Comment and Discussion

Comment by Ralph A. Winter: Scott Harrington has highlighted the most puzzling aspect of the property-liability insurance market: its sporadic volatility. Premiums are relatively stable, even declining, for long periods (such as 1977-84 or 1988-2001). Then policyholders renewing their policies find that their premiums have skyrocketed. The number of transactions drops dramatically in these hard markets: for many lines, insurance is described as unavailable—meaning unavailable at moderate prices. In 1985-86 and again in 2001, many policyholders renewing their policies saw their premiums double or triple and their coverage drop. Harrington discusses several hypotheses in light of the evidence and then offers some innovative evidence regarding a particular source of instability in the market, which he refers to as "aberrant pricing."

In this discussion of Harrington's paper, I comment on several aspects of the insurance cycle and the evidence he presents:

• • The specific empirical features of insurance market dynamics that must be explained by a theory of the insurance cycle,

• • Several of the theories that have been offered to explain these features and how they stack up against the evidence, and

• • Harrington's evidence of "aberrant pricing" and its role in the insurance cycle. I characterize his results as evidence of information heterogeneity across firms rather than as aberrant pricing, but I agree that his evidence reveals a potentially important source of volatility.

I start with the data. The data on the insurance market are unusual in that we have no direct and reliable time-series evidence on either prices or quantities. We are studying a cycle in a market without reliable data on the two most basic variables. We have, however, time series on revenue: net premiums written. This series is often used as a proxy for prices and sometimes as a proxy for the amount of insurance written. We have to keep in mind that it is neither. The net premiums written time series is revenue: the product of price and quantity.

We can, however, obtain general information on price changes through surveys or—because the changes can be sudden and dramatic—simply through reports in the press. Thus we know that in 1985 and 1986 prices jumped and quantity dropped dramatically. The press was replete with reports of soaring premiums and declining coverage. The March 24, 1986, cover of Time magazine stated, "Sorry, America, your insurance has been canceled." We can also combine the strong anecdotal evidence with the available data on revenue. Revenue for liability lines tripled over the 1985-86 period, for example. With the information that the quantity of insurance coverage had dropped—some policies were canceled, some coverage limits were reduced severely, and for many coverage moved from occurrence to a claims-made basis—this jump in revenue meant that prices increased by much more than triple over two years. In short, during hard markets, (1) prices increase sharply, (2) quantity drops, and (3) revenue increases. Hard markets have two additional features as well: (4) the non-linearity of prices increases sharply, in the sense that the marginal price of obtaining high coverage rises even more than the price of basic coverage during tight markets, and (5) hard markets tend to affect particular lines more severely than others, although the recent jump in premiums seems to be more widespread than previous episodes. The task of any economic theory of insurance market dynamics is to explain the unusual dynamics of pricing (stable or declining premiums for some years, interrupted by sharp jumps in premiums) and other features of the market, especially the cyclical behavior of quantity and revenue. Quantity is strongly countercyclical to price movements, and revenue is procyclical.

Harrington starts, as he should, with the most fundamental theory of insurance market pricing: the perfect capital markets theory. He puts it succinctly: "With rational insurers and policyholders, competitive insurance markets, and frictionless capital markets, insurance premiums will equal the risk-adjusted discounted value of expected cash outflows for claims, sales expenses, and other costs." It is useful to express this hypothesis more precisely, by noting that "expected" in this statement refers to the expectation conditional on all information available at...