Insurer death-claims were in line with prior trend during first covid wave

My last covid-related essay. And a note of caution on the charlatans who populate this field.

Summary

This study analyses death-claim data across nine life-insurers, accounting for 95% of India’s insurance death-claims. Period considered is FY21 (April 2020 to March 2021), which squarely overlaps with India’s first covid wave. While an imperfect surrogate, death-claims constitute a large, relevant, robust dataset that can shed light on how all-cause mortality trended. This study finds that insurer death-claims were in line with prior trend during first wave.

Introduction

In India, it is challenging to reliably estimate excess deaths during covid period, due to delayed release of all-cause mortality data. As nationwide CRS (Civil Registration System) data is only available till 2019, official mortality statistics are lacking for 2020 and 2021. This compels researchers to attempt this problem using either partial or surrogate data.

Partial data is usually an aggregation of state-wise death statistics, not officially released, but sourced by media from CRS of respective state governments. Such data is available, with varying degrees of granularity, for over ten states for 2020. While this constitutes universe of all registered deaths within each state, aggregate falls short of pan-India coverage. Further, some states have month-wise data while only an annual summary estimate is available for others. In many states, this data is inconsistent with that in CRS-2019 report, introducing an additional source of error. Extrapolation from few states to pan-India is another challenge.

Surrogate data is usually pan-India, linked to deaths and sourced independent of CRS. NHM (National Health Mission) captures data from hospitals across India. While its purpose is broad-based gathering of health indicators, it does capture some morality data. However, coverage is biased towards government hospitals and rural India, making it unrepresentative. Data capture during periods of disruption is also a problem. Another surrogate is periodic consumer surveys that include questions on whether someone in the family (a nebulous term) died recently. Surveys come with typical problems of sampling, reduced response rate during disruptions and correlating survey results to real-world outcomes. Both surrogates are viewed as less reliable than CRS data.

As with any estimation using sub-optimal data, analysts adopt multiple approaches, hoping that triangulation reduces errors. Typically, analysis of state-level CRS data is primary method. NHM and survey data are more of a consistency check. In such multi-pronged approaches, an additional metric that can boost reliability of inferences is welcome. This paper introduces insurance death-claims as a surrogate metric. It presents the first detailed analysis of insurance data during India’s first covid wave.

Background on Life Insurance and death claims

India’s life insurance industry had 325 million in-force policies, as on December 31 2020. While there is some double-counting (some have more than one policy), this amounts to one policy for every third adult in India. Industry settled over 2 million death claims in 2020, or nearly 25% of 8 million estimated deaths in 2020. This is significant coverage, even if short of comprehensive.

Historically, poor were underrepresented in this coverage. However, that bias has reduced over time. A sizeable part of insurance growth over past decade has been through policies that are issued as part of microfinance or other loans. Government supported insurance schemes such as PMJJBY (Pradan Mantri Jeevan Jyoti Bima Yojana) are focused on covering the poor at subsidized premia and have cumulative coverage of over 100 million lives. Rural and social sector account for nearly half of all leading life insurers business, in terms of lives insured.

While not perfectly representative of India’s overall demographics, life insurance covers a large part of India’s adult population. Insurer death claims are worth studying, as a reasonable though imperfect proxy for all-cause mortality. Death claim data from life insurers is very robust. Each claim is backed by a paper trail, including documentation of cause of death. Audited death-claim numbers are shared every quarter by every insurer as part of statutory public disclosure. Since each claim is meaningful for both parties, involving transfer of owed money, individual datapoints are way less error-prone than a mere database entry or survey response.

In summary, death claims are a reliable, large, pan-India, mortality-linked dataset, whose analysis can provide a directional sense as to how deaths trended during covid period.

Methodology

This study uses number of claims settled, and not claim amount, as the aim is to track number of deaths rather than value of insurance. It uses fiscal years (FY) as its reference periods, to align with insurers’ reporting cycles. It zooms in on FY21 (Fiscal Year 2021, from April 2020 to March 2021), as it squarely overlaps with first wave period. First wave started in March-2020, peaked in September-2020 and petered out by February-2021. While second wave started in March-2021, its effect on deaths was significant only from April-2021. Accounting for three-to-four-week lead-time to register deaths and file claims, FY21 encompasses first wave deaths while excluding second wave deaths.

