The Plural of Anecdote is Data.
The line above comes from a story that, according to Gregg Jarrell, one of my former colleagues at the University of Rochester, George Stigler told him in the 1970s.
If you have a lot of anecdotes, then beyond some point they become data.
I thought of that based on a story I heard a few days ago about a death attributed to Covid. I’m going to hide a few details because I don’t have permission to tell the story with the name and place.
Here’s the story:
A friend who’s a very careful empirical economist told me that a relative’s wife died recently. The cause was clearly a heart attack. But when they tested her, they found Covid. The widower was told that FEMA, the Federal Emergency Management Agency, pays for funerals if the person died of Covid. So did he want the cause of death to be written down as Covid.
What do you think he answered?
I wonder how many other cases there are like this.
Here’s a quote from a June 6, 2022 press release from FEMA:
WASHINGTON — FEMA has provided more than $2.5 billion to over 389,000 individuals and families to assist with COVID-19-related funeral costs for deaths occurring on or after Jan. 20, 2020. This assistance helped pay for the funerals of more than 398,000 people who died from COVID-19.
398,000 people by June 6 is close to 40% of the total number of Covid deaths in the United States. What fraction of those deaths were really due to Covid? I don’t know. We may never know. But the answer is crucial.
Oh and, by the way, why the hell should we taxpayers we forced to pay for those funerals?
Note: The pic above is of George Stigler.
UPDATE: This isn’t about the United States but the story of Britain is also interesting. I’m not saying the incentives are the same but the data are quite shocking. HT2 Tyler Cowen.
READER COMMENTS
Garrett
Aug 17 2022 at 9:00am
Another way to look at the impact of the pandemic is excess mortality. This measure became negative in the US, for the first time since February 2020, in March 2022.
Daniel
Aug 17 2022 at 10:05am
Garrett makes a great point – if reported-COVID deaths (1.03M) are “untrustworthy” as an indicator of deaths due to COVID we can turn to cumulative excess deaths (which don’t have misreporting concerns) as an alternative indicator of deaths due to COVID (presuming our baseline projection of deaths is pretty stable and that 2020/2021/early2022 are unique primarily due to COVID). Cumulative excess deaths are sitting at 1.08M (as of June), so I’d say the reporting is about as “untrustworthy” as reading an expert’s blog on their topic of expertise – you know it’s not going to be perfect, but it’s pretty informative.
https://ourworldindata.org/grapher/excess-mortality-raw-death-count
https://ourworldindata.org/grapher/cumulative-excess-deaths-covid?country=~USA
robc
Aug 17 2022 at 1:19pm
Kevin covered it a bit, but excess deaths need to be split into two parts:
Excess deaths due to covid.
Excess deaths due to reaction to covid.
I don’t know how to split it up, but when you go away from raw deaths and convert instead to years lost, I think the second is larger than the first. The years lost due to economic effect is huge, if nothing else.
JFA
Aug 17 2022 at 2:04pm
You can split up excess deaths by state and look at how laxer states did relative to stricter states. There is not much of a difference. Overall, the excess deaths over the past 2.5 years have been due mostly to Covid-19. Go to this website (https://www.cdc.gov/nchs/nvss/vsrr/covid19/excess_deaths.htm) and you can choose the state you want to view. I think clicking on Florida or Texas and seeing the massive spikes in excess deaths in the second half of 2021 (that follow the delta and omicron waves) should probably put your mind at ease that lockdowns weren’t causing the excess deaths.
robc
Aug 17 2022 at 4:29pm
A year of life is worth about $7.5MM dollars.
How many dollars were lost? A quick google search suggest roughly $16T. That would be about 2.1 million years from economic impact alone.
That may or may not have anything to do with lockdowns. Some of it would be from people changing behavior on their own.
David Henderson
Aug 17 2022 at 7:33pm
robc writes above:
A year of life is worth about $7.5MM dollars.
