It's Gonna Lose Money, Jim

I won’t lie, when post-COVID short-term inflation was the big headline across the industry, I really struggled with thinking through the implications on pricing Property (re)insurance. It wasn’t the assumptions about claims inflation really, it was how this sudden spike in inflation was going to affect premiums.

When I was studying for the rate making exam, I started to learn about exposure bases and to what extent certain bases are inflation sensitive. For example, payroll or home insured value would be considered inflation sensitive, but an exposure base like car-years would not be. In theory then, property valuations¹ are a nice exposure base because they are supposed to automatically self adjust for inflationary pressures. This is all well and good, except that theory quickly comes up against reality.

To get our foundations in order, let’s go through a simple thought experiment. Assume your home’s current and accurate valuation is $100,000. If the replacement cost of everything goes up by +10%, then your home’s new valuation should be $110,000. Simple enough. Now, just because your home valuation went up, it doesn’t mean the propensity to loss has somehow changed. For the nit-picky folks out there, I’m ignoring the argument about a one year difference in home age.

Because the propensity to loss hasn’t changed, the theoretical severity distribution of the losses should have just shifted up as well by +10%. Said another way, the expected loss cost should have increased by +10%. Therefore, in the most simplistic rating algorithm of Loss Cost = (valuation)(base rate), my loss cost will track very nicely with inflation.

Where is the problem then? There are a few things going on that throw a wrench into the whole process. In my example, I assumed I somehow knew the exact amount by which my replacement cost was going to increase. If an insurance company writes a hypothetical home with a valuation of $100,000, they are really worried about the valuation at the average loss date during the policy term. To keep it simple, let’s assume the policy is effective 1/1/2025. Ignoring seasonality, my average loss date then is 7/1/2025. I need to have perfect foresight then into 6-months worth of inflation. Historically this wasn’t a big deal for property insurance because inflation was so low and steady. However, post-COVID, we saw a dramatic shift in inflation.² As evidenced by the performance of homeowners insurers, it’s safe to say most people didn’t see this coming.

So, perfect foresight is a problem, but I think most of us knew this because rate indication work is always trying to forecast future trends. There is a more subtle issue going on though. If you’ve worked in Property insurance long enough, you’ll often hear that as risks get larger within a certain rating class, the rates start to decrease. The argument is usually something along the lines of this. If you compared a 1 story office building to a 10 story office building, there is a smaller chance the 10 story building will see a total loss because by its size alone it would take longer for the fire to completely burn the building. This is a line of reasoning I can get behind. But how is the insurer defining “size”? What you should be very careful about is whether the size element in your rating mechanism is being driven by valuation in some way.

Let’s go back to my $100,000 home example. Just because my home valuation inflated to $110,000 doesn’t mean its size changed; it’s still the same home. The only thing that changed is things got more expensive. However, if your pricing algorithm now takes the $110,000 and uses it for a size discount, you’re underestimating your loss potential. I encourage you to run some pro-forma analyses on your algorithm. Create a premium vs. valuation curve for a given risk, holding the other variables constant. How much is an increased valuation actually flowing into additional premium?

So part property premiums inflation sensitive? Kinda sorta. In order for them to be truly inflation sensitive, remember that two key things need to happen.

1. You need perfect foresight in estimating valuation changes.
2. Your pricing algorithm needs to separate “size” from valuation.

What I’m pondering about:

• A smart actuary pointed out that I’m not talking yet about layering just yet. There is a whole other blog post on just exposure curves I’ll write up at some point.
• I also encourage people to read and analyze rate indication filings or read any other study on premium trend. Yes premium trend is also considering other changes in a portfolio/policy, but you’ll also quickly see how the relationship between premium and valuations is tenuous at best.

1. I use “valuation” to be the same as replacement cost. ↩︎
2. https://fred.stlouisfed.org/series/WPUSI012011 ↩︎