to draw inferences about suicide rates. I think we’d need to look at suicide rates in not-at-risk populations, eg, 25-54, and say, the 18-24 age groups. That would be more enlightening regarding underlying trends.

    1. pgl

      This clears up what you meant with

      “Steven Kopits
      April 2, 2021 at 9:57 am”

      I still say let’s see what our host thinks about this.

    2. baffling

      suicides were the only category with any meaningful drop. and seniors account for about 8ooo suicides per year. this requires that an awful lot of would be suicides in seniors were cut off by covid instead. i would have expected to see a bigger drop in the other categories where seniors have large numbers of deaths-but that did not occur. i do not think the data is saying what you are implying, steven.

  13. Moses Herzog

    This is what happens when people get obsessed with both speed, and making money:

    Any positive thoughts I had on AstraZeneca, at least in the short-term, have been thrown out the window. And really at this point, I regret ever having made positive remarks about the AstraZeneca vaccine. I was wrong about AstraZeneca, and apparently they can’t be counted on to do anything correctly.

    I would go with Moderna I think at this point, if I had a choice between the different vaccines and would only use J&J if there were no other options.

      1. baffling

        lots of bosses think just because they mandate something to occur, it should occur without problems. many have lost the understanding of knowledge and training that takes time to acquire and deploy. just in time manufacturing may apply to product, but it does not apply to labor.

Comments are closed.