If $7.5MM means $7.5 million, then you’re off by at least an order of magnitude because the standard VSL is about $11 million, and that’s for a life, not just a year of life.
robc
Aug 18 2022 at 12:55pm
You are correct. I searched for year of life, but got back the value for life.
My link may have been old, hence the 7.5 vs 11 difference. Plus, you know, its all ballpark.
I saw as high as 300k for a year. That is high if a life is $11M. But I would rather err on the side of caution.
So recalculating, $16T divided by $300k is 53 million years of life lost due to economic effects of covid.
Daniel
Aug 17 2022 at 10:03pm
Indeed, as I said above, there are some assumptions involved in using excess deaths as an indicator of COVID deaths. Luckily, they’re pretty reasonable. Excess deaths due to the reaction to COVID is likely to be small and we can further validate excess deaths as an indicator of COVID deaths.
Other patterns of covariation help here – as JFA says, you can look at heterogeneity across states to get an idea. You can also just look at the pattern of excess deaths over time (not just the cumulative number, but the WAVES in it) that suggest it’s not stringency of response (lockdowns and case rates diverged if you recall) but epidemiological dynamics driving it. This is a pretty basic analysis (see Scott and David’s discussion below), but I think one has to have a pretty warped prior on this to think true COVID deaths make up less than a large supermajority of the reported deaths (i.e., that reported deaths can’t be trusted).
Here’s another basic start (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7539724/): 150 patients had COVID listed as the official cause of death; of those, only 2 had neither pulmonary infiltrates on chest radiography nor oxygen therapy, but they had serological evidence of systemic inflammation (we saw COVID causing this) that could have plausibly led to their renal failure. Small sample size, but very high, if not 100%, hit rate.
CDC (https://www.cdc.gov/nchs/covid19/mortality-overview.htm) says 90% of COVID death certificates have COVID as the underlying factor. 10% have COVID as a contributing factor. Some information on those determinations: https://www.cdc.gov/nchs/pressroom/podcasts/2022/20220107/20220107.htm
Jonathan
Aug 18 2022 at 4:59pm
Healthcare planners sometimes use Quality-Adjusted Life Years (QALYs) to assess the economic benefit of interventions. This Willingness to Pay Study (WTP) came up with a value of $2,799 to $3,599 per QALY. I think planners use larger numbers – $50,000 [oral communication from a Stanford cardiologist] to $250,000.
My back-of-the-envelope calculation of the QALY valuation implied by the lockdown costs early in the Covid-19 era was upwards of $5 million. The non-pharmaceutical interventions were stupendously expensive, reducing GDP by a significant fraction. And the QALY cost was affected by the steep age gradient in Covid mortality – 3/4 were over age 65 – and by the 80% comorbidity rate that reduced quality-of-life (at least 40% of fatalities were in nursing home residents).
Ryan M
Aug 19 2022 at 4:14pm
I’m not so certain about this. There are many factors of “covid response,” which those looser jurisdictions were not able to avoid. Even in states that remained open, the medical establishment, for instance, tended to respond with some uniformity (terribly misguided, in my opinion) pretty much across the board.
When we’re discussing “excess deaths,” it is obviously necessary to measure “excess” against the expected number of deaths, which we generally assume is going down over time. An interesting experiment would be to see what would happen to the actual death statistics if – absent covid – the changes made to our medical establishment/practice were nevertheless implemented. That cannot be done, but I think it gets us thinking in the right direction. Even this assumes that trends in death statistics would have remained steady, such that we have a reasonable metric against which to measure “excess” deaths. I think those two factors- a) the number against which we weigh excess being unreliable, and b) the actual impact of covid-response, even in states that remained largely open – support the author’s initial position that our understanding of covid mortality is likely far less accurate than we imagine.
