India’s top-8 life insurance coverage corporations settled 1.45 million death-claims in Apr-Sept 2021 throughout second covid wave,

Step-1: Historic trend

From FY14 to FY19, death-claims settled by these eight life insurers grew 6 per cent a year to reach 1.6 million in FY19. Before covid hit, first 9-months of FY20 saw death-claims grow 14 per cent over same period a year ago. Had nationwide lockdown not impacted final 3-months of FY20, death-claims were on track to cross 1.8 million in FY20.

Unfortunately, nationwide lockdown in last 2-weeks of FY20 impacted settlement of death claims, causing January-March 2020 death claims to fall sharply from prior 14 per cent growth to negative growth.

Actual FY20 death-claims ended up at 1.7 million due to lockdown disruption. As these disrupted claims were most likely spilled over to FY21, both FY20 and FY21 death-claims were distorted by covid/lockdown.

Since baseline is supposed to incorporate pre-covid trend, this analysis ignores covid/lockdown-distortion. Pre-covid trend corresponds to ~1.8 million death-claims in FY20 and FY14-20 growth rate of 7 per cent a year.

Step-2: Baseline for April-September 2021

Had above trend continued, FY22 was likely to witness settlement of nearly 2.1 million death-claims by these eight insurers. Since our period of interest is only first half of FY22, baseline death-claims for April-September 2021 is little over 1 million death-claims (1.04 million to be precise, though this isn’t an analysis amenable to decimal point precision).

Step-3: Calculation of ‘excess’ by comparing actuals to baseline

During April-September 2021, India’s eight leading insurers settled 1.45 million death-claims. This is 40 per cent above the baseline that we arrived at after incorporating pre-covid data and trend.

Unlike in FY21, when death-claims were in line with historic trend, H1 of FY22 saw a dramatic deviation from historic trend. This is consistent with what we know about severity of Delta wave and analyses of partial state-level death registration data.

What does this imply for excess deaths during Delta wave?

There is no assurance that life insurance death-claims are perfectly representative of India’s all-cause mortality. However, they are a relevant surrogate. An extrapolation from death-claims to registered-deaths is best viewed as an if-then statement.

Over the 6-month period from April-September 2021 (H1FY22), India was expected to witness ~4 million registered deaths had historic trends continued. If India’s registered deaths were also 40 per cent above trend, then India likely witnessed 1.6 million excess deaths over April-September 2021 during second covid wave.

Note that not all excess deaths are covid deaths, as other causes of mortality can also be impacted due to severe pressure on healthcare system. Mechanically dividing 1.6 million by reported covid deaths to arrive at an ‘undercount factor’ is meaningless.

Why is this number well below recently published estimates of over 3 million excess deaths?

Because, and there’s no kind way to say it, 3+ million estimate is plain wrong. Without repeating my entire critique of such papers, they suffer from two major flaws.

First, they do NOT incorporate trend. A recent paper, ironically published in Science magazine, simply subtracts 2020/21 deaths from 2018-19 average without incorporating trend growth. In states that are considered, trend growth is often higher than India’s 3 per cent. Mechanically applying this method to 2019 data yields nearly 1 million excess deaths well before covid, clearly indicating a flawed method.

Second, they cherry pick data in terms of custom-selected months within 2020 or 2021. Since lockdown disrupted death registrations, some of these spilt over to subsequent months. Most analyses focus their analysis on June-September or July-Oct 2020, although covid first wave started in March. This counts deferred registrations as excess deaths. Above flaws in method lead to overstated, meaningless estimates. Innumeracy and bias make most estimates fodder for sensationalist headlines, not serious analysis.

What could be the errors in my analysis?

Most obvious flaw is that deaths and death-claims need not follow the same trend, as segment covered by insurance may not be representative of India’s population.

A second flaw could be seasonality, if any. This analysis assumes April-September baseline that’s 50 per cent of full-year baseline. If April-September account for less than 50 per cent of full-year deaths/death-claims (long run data is unavailable to reliably figure this), Delta wave death-claims could be more than 40 per cent above-trend.

Lastly, deaths and death-claims can be mean-reverting. A phase that saw above-trend deaths could be followed by a below-trend phase. As an illustration Tamil Nadu’s online CRS database shows lower deaths in January-2022 than any of prior four Januarys, despite Omicron. In this specific analysis, a final analysis of full-year FY22 could show a divergent trend from first-half of FY22.

As an alternate scenario, we could arrive at a more conservative baseline by starting with FY19 death-claims (for which no covid/lockdown-adjustment is required) and growing it at 6 per cent rather than 7 per cent. In that scenario, April-September 2021 death-claims are ~50 per cent above trend, implying ~2 million ‘excess’ deaths.

Instead of viewing 1.6 million ‘excess’ deaths as a point estimate, it could be viewed as a range, say 1.5 to 2 million although I am biased towards lower end of that range.

Insurance death-claims are 40 per cent above-trend for April-September 2021, a period that covers India’s second covid (Delta) wave. If deaths followed a similar trend, Delta wave likely resulted in 1.6 million above-trend or ‘excess’ deaths.

Note that not all above-trend deaths are covid deaths, given healthcare system strain exacerbating other causes of death. As this is an imprecise estimate, it is best to treat 1.6 million as a figure with an error-band around it. Despite that, this is well below recently popularized estimates of >3 million. This is unsurprising given methodological flaws behind such sensationalist claims.

This piece was originally published on Anand Sridharan’s Substack and has been republished here with permission.