Dean Baker and Matt Harmon, writing for the Center for Economic and Policy Research blog, analyzed various methodologies of measuring price changes in auto insurance: the Consumer Price Index (CPI), the Personal Consumption Expenditure deflator and what they refer to as I.I.I. data (average expenditures published by the NAIC). Their comments appear to be driven by the April 2018 report of the Bureau of Labor Statistics that the price of auto insurance in the CPI rose by 9.0 percent over the price in April 2017. This observation leads to the claim that auto insurance “has passed medical care as a driver of inflation.”
The blog post takes a number of tangents—on health insurance, on the different methodologies used by the CPI and the PCE deflator—and then turn to the observation that what people actually pay for auto insurance isn’t increasing very much at all. They infer that because the cost of auto insurance is rising in the CPI, and at the same time the Bureau of Labor Statistics consumer expenditures survey shows that auto insurance spending remains constant, consumers must be offsetting higher prices by buying less insurance.
In the comments section of the post, I.I.I.’s chief economist, Dr. Steven Weisbart addresses Baker and Harmon’s analyses of the various measurements of auto insurance costs and refutes the assumption that policyholders are assuming more risk as opposed to comparison shopping for cheaper polices, for example.
His full response appears below.
Dean Baker and Matt Harmon are correct that “there are aspects to the issue [of measuring inflation in insurance] that are informative about how we measure and think about inflation,” but their analysis is mistaken.
The CPI does, in general, aim to measure price changes in “quality-adjusted” goods and services. Its method for doing this for auto insurance is, unfortunately, seriously deficient. This is because auto insurance premiums are expected-cost driven. (By law, insurers cannot charge to make up for losses in prior years or charge in one state to make up for losses in another state.) Further, premiums are set based not only on expected claims (and claims adjustment expenses, including litigation defense for some third-party collisions), but also on expected investment income from the period between collection of the premium and the payment of a claim. The BLS methodology for determining current prices for auto insurance does not—indeed, cannot—capture these forces. Without recognizing them, premium increases in the current year over the prior year is mistakenly perceived as inflation.
In the last few years, there has been a dramatic upsurge in both the frequency and severity of auto insurance claims, both private passenger auto and commercial auto. Although severity (the average dollar cost of claims, unadjusted for quality improvements in recent years) has been increasing for a long time, increases in frequency (the number of claims per exposure unit) have been unusual and have been rising sharply. The Insurance Information Institute has discussed this in a white paper, identifying some of the major causes of these increases as increased congestion from the continued growth in the number employed, increased distracted driving, and higher speed limits in some cases. Auto insurers did not foresee these changes. They also (with many others) did not foresee the continued low interest rates that delivered lower investment income than they would have earned (and which they would have used to keep premium increases down) and in recent years increased premiums to try to get “ahead of the curve.” These are not inflationary increases in a quality-adjusted financial service.
Moreover, the BLS measure doesn’t try to capture what people pay. It created a hypothetical buyer and asks a panel of insurance companies what they would charge that buyer. The response doesn’t capture discounts that many insurers offer, such as for insuring both auto and home and other coverages with the same company, or for being a long-term policyholder, or being accident-free. As Baker/Harmon recognize, increases in what people pay for auto insurance have been much smaller than the BLS inflation measure. They infer that this means policyholders are assuming more risk. That’s possible, but other inferences are equally possible—such as that comparison shopping has led them to find the same coverage for a lower premium, or that some coverages are no longer cost-effective as their cars age.
Finally, the 9.0 percent year-over-year increase cited by Baker/Harmon is a bit of a cherry-picked datum. On the day the blog post was published, BLS released its CPI report for June 2018. It shows the CPI for auto insurance at 7.6 percent. Further, in two months the auto insurance component of the CPI will likely drop further, since the 12-month figure includes an unusual 0.9 percent increase in August 2017 that is unlikely to be matched in July or August 2018. Note that in the four most recent months of 2018 increases in the auto insurance component of the CPI were 0.3 percent in March, -0.2 percent in April, 0.4 percent in May, and 0.3 percent in June.