Kevin Corcoran
Aug 17 2022 at 12:10pm
Excess mortality is a useful data point, but there are limits to how much we can infer “excess mortality during the Covid pandemic” to “excess mortality caused by Covid infection.” There are other sources of excess mortality during this time. Here’s a few quick examples worth considering. For many months (and longer, in some areas) hospitals and doctors offices were strictly limiting the services provided to patients, and that reduced patient care translated to cancers caught later (sometimes too late), other medical conditions being overlooked, diseases not being treated, etc. This would account for some portion of that excess mortality. The economic damage caused by many tens of millions of people losing their jobs as the economy was shut down would also have deleterious health effects, especially at the margin. Some argue that “deaths of despair” increased during this time as well. All the deaths that occurred to these factors would appear as “excess deaths.” It’s too soon to say how many deaths can be attributed to these causes, but just looking at excess mortality numbers is misleading.
TGGP
Aug 17 2022 at 12:47pm
If you read Robin Hanson or Greg Cochran (both of whom made multiple successful bets on deaths during COVID), you’d know that a whole lot of medical care has a negligible effect on actual health. Vaccines are, of course, an exception to that.
David Henderson
Aug 17 2022 at 1:07pm
Hmmm. I’m not familiar with Greg Cochrane, but I’m quite familiar with Robin Hanson’s thinking. I seem to recall that Robin summarized the effect of medical care the way many advertisers summarized the effect of advertising: we’re convinced that half of it is wasted but we’re not sure about which half.
If a huge percent of medical care is delayed by months or a year, there is likely to be excess mortality due to that fact.
JFA
Aug 17 2022 at 1:29pm
“If a huge percent of medical care is delayed by months or a year, there is likely to be excess mortality due to that fact.”
That is possible, but if that were a large driver of excess mortality, you probably wouldn’t expect excess mortality to be so correlated with Covid case rates.
Kevin Corcoran
Aug 17 2022 at 1:49pm
Actually that is what you would expect to see, if the length and severity of lockdowns were themselves correlated with Covid case rates.
JFA
Aug 17 2022 at 2:37pm
David was suggesting that delaying treatment (by months or a year) would cause increase in excess mortality. So that would mean that if you delayed your medical visit then you would die later. There’s no reason why these later deaths would correlate with case rates.
What you are suggesting is that people were delaying treatment and dying at the time the decision to delay treatment was made. If that were the case, then yes… you would expect excess deaths to correlate with case rates. But I think that is a harder case to make. Nationally, hospital visits dropped 30% (relative to 2019) by mid-April of 2020 and then steadily recovered throughout the year. It was at 13% below 2019 levels by January 2021 and 3% below 2019 levels by June of 2021. It’s essentially a steady rise in hospital visits with little to no variation with case rates. If delaying treatment caused immediate excess deaths, then you would expect a mound of excess deaths from March 2020 through July 2020 with little undulation after that. Instead, what we see is that even with hospital visits near parity with 2019 by the summer of 2021, there are massive spikes in excess deaths starting in July of 2021. This suggest people were getting the care that would immediately benefit them and delaying the care that could wait.
You can also see in the data that deaths from cancer remained steady and did not see spikes relative to previous years.
You also noted the economic damage might have led to more deaths of despair. This may account for some of the excess deaths, but I would also point you to research showing that family balance sheets (at every decile of the income distribution) were in better shape at the end of 2020 than at the beginning. Then there’s the fact the unemployment was at very low levels throughout 2021 even though excess deaths undulated throughout the year. I don’t think it’s reasonable to associate a large chunk of the excess deaths to deaths of despair.
As I mentioned elsewhere, check out excess deaths in Texas and Florida. They correlate very well with case rates but I don’t think anyone would say they were anywhere near lockdown during the delta and omicron waves.
Michael
Aug 18 2022 at 5:18pm
A way to get at that question of attribution is to look at regional and temporal differences in Covid-19 policies versus death rates. Not all states behaved even remotely the same, many policies were different in the later waves vs the earlier.
To a first approximation, were n’ttheir more restrictions in blue states than red ones? I think one would expect more deaths in blue states if restrictions were causing the deaths. I don’t think that is what the data show.
