Home > Uncategorized > Why We Can Be Pretty Sure China is Lying About COVID-19

Why We Can Be Pretty Sure China is Lying About COVID-19

This post has nothing to do with inflation (although to the extent that China’s problems are worse-than-advertised, the supply shock could be worse-than-expected and the resulting inflation impulse larger-than-expected). However, as a longtime participant in the financial markets – which, over the last decade or two has increasingly meant an observer of China and Chinese data – I have been absolutely flabbergasted at the evident willingness of press, government officials, and healthcare experts to accept at face value the reports China has released about the development of COVID-19. At worst, some observers will allow that these numbers might not be “verifiable,” but this is generally expressed as sort of a minor issue. Now, I realize that there are potentially partisan reasons to slant interpretations one way or t’other, but I don’t think this is a case of ‘ a little bit of wiggle room’ in the figures.

Just for fun, I decided to take what China has said about its experience with COVID-19 and use what we know about the development of the virus in the US to project what the true cases probably are. Obviously, as with any model, there are wide error bars…but it’s simply implausible that China is experiencing anything like what they claim to be experiencing. Here’s the math of it.

We know from the US and the rest of the world that the virus cases grow around 33% per day until they downshift “at some point.” See for example this outstanding chart from the Financial Times from last week.

That point of shifting is somewhat speculative and probably depends on a lot of things. But for our model we’ll assume that the virus grows in China at 33% per day and downshifts to a 2% growth rate (which is roughly what South Korea’s growth rate in new cases is now, so we are being very generous) at the point China claimed that the growth rate of new cases was starting to decline, roughly around February 11.

We also can infer from US numbers that the death rate is around 5%, with deaths taking about 5 days and recoveries about 17 days. So, we look backwards about 5 days to see the number of open cases, and realize about 5% of those will die. Similarly, we look backwards about 17 days and realize that about 95% of those cases eventually recover (recoveries take longer because they tend to be cases that were caught earlier, plus a ‘recovery’ is not defined until we get two negative tests while a death is pretty clear). There are a number of combinations of period lookbacks/mortality that work, but that’s about the best case. We can’t get a model that is consistent with the number of deaths we have in the US with a longer resolution time unless the fatality rate is much higher; if the fatality rate is lower than 5% it implies that cases resolve even faster than 5 days after detection which seems unlikely. Globally, the ratio of deaths to (deaths + recoveries) is about 16% (see here under “closed cases” for a source of that data), so 5% is quite conservative. Similarly, we can’t have the low number of recoveries we have unless recoveries take a lot longer than deaths to resolve, or the recovery rate is a lot lower (death rate is a lot higher) than 5%.

And that 5% is with the US having some advance warning, and outstanding medicine. I don’t have any reason to believe it would be lower in China. So those are the parameters I’m going to use. 33% growth rate of infections downshifting to 2%, 5% mortality rate, with resolutions happening in 5 days for death and 17 days for recovery.

The first case in China dates from 12/16/19.

Growing at the aforementioned rates, from 1 case on 12/16/19, China ought to have been around 72 cases by end of December. They admitted to 27. Since at low numbers the growth rate has a lot to do with idiosyncratic details of the particular cases, we will re-set to 27 on 12/31, to be generous. But recognize that there is some reason to think China was low by a factor of 2-3, two weeks in.

Growing at the aforementioned rates, China ought to have been around 14,300 by January 22, when they told the WHO they had 547 cases. (From here on, all of the data comes from Bloomberg whose numbers differ slightly, but insignificantly, from the Johns Hopkins data). So they’re off by a factor of about 25 from what we would expect, based on the experienced growth rates of other countries.

Around February 11, China claims to have had 44,653 cases. By our growth rates, it should have been more like 4.3 million. China must have had spectacular medicine! (even if we grow the 547 number from Jan 22, they would have been at 125k by Feb 11). So it looks like by Feb 11 China was already off by a factor of between 3 and 100. The 3 requires us to believe that they were being completely honest on Jan 22 when they said 547.

So let’s assume the new cases downshift on Feb 11 to only 2% growth. Even starting from their 44,653, we would see 109k cases by now (March 27) if growth rates were only 2%. Literally the only way to get to China’s figures is to say that the transmission rate was never very high, despite the widespread travel around the Chinese New Year and the fact that as the origin of the pandemic is it reasonable to conclude that their recognition of the danger of this disease would take a little longer. And if all of those things are true, then the aggressively autocratic crackdown seems really over the top, given the vanishingly-small prevalence of the disease in a country of 1.4 billion people.