This study limits itself to top nine insurers, who account for 95% of industry’s death claims over FY14 to FY21. These include LIC of India and top eight private insurers (HDFC Life, ICICI Prudential Life, Kotak Life, Bajaj Allianz Life, Aditya Birla Sun Life, Max Life, Pramerica Life, SBI Life). This allows each data point for each year to be derived from companies’ statutory filings. Further, it is important to adjust for lockdown effect, as nationwide lockdown in end-March 2020 shifted claim settlement from end-FY20 to FY21, distorting both years’ metrics. Going company-wise allows quarterly data to be extracted for this phase. By extracting data from FY14 onwards, covid period can be compared to medium-term trend before covid.

It is crucial to note that death claim growth is higher than growth in registered deaths. As elaborated below, death claims grew 7% a year before covid, while registered deaths grew 4% a year over the same period. Difference is due to rising life insurance penetration, especially driven by PMJJBY, rural push by insurers and social-sector insurance linked to microfinance and other loans. Given this, it is incorrect to directly compare death claim trend to death trend over any period. Correct methodology requires comparing death claims trend in period of interest to prior trend in death claims. Having established deviations from trend, if any, within death claims, linkage to all-cause mortality can be cautiously attempted.

Data shows three distinct phases: pre-covid, lockdown-disruption, first wave

Figure 1 shows growth in insurance death claims over time. First phase is pre-covid, with no distortions from either covid or response to covid. This extends from FY14 to FY19 and over first nine-months of FY20 (9MFY20 or April-December 2019). Growth averaged 6% a year over FY14-19 and 14% during 9MFY20. Second phase is lockdown-disruption, over last quarter of FY20 (Q4FY20 or Jan-March 2020). Sudden collapse in growth from 14% to -4% is seen in Q4FY20. This dramatic swing has nothing to do with covid or deaths. It is the consequence of strict nationwide lockdown imposed during last few weeks of March 2020. Lockdown disrupted claim filing, processing and settlement. Since insurance money is consequential to families, disrupted claims were not dropped but delayed. Claims pertaining to deaths that happened in Q4FY20 saw delayed settlement that spilt over into FY21. That leads to third phase, over FY21, that witnessed dual impact of both covid and spill-over effect of delayed Q4FY20 claims. As a result, FY21 saw elevated growth rate in death claims, at 17%.

Both FY20 and FY21 numbers are distorted due to lockdown and deferment of claim settlement. This distortion has to adjusted for, to understand true trend in death claims during first wave period (i.e. FY21). As is evident, true growth rate during first wave is an amalgam of -4% in Q4FY20 and 17% in FY21. 

Adjusting for lockdown-effect, FY21 death claim growth was in line with prior trend

Had 14% growth of 9MFY20 continued into Q4FY20, an additional 95,000 death claims would have been settled in FY20 itself, instead of getting pushed out into FY21. To account for this effect, this analysis adds 95,000 claims to FY20 and subtracts the same from FY21 to derive ‘adjusted’ death claims for both years. As shown in Figure 2, this restores FY20 growth rate to 14% and shows normalized FY21 growth to be 6% over FY20. FY21 growth of 6% is in line with prior trend growth of 7% (over FY14-FY19 and 9MFY20).

While it is evident that lockdown distortion needs to be corrected, exact quantum of correction is subjective. What would Q4FY20 have grown at, had covid-lockdown not happened, is an unknowable counterfactual. While this study views continuation of 9MFY20 growth as most likely scenario, it is worth doing a sensitivity analysis. Assume that Q4FY20 did not continue with 14% growth, but grew at a slower 7-8% (in line with pre-covid trend or FY19 growth). In that scenario, FY20 growth reduces to 12% and FY21 growth increases to 10%. While 10% is above 7% pre-covid trend, it is not a complete outlier. Any real-world dataset comes with natural fluctuations around a trend. Unless fluctuations are sizable, its better to view them as part of a likely range of outcomes rather than mechanically calculate ‘excess’ for every blip.

Across a range of reasonable assumptions for lockdown-effect, FY21 growth in death claims is roughly in line with prior trend. Growth was higher in two, if not three, of the four years that preceded covid-impacted FY21. Trend in death claims is not suggestive of mortality running above trend during first covid wave.

Death claims classified as covid claims

Death claim procedure entails paperwork mentioning cause of death. Data is available on death claims where ‘COVID-19 and related complications’ was listed as cause of death. According to RBI’s financial stability report, life insurers settled 21,854 such claims in FY21. This amounts to covid as reported cause of death in 1.1% of all FY21 death claims. In comparison, India’s 163,000 reported covid deaths as on March 31, 2021 accounted for 2% of all deaths. This could be interpreted as evidence of reliable covid death reporting. However, difference of nearly 2x feels too large to be reliable. It is possible that insurance death claims haven’t comprehensively captured cause of death data, especially since claims are settled irrespective of which ailment caused death. That said, it is noteworthy that covid constituted a small fraction of both death claims and deaths during FY21.