Hanoch
Aug 19 2022 at 12:13pm
I am wondering how useful mere excess death statistics can be given that, at the same time as covid, opioid-related deaths have also been increasing significantly. See, e.g., https://www.cdc.gov/opioids/data/analysis-resources.html#anchor_trends_in_deaths_rates
JFA
Aug 19 2022 at 12:33pm
Here’s some data on overall drug overdose deaths: https://www.cdc.gov/nchs/nvss/vsrr/drug-overdose-data.htm. In the 12 months leading up to February 2020, there were about 73k overdose deaths. In the 12 months leading up to February 2021, there were 96k overdose deaths. In the 12 months leading up to February 2022, there were about 106k overdose deaths. For simplicity assume that without the pandemic the flat trend of the previous 2 years of around 70k overdose deaths per year would have continued. That would mean around 62k extra deaths due to overdose deaths. That’s out of around 1.1 million overdose deaths. The change in overdose deaths is also not correlated with Covid cases or deaths.
This would suggest mere excess deaths are still a useful statistic.
JFA
Aug 17 2022 at 10:41am
It has been repeatedly shown that Covid has negative cardiac effects. You can do a literature search or look at the excess deaths by cause subgroup. Deaths from ischemic heart disease and hypertensive diseases have been elevated since the pandemic began (as have overall deaths). The question here is whether the woman would have died had it not been for covid. I also find it hard to believe that coroners and doctors are asking enough people what cause of death they want to put on the death certificate for it to make a big difference in national level numbers.
Just to forestall the objection that blames “lockdown” on excess deaths: those states that instituted fewer measures have massive excess deaths (which correlate very well with case rates), just like everyone else.
David Henderson
Aug 17 2022 at 11:25am
You write:
Why do you find it hard to believe?
JFA
Aug 17 2022 at 11:46am
Why would the doctor or coroner consult the deceased’s family member for what to put on a death certificate? I have had many family members (thankfully most of them were late in life) die in my lifetime, and I don’t recall any family members being asked what they wanted on the death certificate. To my knowledge, that is not standard operating procedure.
Vivian Darkbloom
Aug 17 2022 at 12:25pm
“To my knowledge, that is not standard operating procedure.”
I’m pretty sure that David’s point is that it isn’t and shouldn’t be.
So, “why on earth would they do that”?
David’s post explicitly provided an answer to that—it is because FEMA pays for the funeral if Covid is listed as the cause of death.
Once again, incentives matter!
JFA
Aug 17 2022 at 1:25pm
Vivian, it’s still unclear to me why the doctors would ask. Could they just not put it on there like any other multi-cause death without asking the family? Seriously, how many doctors are there sitting at their desks wondering whether to put Covid as a cause of death and say to themselves “Let me ask the family what they think”? I’m not saying this never happens, but David’s point was that this is happening enough to where the numbers cannot be trusted (hence the title “Why We Can’t Trust the Covid Death Statistics”). What I said is that I doubt this is happening enough to make a big difference in the reported figures (especially when excess deaths are in line with the number of Covid deaths).
Also, I’ll reiterate that David’s anecdote starts off with someone whose cause of death was “clearly a heart attack” but also tested positive, and ends with the suggestion that putting Covid on her death certificate is inaccurate (maybe even fraudulent). I think this is a mistaken view as 1) covid has been shown to cause cardiac distress (which can lead to heart attacks) and 2) it implies that only deaths with Covid listed as the only cause of death should be counted as Covid deaths. Many deaths are multi-causal. Should we classify all those deaths from a rare form of pneumonia back in the 1980s as due to only pneumonia rather than due to AIDS since the cause of death is clearly pneumonia?
Vivian Darkbloom
Aug 17 2022 at 2:05pm
Of course, they could just put the cause of death on the certificate solely using their professional judgement. However, I can imagine a physician dealing with a grieving family thinking that maybe this might help them out a bit.