Going back to our original trends of 33%, downshifting to 2%, and with the death and recovery rates I am estimating: I think the actual number of cases in China is more like 10 million, with 465k dead and 7 million recovered. Honestly, it’s hard to explain why their traffic and power usage is recovering so slowly if only 0.006% of the country ever contracted the virus and an even tinier fraction died. The economy should have immediately sprung almost fully back when the quarantine was lifted. Most people in China wouldn’t even know anyone who had been infected…only 1 person in 17,000 ever got the disease. The numbers would just be too small to notice.

We don’t just have to believe China is inherently deceitful…we just need to believe that the country doesn’t have miraculous medical powers. And we need to believe in math. There are other reasons why the numbers could be technically accurate, and yet not illuminating. It might be the case that China simply isn’t testing people very aggressively, so that it is the case that only 81,340 people have tested positive, and only 3,292 of those have died. In that case their numbers would be accurate but not necessarily the whole truth since the country would also be having the small issue of seeing a bunch of other bodies piling up for “unknown” reasons. That inconvenient fact would eventually become hard to not notice, which is one reason why it might make sense to expel external journalists…

Categories: Uncategorized
  1. Eric
    March 28, 2020 at 1:14 pm

    “And if all of those things are true, then the aggressively autocratic crackdown seems really over the top”

    Given that china exhibits really over the top aggressive autocratic crackdowns on just about everything, why would you rule this out?

    • March 28, 2020 at 3:46 pm

      welding people in their homes? Dragging people out of their homes? I don’t know, that doesn’t seem out of the range of policies I would expect a Communist regime to deploy, but it does seem out of the range of policies I would expect for something which, according to their numbers, has only killed about 2% as many people as the flu kills every year there.

      • eric
        March 28, 2020 at 4:09 pm

        I dunno. They have rounded up circa 1,000,000 muslims in concentration camps because of a fear of terrorism that couldn’t possibly kill anywhere near any of those numbers. But also, at best that shows that back in January they were extremely worried about the virus. Maybe now they have figured out that it kills about as many as the flu and they are just letting it do its thing like we always do with the flu.

      • March 30, 2020 at 10:49 am

        Maybe! Of course the evidence is that it kills closer to the experience with SARS, than the flu.

  2. onebir
    March 28, 2020 at 2:19 pm

    “We know from the US and the rest of the world that the virus cases grow around 33% per day until they downshift “at some point.” See for example this outstanding chart from the Financial Times from last week.”

    But we don’t, because of counter examples including HK, Singapore, Japan & South Korea – and perhaps other countries, which haven’t crossed the 100 case threshold to make it on to this chart, despite some imported cases.

    Jeremy Howard provides a reasonably compelling hypothesis for why (widespread use of face masks) here:

    I’ve also seen Chinese company COVID-19 policies mandating (amongst other things) in -office mask use, disinfection on arrival at office, segregation of any one who shows a temperature, taking lunch separately at workers desks. I imagine these are based on country wide guidelines. So it seems reasonable to believe their social distancing – along with that in other countries who got a rude awakening during the SARs epidemic – is significantly more effective than in most other countries.

    I don’t doubt there’s a larger ongoing problem in China than the stats reflect, and I too would like to get a handle on just how big. But a top down approach based on the assumption that the West sets the gold standard for epidemic control probably isn’t a good way to do that.

    • March 28, 2020 at 3:41 pm

      Interesting that you mentioned South Korea because to get a handle on the difference i also analyzed South Korea. They evidence a lower mortality rate (around 2%), but interestingly with a longer time to resolution of “recovered” (24 days) and death (12 days) to fit the model. And the compounded growth rate is lower as you note, around 22%. What I wrote in an email about this on Friday is exactly to your point: “Transmission rate is lower, for reasons no one has been able to figure out. Might have something to do with the fact that it’s de rigeur in Korea (and Japan) to wear masks in lots of public places.)

      It isn’t as widespread a practice, though, in much of China. Then, in the same email, after mentioning the longer resolution time frames, I added “And that says to me that either (a) they’re using a much more sensitive test, so they’re catching very low levels of the virus and therefore it takes longer to declare someone ‘cured’, or (b) they’re testing much earlier, so that they’re recording someone as sick very early, isolating them early, and then it takes much longer to resolve as a result. In either case, the slow growth/transmission rate is tied to the mortality rate given that someone is sick. Unfortunately, that means that if we observe a country with a high transmission rate (e.g. the US, Italy, basically everyone else) then we also expect to see a higher mortality rate.”

