COVID-19 ISO Insights

COVID-19 - A First Look at Distorted General Liability Actuarial Patterns

March 1, 2021

By: John Buchanan, Marni Novack, Tim McCarthy

Media reports indicate that the first reported case of COVID-19 in the U.S. was in late January 2020, and the first fatality from COVID-19 in early February. With cases that were on the rise, many states started shutting down in the second half of March and the first week of April. By mid-May, more than half the states remained sheltered, while many others reopened to various degrees. Some states never sheltered. Reopening, pausing, and reversing responses continue to vary dramatically through the first quarter of 2021, along with significantly varying degrees of restrictions and openness of bars/restaurants, schools, hotels and entertainment facilities. There have been more than 27M cases and 500K deaths in the US from COVID-19.

Over the past year that the US has been affected by the pandemic, much of the economy has been shuttered for some period of time leading to significant impacts on the insurance industry. During our investigations of the impacts, we have found that some insurance markets have had increased actual loss activity compared to the expected levels, while many others have had sometimes significant reductions. For this brief overview, we looked at this impact on General Liability for the first three quarters of 2020 to determine how overall loss levels appeared to have been impacted by the pandemic, compared to what we may have expected in a no pandemic world.

It is very important to note that while loss activity has been reduced thus far on an overall basis due to significant drops in real exposure, there is expected to be a similar impact, though not necessarily to the same degree, from a premium perspective. These potentially significant reductions may result from future premium audits for General Liability policies that were in effect during the period of reduced economic activity in the first three quarters of 2020. Exposure bases which could be subject to audit represent a large percentage of the ISO General Liability classifications. We may not fully know the impact until well into 2021.

While we focus a good deal in the following sections on what has happened so far related to Covid, it also should be remembered that insurance involves estimating the cost of providing insurance coverage for a future period. Unlike many industries, the cost of the good/service provided is not completely known at policy inception with insurance. Historical insurance experience may tell a good deal, but shocks to the economy, and major short-term changes in risk exposure, must be viewed carefully.

The economy during a period where Covid vaccines are readily available could be expected to be much more like the pre-Covid period. As a result, the reduced claim activity seen during the first nine months of 2020 may not be fully indicative of what will occur for insurance policies that are being issued in the second half of 2021. There will potentially be a surge in demand for certain activities when life gets closer to normal. As an example, friends who enjoy going out to dinner before a concert may be doing that more often than in the past when they finally feel comfortable to start going out to dinner in public again. These issues will pose a challenge to the insurance industry. Having the appropriate historical data, as well as working to understand the future world where life returns to some degree of normalcy, will help solve those major distorting challenges.

A. Overall reported loss impact due to COVID-19

Based on the ISO statistical database, the total actual number of General Liability incurred claims for the first three quarters of 2020 are about 34% lower than expected. Total dollars of incurred indemnity are about 26% lower than expected. Using the most recent three years that were not impacted by COVID (2017-2019) as a base, and after adjusting for seasonality, we expected about 72,000 claims to be reported. However, only about 48,000 were actually reported. Additionally, we expected $615M of incurred indemnity, while only $453M was reported. Across all General Liability classes, claims fell about 20% in the first quarter, dropped sharply by 40% in the second quarter, and then rebounded a bit to a fall of 25% in the third quarter compared to what was expected.

Figure 1

As can be seen in Figure 1, the ratio of actual vs. expected total GL claim counts for the 2nd quarter 2020 is 53.6% (Covid Retention Ratio). Within GL, the four most affected class groups were Entertainment and Recreation, Hotels and Motels, Restaurants and Bars, and Schools, which fell between 70% to 85% compared to expected. Overall, for these four class groups, the reductions in the first, second, and third quarters were 22%, 74%, and 48% respectively. These reductions are equivalent to an overall drop of 50% year to date, compared to what would have been expected. It is intuitive that these classes are the ones that fell the most, as they were the activities that were dramatically curtailed. To the extent schools and restaurants were open through the 3rd quarter of 2020, operations had also significantly changed in many jurisdictions. An increase in delivery/take-out orders and the prevalence of remote learning had reduced the expected risk for Premises/Operations incidents compared to pre-COVID periods. While these four categories had fewer reported losses, we also saw that average severity for these class groups was about 28% higher than expected, so the losses that did occur seemed to be larger. On the other hand, we found some other GL class groups, such as Auto/Transportation, Emergency/Government Services, and Food Processing had higher than expected losses for some time periods in 2020. It is also intuitive that some of these class groups would have worse experience under the pandemic, as heightened activity would have been expected in these segments due to COVID.

B. Deeper dive into impact by type, size, shelter status, etc.

Within the overall General Liability data, we also looked at some other splits of the data by class group. One such split was by claim size. When looking at larger claim sizes, those greater than or equal to 10k and 25k, we observed that the drop-off in claims was not as large as for the smaller claims. As can be seen in Figure 2, the drop-off for total General Liability in the third quarter of 2020 for claims greater than or equal to 25k was about 15% while for all claim sizes, it was about 30%, with Food Processing, Hospitals and Nursing Homes, Offices and Banks, and Residential actually seeing more claims above 25k than expected in the third quarter.