Divergent trends at sub-segment level

While total death claims settled is the metric that matters, there is a divergence between trends in LIC and private insurers. This is large enough to deserve mention, although exact reasons behind the same are unclear. From FY14 till 9MFY20 (pre-covid phase), LIC’s death claims grew -1% a year, while private insurer claims grew 27% a year. After adjusting for lockdown effect, FY21 showed a sharp deviation from trend for both categories, in opposite directions. LIC’s death claim growth in FY21 rose sharply to 16%. Private insurer claim growth fell sharply to -6% in FY21.

LIC and private insurers have somewhat differing segment focus. LIC dominates a segment called ‘individual life’ policies, where LIC sells policies directly to consumers, one at a time. Private insurers dominate a segment called ‘group life’ policies, where policies are sold to consumers under a broader umbrella. It could be through tie-up with microfinance lenders, bundled with loans, or as part of government’s PMJJBY scheme. While it is possible that LIC’s clientele is older, seeing higher mortality, there is no evidence to support this conjecture.

While LIC is a large company, it is one company. If above data is looked at company-wise within private insurers, there is wide divergence at individual company level. Four showed higher than trend growth in FY21 and four showed lower than trend. At a single company level, whether LIC or others, idiosyncratic issues impact performance as much as thematic issues. Further LIC is coming off two years of below-trend claims. Average claims over FY19-20 are 4% lower than that over FY14-18. LIC’s FY21 adjusted claims are only 7% higher than its FY17 claims.

Other evidence shows high growth within segments that are less likely to see covid impact. PMJJBY policies are limited to those aged between 18 and 50. PMJJBY claims grew 32% in FY21. Even adjusting for lockdown distortion, this is higher growth than either industry or LIC in a younger segment less likely to be covid-impacted. Excluding PMJJBY claims reduces industry growth by nearly 2% in FY21.

While there are segment-level divergences in either direction, causality for the same is unclear. In other types of data as well (e.g. state-CRS, NHM, survey), granular inconsistencies abound even as aggregate trend is relatively steady. As an illustration, districts show wild swings even as state trends are stable. This study simply logs internal divergences in the spirit of transparency. As aim is to study all death claims, not some, inferences are limited to aggregate data.

Conclusion

The big question analysts are trying to answer is ‘How did aggregate deaths trend in a pandemic year’. While precise answer will have to await publication of CRS-2020 report, analysts have attempted provisional answers using partial data and surrogate metrics. One such metric, hitherto unresearched in Indian context, is death claims across life insurers. This is a large, robust and relevant dataset. This paper presents the first systematic study of insurer death claims during FY21, which overlaps with India’s first covid wave.

This study finds that death claims, across nine life insurance firms that account for 95% of all claims, grew 6% in FY21. This is in line with prior trend growth of 7%. While there is some subjectivity in adjusting for lockdown disruption, sensitivity analysis shows FY21 to be roughly in line with historic trend under a range of reasonable assumptions. There is divergence between trends at LIC and private insurers, although causes for the same are unclear.

In countries that witnessed severe covid in 2020, all-cause mortality witnessed a significant deviation from prior trend. With India having less developed registration systems and more noisy data, a key determinant of first wave severity is whether such large deviation was seen, across relevant metrics. To that extent, insurance death claim data is not suggestive of above-trend mortality during India’s first covid wave.

Postscript

This is my last essay on covid data analysis. Even this was to complete an analysis I had partially presented earlier. While I enjoyed the intellectual challenge of applying my day-job skills in a different domain, I did not enjoy the company. So, I’m out.

When it comes to excess death articles, analysts are from fields out of touch with reality, without skin in the game. They are far from excellent in any domain but claim to straddle many. Innumeracy, unreliable data and shoddy methods abound. Motivations are suspect, with a bias for sensationalism over soundness. Maximizing excess deaths gets headlines, interviews and quotes in global media. Sober acknowledgement of methodological challenges, error-bands and realistic ranges does not. So-called researchers are no more than TV anchors hiding under the veneer of academic credentials.

Curiosity and accident brought me here. Having been used to working with an excellent investor and outstanding business owners/managers, it felt like going from enlightenment to dark ages. So, I leave you with a note of caution. Be sceptical, even cynical. Don’t take touted numbers at face value. As a rule of thumb, they’re inflated by 100%, which makes them nonsense. If you take two minutes to extract just the data embedded in any article and compare it to history, you’ll do way better than the ‘expert’ who wrote it. This domain is a perfect reminder of the defining message of the enlightenment. Nullius In Verba. Take no one’s word for it.