      Now, regarding their COVID-19 policies note that those don’t pre-date the epidemic, so they can’t be a reason for lower transmission rates. Note that i don’t doubt that they eventually “bent the curve” with their policies (also, not noted, the fact they were welding entire neighborhoods in their homes), just that it isn’t zero as they claim. And it’s inconceivable that in the midst of nation-wide travel and celebrations in the middle of the pandemic’s spread they managed to only spread the disease to 80,000 people. I suspect instead that they just stopped testing, and never tested very widely.

  3. Jan Robyns
    March 28, 2020 at 2:21 pm


    I agree with China statement but you clearly havent thought about your statistics at all. You are making HUGE errors in your death rate etc. I love your inflation stuff but the above is so cringingly bad I had to stop reading it. Sorry to put it this way.

    Kind Regarda



    • March 28, 2020 at 3:33 pm

      Sorry Jan, but your math is incorrect. I’m using publicly available data showing the death rate. You’re thinking about the CSR, which is not the right way to look at mortality with long-incubation illnesses. Not only have I thought about my statistics, I’ve spent an enormous amount of time doing so over the last couple of weeks. If anything, 5% is too low since the REALIZED death rate among closed cases is more like 16% (which is well below MERS, btw, so not at all strikingly high). Now, if you want to include all people who have been exposed but not positive tests you may be right that it’s lower…but that’s also not a relevant statistic. So, while these figures aren’t “right” (no projections are), they’re the right order of magnitude. Sorry.

  4. Pradeep Jhanjee
    March 28, 2020 at 9:05 pm

    Hi Michael, Many, many, people who study viruses and their progress through populations for a living, are of the opinion that once the dust settles the death rates will be below 1%. There are many assumptions being made around how Covid-19 deaths are being attributed and how many people actual get themselves tested. Kind regards, Pradeep

    • March 30, 2020 at 10:48 am

      Thanks Pradeep. It would be terrific news if this virus had only a small fraction of the lethality of SARS (10%) or MERS (34%). Sadly, many many people who study viruses – including some who think that eventually the death rate will be below 1% (after years, after the development of vaccines and treatments, and after measuring all people who have been exposed rather than all those who have presented themselves for diagnosis) – don’t think that is likely to be the case as there is no real strong reason to think that this is fundamentally a less-dangerous virus. There are also some epidemiologists who think the ultimate fatality rate will be above that of SARS (so far, it is).

      Some of the people who say 1% mean that 1% of all those who are exposed, since many of those people won’t get tested because they will have mild symptoms. But while this is an academically interesting answer, it is useless for modeling since we have no way to know how many people have been exposed, contracted the disease but didn’t have symptoms. Obviously, if that number is 5x as high as the ones who have tested positive, but the mortality rate is 1/5 of my current estimate, it doesn’t matter (same number of dead people) except that you can’t model it.

      The good news is that in the US we appear to be already developing successful treatments as that 17 days is already down to about 13 (recoveries have surged) and the death rate has dripped to around 4.5%. But that’s quite apart from the question of China’s experience, which is the concern of this article. Thanks for writing.

  5. onebir
    March 29, 2020 at 4:04 am

    “And it’s inconceivable that in the midst of nation-wide travel and celebrations in the middle of the pandemic’s spread they managed to only spread the disease to 80,000 people. I suspect instead that they just stopped testing, and never tested very widely.”

    Agreed – however for about month after this, everywhere, was in a very strict lockdown. (Including quite minor places – ever heard of Zhaotong? 😉

    This will have curtailed further transmission after the spring festival burst, and allowed most cases to become non-infectious. Let’s say their case count was multiplied by X ( 0 < X <<1 )

    My point is that we don't know X.

    The measures in place – including mandatory mask usage – seem strict enough to strong suppress further spread (as evidenced by that Jeremy Howard video I linked to earlier).

    But X is crucial, and we just don't know its value.. I think the modelling you did can only put a loose upper bound on the size of the remaining Chinese problem; to get a better estimate would take bottom up work (as far as I can see).

  6. onebir
    March 30, 2020 at 5:46 am

    Bottom-uppish highish-frequency work: Financial Times China Economic Activity Index:
    (~ 1/2 way down page)

  1. No trackbacks yet.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.

%d bloggers like this: