Covid-19: Measuring the frequency impact

Dynamics specific to P&C have been greatly changed by the coronavirus outbreak. For all the focus on the potential tail risks, the likeliest outcome is a period of benign frequency that allows carriers to catch up to trend, altering the near-term industry paradigm. This doesn’t change our “worst of all worlds” view emerging from the macro backdrop over the medium term, but it does lessen the pressure on underlying loss trends in the nearer term (despite some near-term cat-like losses in some lines).

Importantly, we expect the low frequency to be recognized over time and not in initial picks, lowering the risk of an offsetting regulatory action. We think this will manifest comparably to the post-9/11 market when pricing doubled but loss picks remained close to stable, with excess reserves released over a long period.

In short, we have reached the “repair” stage of the underwriting cycle. But instead of conducting this through the typical hardening pricing dynamics, carriers are being given a get-out-of-jail-free card for having recognized too much of the benefit of a post-recession deflationary period into their pricing and reserving through the later part of the economic cycle from 2014 through 2019. So long as they can keep their loss picks elevated out of “conservatism”, they may even get some continued modest pricing benefits too, limited perhaps only by insureds’ ability to pay.

We expect this frequency benefit to manifest predominantly in auto and commercial auto, but also in workers' comp (in spite of prevailing conventional wisdom) and across several other liability lines. It’s as simple as saying less “stuff” happening means less exposure, less risk and fewer claims.

Auto: With streets so empty, auto accidents have already declined substantially

Lockdowns resulting from Covid-19 will have a profound effect on auto accidents, specifically in personal auto as many are working from home, furloughed or unemployed, and non-essential business is on pause. This is likely why auto-exposed names like Progressive have outperformed the market and their peers.

While the impact of Covid-19 on auto frequency may seem obvious, we thought it’d be interesting to try to quantify and visualize the impact, in part because the sheer velocity of change makes this period unlike any other historical analogy we might draw upon.

The easiest source of visualization comes through Google’s traffic data. We’ve taken the usual Monday 16:00 traffic from Los Angeles, Chicago and New York and put it against current results. The figures show what many are likely thinking – little to no traffic with green routes no matter where one looks.

In addition, given the unprecedented nature, we also wanted to try to approximately quantify the frequency benefit rather than simply observe it directionally. We have attempted to do this via access to real-time accident data that is available through some state DOTs or similar local government entities.

Our sample of states, which makes up 23% of 2019 industry DWP, showed that auto accident frequency was down by 31-67%. However, importantly, this sample included a range of states, including many that have only recently begun to experience the impact of social distancing measures through school and non-essential business closures or stricter shelter-in-place restrictions. In fact, even those states with more severe drops only had such measures in place for half the month, likely indicating that April will be even more benign. Note, we will get our first data point from an insurer on April 15 when Progressive reports March earnings.

Note, personal auto makes up 36% of industry premiums.

Other sources of data journalism using cellphone data have shown that the average distance traveled per journey has also declined precipitously. For example, in Daytona Beach, Florida – an area with no travel restrictions in place for most of the month – the average journey more than halved from 4.4 miles to 1.9 miles. In Seattle, where stricter policies are in place, the average journey fell from 3.8 miles to just 68 feet (i.e. likely by foot). As a general rule of thumb, this also should reduce severity due to a reduction in long-distance highway travel and a greater percentage of collisions being at lower speeds.

You can find a useful data resource here for more examples of reduced mobility and non-essential travel, which is broadly consistent in direction and magnitude of our survey of actual accidents. This data also shows that most of the impact has been during the second half of the month only.

We’d note that there may be some risks that are pushed further out in spite of the near-term benefit. It is reasonable to assume there could be a change in consumer behavior coming out of the crisis, either short-lived or else perhaps for a reasonable stretch depending on the impact on consumer psychology (determined by the length and depth of public health outcomes).

For example, there may be changes around willingness to use public transportation and the adoption of substitutes like TNCs or personal vehicles. This could add frequency risk if not anticipated in pricing (e.g. assuming mean reversion may undershoot it). This could be particularly problematic if regulators push too far and too long on bleeding frequency benefits into pricing.