And, speaking of incentives, here is something David didn’t mention: Hospitals are paid 20 percent more for treating Medicare “Covid patients” even in absence of a positive Covid test. Maybe the doc and the hospital administration also had more than just the exercise of professional medical judgement in mind?
https://www.aha.org/special-bulletin/2020-03-26-senate-passes-coronavirus-aid-relief-and-economic-security-cares-act
I recall visiting Venice, California a number of years ago and walking along the boardwalk there. It was quite an eye-opener for me because I hadn’t been there in quite some time. I was wondering why all the advertisements for medical marijuana certificates posted along the boardwalk by “physicians”. Was this simply a matter of the exercise of “professional medical judgement” or was there something else at play?
The federal government (and perhaps states as well) have created quite a number of incentives for “with covid” cases to be reported as “from covid” cases. This makes the reported statistics subject to question, and for good reason. As David clearly stated, he doesn’t know the extent of this, and neither do I. But, to dismiss that incentives might skew the reporting strikes me as more than a bit naive.
Kevin Corcoran
Aug 17 2022 at 2:15pm
JFA asks:
I make my living working with doctors, and based on my experience, my answer to this question would be “quite a lot, actually.” I can’t give a precise quantification, of course, but JFA seems to think that if it happens, it would be a rather obscure occurrence, whereas in my estimation, it’s something I would expect to see a decent percentage of the time.
JFA
Aug 17 2022 at 3:27pm
@Kevin: I’ve texted back and forth with several doctor friends today, and they said they have never consulted the family about what to put on a death certificate. Here’s an article (https://www.medpagetoday.com/opinion/working-stiff/79819) I found entitled “Your Patient Died. Now What?” Here’s a key line: “I can’t begin to tell you how many family members complain that I, the forensic pathologist, am the “first doctor who explained things to me.”” Now the guy is a pathologist, but the statement seems indicative of doctors not asking the patients. Again, I’m not saying families are never consulted… it just doesn’t seem like it’s done that often. But chances are, absent a survey taken by a random sample of doctors, we probably won’t know whether this is common practice or not.
David Henderson
Aug 17 2022 at 2:01pm
JFA quotes me:
Then he writes:
I would want to know more about the correlation. We had high Covid case rates all over the country from March 2020 on. And we had high excess mortality. But we all also had many people cancelling their medical treatments or, worse, being cancelled without their consent, and so missing procedures for months to sometimes over a year. So all of these things were happening at the same time. Is JFA saying that we can know that the cancellation is not an important driver of excess mortality? I don’t see how.
And remember that it doesn’t have to be the main driver. I’m quite open to the idea and, indeed, I believe the idea that the main driver was Covid. But if, say, 100,000 deaths out of the million plus were due to other diseases, that’s a big deal.
JFA
Aug 17 2022 at 3:15pm
I think I was replying to someone above while you were typing this, so (at the risk of redundancy), I’ll reply here.
There’s no doubt that there was a large drop in medical visits starting in March of 2020, but medical visits were steadily increasing after that reaching near parity with 2019 volume by June of 2021.
If delaying treatment were the cause of excess deaths (and delayed treatment or diagnosis resulted in future death), excess deaths wouldn’t correlate so highly with case rates or measured Covid deaths. (I ran the numbers last year, and the correlation between excess deaths and Covid deaths was around .92 if I recall correctly).
“Anti-lockdown” states (for lack of a better term) saw massive spikes excess deaths during the delta and omicron waves. Their percent excess was similar to other states. I do think this should be considered prima facie evidence for lockdowns not causing a large portion of excess deaths (or even the majority as robc suggests).
If delayed diagnosis or treatment were an important factor, I think we would expect to see an increase in cancer deaths, but we don’t.
I think it is also fair to say that some amount of the excess deaths were probably unmeasured Covid deaths. Note that (just like the flu) Covid leads to an increase in the risk of heart attack and stroke. So while I would not say that Covid is the primary cause of death of someone having a heart attack 3 months after testing positive, I think it would be foolish to say that that death is unrelated to Covid. Given the elevated deaths caused by circulatory diseases, I think it’s important to consider how many of these excess deaths Covid contributed to.
