Three New Estimates of Deaths in India during the Pandemic

This opinion piece was first published by the Indian Express under the title “Why we need to count the Covid dead.”

Of another existential threat, Bob Dylan accusingly asked, “How many deaths will it take till we know that too many people have died?” This Covid pandemic in India has seen a curious but no less tragic inversion of that sentiment: only a sense that too many died in the second wave has really galvanized efforts to find out the true number of deaths.

The official Covid death count as of end-June 2021 is 400,000. The reality is, of course, catastrophically worse. Unlike in other countries, authoritative excess death estimates based on official data have not been available because government recording of deaths, especially at the center has been lagging. As a result, thus far, and with some exceptions, attempts to capture the somber reality have been inadequate.

Recently, the heroic and tireless efforts of a number of journalists, newspapers (English but especially vernacular), and researchers—too numerous to name—has led to fuller and better cataloguing of mortality during the second wave. And we are now, for the first time, getting data-based estimates of excess deaths at an all-India level. These deaths cannot be strictly attributed to Covid per se, as many of the relevant data sources tell us nothing about cause of death. Rather, what we observe are deaths that happened in the wake of the pandemic in excess of some baseline number of deaths in previous years, referred to as all-cause excess mortality.

In new research, we provide three different estimates of such excess deaths based on three different data sources, each requiring different assumptions and methodologies (table below). We do not favour any one estimate because each has merits and shortcomings discussed below.

Table. Estimates of All-Cause Excess Mortality (millions)

Data Source/methodology Wave 1 (May 2020 -March 2021) Wave 2 (April 2021 –May/June 2021) Total
1.States' Civil Registration Systems (CRS) 2 1.4 3.4
2.Indian sero-prevalence surveys plus international age-specific infection fatality rates 1.5 2.4 4.0
3.Consumer Pyramid Household Survey (CPHS) of CMIE 3.4


Official 0.16 0.24 0.4

Comparing the results side by side, however, provides three key take-aways.

First, unsurprisingly, there is considerable uncertainty within and across estimates. The central estimates range from about 3½ to nearly 5 million with substantial error margins around them. The share of deaths in the first versus second waves also varies across estimates. It is imperative, therefore, that research continue to estimate more precisely Covid-related deaths. It is equally imperative that government aid this effort by making publicly available all the data on sero-surveys and deaths it has generated.

Second, the first wave seems to have been more lethal than is popularly believed. Because it was spread out in time and space, unlike the sudden and concentrated surge of the second wave, mortality in the first wave appeared moderate. But even the CRS data suggest that up to 2 million might have died in that period. In fact, not grasping the scale of the tragedy in real time in the first wave may have bred the collective complacency that led to the horrors of the second wave.

Finally, and perhaps the most critical take-away is that regardless of source and estimate, actual deaths during the Covid pandemic are likely to have been an order of magnitude greater than the official count. True deaths are likely to be in the several millions not hundreds of thousands, making this arguably India’s worst human tragedy since partition and independence.

How did we arrive at these estimates? The first estimate is based on civil registration of deaths in the various states (CRS) and as of this writing, this data is available for 7 states (Andhra Pradesh, Bihar, Chhattisgarh, Karnataka, Kerala, Madhya Pradesh, Tamil Nadu and Uttar Pradesh). This data source has the merit of being based on official data and being relatively timely.

But there are problems because the CRS undercounts deaths even in normal times: in 2019, on average, only 86 percent of deaths were registered compared to estimates from subsequent official surveys and this varies considerably across states. Extrapolating these data for India requires assuming that the pattern of mortality and under-counting for these seven states (accounting for about half of India’s population) are applicable to other states.

If we assume the rate of undercounting did not change during the pandemic, we get an estimate of around 2 million excess deaths in the first wave of the pandemic, and an additional 1.4 million deaths in the second wave. If underreporting of deaths has increased during the pandemic, the toll may be higher. Conversely, if we assume the CRS has suffered no undercounting during the pandemic, however unlikely that may be, our estimate would fall as low as 1 million total excess deaths across both waves. Crucially, CRS data mostly stop in May, and given the lags in recording deaths, it is almost certain that not all second wave deaths have been captured.

A second estimate can be derived from the simple arithmetic that excess deaths are number of infections times by number of deaths per infection, called the infection fatality rate (IFR). The government has generated plausibly reliable data on infection rates. Several sero-prevalence studies have been done for different states and cities but two nationally representative ones are the third sero-survey done in December 2020-January 2021 and a recent WHO-AIIMS survey covering the period, mid-March to early June, 2021. They suggest infection rates of about 25 percent until mid-March (first wave) and about 65 percent by end-June (second wave).

Since mortality estimates for India are less solid, so are IFR estimates. Therefore, we rely on international estimates of IFRs. Recently, the United States’ Center for Disease Control (CDC) disseminated its best estimates of age-specific IFRs which can be combined with Indian demographic characteristics and the age-pattern of Indian infection rates to derive a plausible measure of IFR for India. The underlying assumption here is that the likelihood that any given infected person will die is the same across countries so that the international differences in aggregate IFRs are driven by the age structure of population and the age pattern of infections.

Applying international IFR estimates to India’s demographic structure and seroprevalence rates implies around 1.5 million Covid deaths in the first wave, and an additional 2.4 million deaths in the second wave.

Our third estimate is based on the consumer pyramid household survey (CPHS) conducted by the Center for the Monitoring of India’s Economy (CMIE). The survey asks whether anyone in a household died in the previous four months and the latest period for which this survey was conducted is June 2021, allowing us to capture most but not all of the second wave.

The important caveat is that death estimates from the CPHS pre-Covid do not track closely estimates from other official sources. Perhaps even more important is that the CPHS shows a big and inexplicable spike in mortality in 2019 before Covid. If some of the measurement errors from the CPHS pre-Covid carry over to the Covid period, the reliability of the excess deaths estimates is not assured. The CPHS also does not as yet capture mortality in the entirety of the second wave. Taken at face value, we calculate that CPHS data imply roughly 3.1 million excess deaths in the first wave, and another 1.5 million in the second wave.

All these numbers are far from definitive. A collective understanding of and engagement with all of these mortality data sources and others (most recently, the National Health Mission which has released new data), warts and all, is necessary. As a country, India must confront the scale of tragedy, to draw lessons, and to etch remembrance of it in the nation’s collective consciousness to foster a “Never Again” resolve. The counting—and the attendant accountability—will count not just for today but for the long future.


CGD blog posts reflect the views of the authors, drawing on prior research and experience in their areas of expertise. CGD is a nonpartisan, independent organization and does not take institutional positions.