Like personal auto, commercial auto (6.4% industry DPW) should also benefit from there being fewer cars on the road. Ohio discloses the number of accidents involving commercial vehicles, and in March there was a 44.7% decline in commercial-related auto accidents YoY.

Going further, truck tonnage will likely decline due to the increasing likelihood of a recession. The longer social distancing is required, the less likely we are to see a V-shaped recovery.

Though there is likely to be more offsets versus prior recessions due to consumer deliveries (more online retail generally, less ability to shop at brick-and-mortar stores), the level of commercial shut-down is unprecedented and is likely to have significant impact on volumes across commercial entities. Truck tonnage has risen substantially following a drop during the 2008 financial crisis, but will likely fall if economic conditions remain bleak. This will likely reduce some of the pressures that have led to worsening conditions for insurers, including less experienced drivers, quality of fleet in service, hours in service etc.

Recall, commercial auto has been a problem spot for a decade for commercial insurers, and a persistent source of loss ratio underperformance and adverse reserve development. This current climate could help slow the treadmill and finally allow carriers to catch up to trend.

We’d also note that SEC filings show that legendary insurance analyst and investor VJ Dowling recently disclosed a 5.6% stake in insurer Protective Insurance, with ~50% DPW in commercial auto. Mr Dowling has made his name in no small part by his ability to recognize a tipping point when he sees one.

Covid-19 will impact lines of business beyond auto, and not always favorably

As is well understood, Covid-19’s impact on employment should also affect lines like workers’ comp, where significant layoffs will translate to lost premium through audits.

For the week ended March 27, a record 6.6 million Americans filed jobless claims, on top of the previous week’s 3.3 million. For context, this is more jobs lost than through the entire great recession in just two weeks.

Unfortunately, the amount reported to date is still well below estimates by economists at the Fed’s St Louis district which project total employment reductions of up to 47 million, or a 32.1% unemployment rate, based on how bad things could get.

Conventional wisdom suggests workers’ comp is negatively affected by recessionary conditions and incentives for employees to shift health costs to their employer. The lost earned premium can also reduce fixed-cost operating leverage and pressure the expense ratio.

We expect this time to be different for three reasons: (a) fewer uninsured Americans versus prior recessions, (b) rapidity of closure of the economy and access to employers’ facilities, and most importantly, (c) the sheer number of employees working remotely from their own homes limits the ability for workplace accidents.

Like pretty much every line across P&C, we expect there will still be a lot of noise around attempts to shift losses to insurers (e.g. “I caught corona through my employment”) that will have some LAE impact, but ultimately we expect the frequency benefit to dominate.

For similar reasons, we expect plenty of frequency good news across other liability lines including GL (fewer slips and falls etc, excess auto exposure benefits). In general, we would also expect a lessening of adverse severity trends over time if the recessionary conditions persist as the late-cycle “deep pockets” psychological bias lessens and eases adverse tort trends.

Finally, we’d like to make two points on these views.

The first is that, in some ways, this is an unverifiable prediction without a pretty long time horizon. We’ll likely not be proven right by reported results, nor are there many sources of evidence that will disprove the hypothesis.

Insurance companies have every conceivable incentive to not recognize good news for as long as possible, and to recognize every ounce of bad news quickly. We expect them to keep their customers and distribution focused on their losses from Covid – real or “potential” – as well as their trailing lack of profitability, investment income pressures and balance sheet hits from investments.

With policymakers looking for deep pockets to use to fill the gaping economic hole, there is a monumental incentive to act like an industry in crisis, one straw away from breaking the camel’s back.

The second is the size of the potential windfall. For example, personal and commercial auto incurred losses combined run at north of $200bn annually. Let’s cut this by 50% for three months = a potential $25bn gain.

Of course, this is back-of-the-envelope stuff, and it then becomes another issue altogether over how much is protected from effective government seizure, and how and when it is recognized in earnings.

Our point is merely to illustrate that, for all the fixation on tail losses and cat-like potential from some specialty lines, there’s plenty of scope for positive offsets too. They’re just not likely to be talked about any time soon.