Figure 2

Another split reviewed was by cause of loss between Bodily Injury (BI) and Property Damage (PD). In Figure 3, we observe that BI claims and dollar amounts generally fell more than that of PD, with BI incurred claim counts falling by about 45% year to date versus a drop-off of about 20% for PD. At this point, it is not clear if this change is due to the type of incidents occurring in a world significantly impacted by COVID or other reasons such as the reluctance of many people to go to a doctor or hospital in the first three quarters of 2020 unless they believe they had COVID-19 or suspected they were suffering from life threatening conditions. This may have delayed the filing of some Bodily Injury claims.

Figure 3

A final split reviewed was the state shelter status in the second quarter of 2020. As seen in Figure 4, we observed that there was a modest impact, with the never sheltered states having somewhat less of a drop-off in claims than all states combined in each of the first three quarters of 2020. A quartile ranking by second quarter openness score also indicated an impact on the reduction in claims.

Figure 4

As expected, states that never sheltered have a smaller reduction in claim counts (29.4% vs. 46.4%) during the highly sheltered 2nd quarter of 2020. The level of business activity would generally have been higher in the “never sheltered” states compared to other jurisdictions with strict sheltering or openness standards during this time period, and as a result claim activity would have persisted at higher level in those states that were never sheltered. This noted impact persisted into the 3rd quarter of 2020.

C. Overall settlement impact

We have been focused on incurred claim counts and losses to this point, but total General Liability settlements for incidents that occurred during this period, for both claim counts and dollars were also affected. The impact for settlements does not appear to be by as much as reported losses. As can be seen in Figure 5, we observed that the settled claim counts dropped by about 26% from what would have been expected, while the settled dollars fell by around 12% below expected year to date.

Figure 5

Since we have a more muted impact from Covid due to settlements, it is possible that companies responded quickly to settling claims in a remote environment. That reaction may have offset any delays in case resolution due to the court system pauses. Perhaps with the reduction in new claims being established, there were more potential resources to settle the claims that were setup.

Overall, the quarterly pattern that we have observed mimics the Covid infection case counts. In particular, the heavy first wave 2nd quarter 2020 Covid cases had the most impact on the heavily shuttered class groups of Entertainment and Recreation, Hotels and Motels, Restaurants and Bars, and Schools. We also did note a higher than expected reduction in claims in the 1st quarter prior to mid-March. Potentially, even though states had not yet mandated shelter rules until later that month, behaviors of worried consumers and travelers was already potentially impacting economic activity and resulting GL case counts. The smaller reduction in claims during the 3rd quarter is in keeping with the pause after the 2nd wave of Covid cases in July, dipping down to a temporary low in mid-September.

The question remains open as to the impact on GL claim data in the 4th quarter of 2020 and beyond due to the massive 3rd wave of cases and fatalities spiking upwards in November and continuing up 5-fold through January. However, the engine of the economy and resulting claim activity was likely less impacted given that society has become better equipped to handle this increase in cases, than it was in the immediate shutdown phase of the second quarter. We will be closely monitoring the 4th quarter 2020 statistics and beyond to evaluate the remaining distorting effect of the pandemic on the claim emergence.

D. Why is tracking these patterns important?

Pricing and reserving actuaries rely heavily on a consistent set of historical experience and statistics, to help make estimates of the future. The classic description is that actuaries look at the rear-view mirror, along with attempting to measure any known distorting factors, to predict what is going to happen in the road ahead. When there are large catastrophic events or loss disturbances such as 9/11 or the Great Recession, actuaries must make significant assumptions to offset these distortions. The distorting impact of Covid will likely cause sometimes major temporary reductions in loss activity, and in some cases increase loss activity. Proper adjustments will require extremely robust benchmarking statistics, and modeling capabilities, to adjust for those impacts.

Besides the normal reporting and settlement patterns affected, new and increasingly sophisticated approaches will need to be developed and mastered. Ranging from significant impacts on accident or exposure year trend calculations, to the impacts on size-of-loss distributions, the amount of analysis and necessary adjustments to the data will undoubtedly be extensive.

Both in the most recent phase of extremely high Covid case counts and resulting diminished economic activity, but also now in the current early 2021 phase where case counts appear to be significantly reducing and pressures are mounting for resuming economic activity under various recovery shapes, then GL and other insurance losses are destined to snap (or ooze) back to prior expected levels. Like with the Great Recession, the impacts and return towards normalcy may take many quarters if not years to assess, even with the best of data.

For more information on ISO’s analysis capabilities regarding COVID-19, please contact one of the authors or contact us here with your inquiry.

Related article
Information Emergence Lag and Wrong Signaling – Going Viral

Authors
John W. Buchanan, FCAS, MAAA, is Managing Director, Excess & Reinsurance at Verisk
Marni Novack, FCAS, MAAA is Actuarial Associate, Excess and Reinsurance at Verisk
Tim McCarthy ACAS, MAAA is Actuarial Director, Commercial Liability at Verisk