So we have the following (probably not exhaustive list) non-Covid buckets for excess deaths:
deaths due to delayed diagnosis and treatment due to Covid measures
deaths due to economic distress (deaths of despair)
deaths not directly caused by Covid but which were indirectly affected by it
Given 1-4 from above, I’m not inclined to give much weight to #1. I probably think 3 is a more significant factor than you do. There was quite a spike in overdose deaths. If you attribute 100% of the increase to lockdown in 2020, then that’s about 50,000 over 2 years (https://www.cdc.gov/nchs/nvss/vsrr/drug-overdose-data.htm).
Andrew_FL
Aug 17 2022 at 2:29pm
The demographer Lyman Stone has some of the best analysis of excess deaths. Unfortunately I couldn’t find a tweet where he breaks down deaths into respiratory vs other causes since October, but you can see that respiratory excess deaths clearly strongly correlate with, and exceeds especially in the early pandemic, official COVID deaths.
vince
Aug 17 2022 at 2:42pm
Let’s not dismiss that the media loves to promote fear, Pharma wanted to promote vaccines, and the government wanted to control behavior. Of course, Pharma also rewards the media and is rewarded by the goverment.
Reporting a higher number also made Trump look bad, a hobby among his enemies. During his campgain, Biden said anyone who allowed over 200,000 Covid deaths should not remain as President. The counting has backfired, with Biden now at over half a million.
Scott Sumner
Aug 17 2022 at 3:23pm
The actual number of Covid deaths is very likely larger than the official total. Many Covid deaths are reported as due to some other factor.
While excess deaths could be due to factors other than Covid, a careful analysis of the data shows that it almost all is Covid.
David Henderson
Aug 17 2022 at 3:32pm
You write:
What’s your evidence for that?
You write:
Could be, but I’m not familiar with that careful analysis. Please provide a cite.
Todd Kreider
Aug 19 2022 at 3:20pm
“The Ethical Skeptic”, who has an engineering degree from M.I.T., has not been happy with the CDC making long delays with posting recent data and relabeling Covid deaths. He also sees an increase in cancer deaths which goes against JFA’s statement: “You can also see in the data that deaths from cancer remained steady and did not see spikes relative to previous years.”
cancer increase:
https://twitter.com/EthicalSkeptic/status/1560456057272635400
renaming cause of death:
https://twitter.com/EthicalSkeptic/status/1560357663678029827
JFA
Aug 19 2022 at 5:33pm
Todd,
Go to https://www.cdc.gov/nchs/nvss/vsrr/covid19/excess_deaths.htm, click on deaths by cause subgroup, click “update dashboard”, then scroll down and choose “malignant neoplasms” under “select cause of death group”. It all wiggles around the 2015-19 trend. There are some weeks when 2022 is higher, and some where it’s not. Again, I’m not saying that no one didn’t get life saving medicine or a diagnosis that was too late. What I’m saying is that it is most likely not a big number, and certainly does not explain the massive increase in excess deaths.
Brendon Xiques
Aug 18 2022 at 9:35pm
I visited the CDC website and saw listed there the fact that 5.28% of total reported COVID deaths were from “COVID” alone. In other words, there were other factors contributing to the other 90+% deaths. Then later the Director of the CDC said, within my hearing, that only about 6% of all “COVID” deaths were from COVID alone. Then, the Director CDC stated that of the other deaths reported as COVID each one had an average, I repeat AVERAGE, of “more than four co-morbidities”. That is a direct quote. I read RFK Jr’s book in which he also asserts that an average of about 4 co-morbidities or underlying conditions were involved in 94% of reported so-called COVID deaths. Last, I met an EMT from New Orleans who stated to me he picked up the body of a man who had shot himself in the head. On autopsy, this man was discovered to have COVID. He was listed as a COVID death. Also, hospitals got paid $30K + for treating COVID patients. “Nuff said about this. And are we supposed to trust our government? Really?
Comments are closed.