The official count of covid-19 deaths from the Indian government currently (2 September 2021) stands at about 440 thousand, a rate of about 315 per million people (see Our World in Data).

This is a much lower death rate than many Western countries where the reported deaths are often between 1,000 and 2,000 per million (UK, USA and several European countries are approaching 2,000 with most others seeing at least 1,000 deaths per million.

And some Indian data has been used to illustrate the benefits of ivermectin. So knowing whether the data is reliable might be useful to other countries.

But there have been longstanding doubts about the reliability of the official numbers from India. Recently, several studies have tried to estimate excess deaths in India as a way of estimating covid-19 deaths without having to rely on the official statistics.

And a recent comparison (July 2021) of different ways to estimate excess mortality concluded this (my highlight):

But all estimates suggest that the death toll from the pandemic is likely to be an order of magnitude greater than the official count of 400,000; they also suggest that the first wave was more lethal than is believed.

These estimates suggest total deaths of perhaps 4 million giving a death rate per million of over 3,000 per million, higher than the USA or the UK and possibly doubling the estimated total deaths worldwide.

Is it plausible that India's official numbers could be understating covid-19 deaths by a whole order of magnitude?

  • You're quoting from the abstract of the report. Shouldn't you rather base your question on the actual text of the report? Perhaps they discuss in detail what "all estimates" are, which may be important if one wants to answer your question? And would you accept quotes from the full report as an answer if they present a plausible reason why the official count may be understating the real number of deaths?
    – Schmuddi
    Commented Sep 2, 2021 at 21:23
  • @Schmuddi I quoted a compact version of the broad claim. The point is to question the claim in an answer. If the detail in the full paper stands up to skeptical analysis, then quoting it is fine (but it can hardly be a definitive answer).
    – matt_black
    Commented Sep 3, 2021 at 0:13
  • 1
    @fizz The key problem is that Indian stats are shonky. I don't think they release "official" excess deaths which is why the paper with the clam used other ways to make estimates. The big question if whether other evidence contradicts the official covid reported deaths.
    – matt_black
    Commented Sep 3, 2021 at 0:15
  • India doesn't mandate death certificates in the first place. It's impossible for the Indian data to be correct, the only question is how far off it is. Commented Sep 9, 2021 at 0:03
  • @LorenPechtel Well, that is the question. Or perhaps a part of an answer as, if they don't have reliable cause of death data, then their official covid death data must be unreliable.
    – matt_black
    Commented Sep 9, 2021 at 0:14

2 Answers 2


I'm honestly not sure how one could answer this question. The CGDEV paper you've quoted that from used 3 different methods to estimate the excess deaths in India... because there's no official data on those, due to the inefficiencies of India's data collection system. Even the (centralized) data that's needed for best estimates of excess deaths in India is delayed from release by several years. So the CGDEV paper does pass a "smell test" that at least they tried to "triple check" their way estimating excess mortality.

So, the best I can offer here (short of an in-depth critique of the paper's methodology) is that there's a JAMA Network Open paper that plots excess deaths vs. confirmed Covid fatalities for a bunch of countries (not including India though).

enter image description here

India would definitely be an outlier with a 10:1 ratio as per CGDEV's estimate... but there are a couple of countries with a similar or worse ratio in the JAMA paper, e.g. Kazakhstan has pretty similar 11:1 and there's even a 28:1 (Uzbekistan).

Also, according to a Times of India article, only 1 in 5 deaths in India had a certified cause of death... in 2019.

There's another estimate of the ratio of excess to registered-covid deaths in The Hindu, signed by the newspaper's data team; this analysis came up with a multiple of 8.2. Still, there's wide per-state variation in this study, between 0.42 (??) and 23.8 (Kerala vs Madhya Pradesh).

So, is the CGDEV estimation totally outlandish? Probably not. Does it require further investigation before a definitive conclusion can be drawn? Probably yes.

  • There's a longer paper similar to the JAMA Net. Open one in eLife. Again it doesn't cover India (or China). Commented Sep 8, 2021 at 21:05
  • "Excess death" doesn't necessarily mean "undetected/concealed COVID cases". Excess death could be due to many other diseases being put behind (eventually leading to higher mortality), and many hospitals prioritizing COVID patients over others, and even slowing down planned patients (surgery, not-yet-critical conditions) since most capacity goes to COVID treatment.
    – Igor G
    Commented Sep 10, 2021 at 13:11

The peaks don't line-up temporally in different geographical places. From the charts below, the U.S. cases (and deaths) are rising on a new wave, while India seems to have recently descended from the previous one, and possibly just begun its ascent on another wave unfortunately.

enter image description here enter image description here

  • 11
    How does this, in any way at all, address the question of whether India's official numbers are reliable?
    – matt_black
    Commented Sep 3, 2021 at 13:07
  • @matt_black To compare the "sum" of two timeseries, we need to consider the whole range (-inf, +inf), unless they are aligned, in which case a subrange would suffice. The OP chose to compare these two timeseries (-inf, Sep-2nd) which is invalid if they are not temporally aligned. So no conclusions can be derived from OP's summations; that's what I mean
    – Pradyumna
    Commented Sep 4, 2021 at 11:23
  • 3
    Nothing in the question involves summing or comparing two time series. The question is about whether the cumulative total Indian deaths are reported correctly. In short, I have no idea what your comment is trying to argue.
    – matt_black
    Commented Sep 5, 2021 at 15:35
  • @matt_black "cumulative total Indian deaths from -inf to sep2" = "sum of the daily Indian deaths from -inf to Sep2." What I'm saying is that because the two timeseries are not aligned with each other, you can only compare the "cumulative total Indian deaths from -inf to +inf" with the "cumulative total U.S. deaths from -inf to +inf". A sub-range (-inf, Sep2] has no justification if the two timeseries are not aligned.
    – Pradyumna
    Commented Sep 5, 2021 at 17:31
  • 6
    The key question as asked does not involve comparing India to other countries at all: it is about whether the official Indian reported deaths are accurate. But, also, it is ridiculous to argue we can only compare death totals if we have them over infinite time: that's ludicrous. We can compare deaths across countries but recognise that some are at different stages of the pandemic. But that is a minor factor at this stage of a global pandemic and irrelevant to this question.
    – matt_black
    Commented Sep 6, 2021 at 10:24

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .