ABSTRACT This study further substantiates the presence of insurance underwriting cycles and analyzes their causes. A generalized least squares analysis of changes in premium levels is used to test the rational expectations/ institutional intervention hypothesis across countries as well as within each country. We also examine the relation between cycle length and the market/institutional features of each country. Finally, a logistic model is used to predict the presence of a cycle based on the market/ institutional features. The results suggest that the rational expectations/ institutional intervention hypothesis explains many aspects of the underwriting cycle, including cycle length. INTRODUCTION The cyclicality of underwriting profits for the property-liability insurance industry has been extensively researched.1 The industry results tend to follow a cycle consisting of alternating uniform periods of rising and then falling underwriting profits. The period in which the cycle "bottoms out" is typically characterized by withdrawal of insurers from some markets resulting in availability crises for some lines (Cummins, Harrington, and Klein, 1992a). Periods of rising underwriting profitability are characterized by increased entry of insurers and expanded coverage for many lines of business. Studies show that the underwriting cycle in the United States tends to follow a second-order autoregressive process with a total period (or cycle) averaging six years (Cummins and Outreville, 1987; Venezian, 1985; Doherty and Kang, 1988; and Smith and Gahin, 1983). But cycles are not limited to the United States; Cummins and Outreville (1987) observe underwriting cycles over long periods of time and in many countries. Explanations for the cycles follow two schools of thought. The first explanation is based on the premise that insurance markets operate irrationally (Venezian, 1985) and/or exhibit market imperfections (Gron, 1994; Winter, 1994), while the second emphasizes rationality with institutional intervention. The argument of irrational behavior suggests that insurance markets are destabilized by phenomena such as extrapolative forecasting and so-called "cash flow underwriting," which can result in prices considerably higher or lower than competitive levels due to erroneous estimates of losses or investment income. The Winter/Gron "capacity constraint" theory posits that cycles are caused by impediments to capital flows that result in alternating periods of excessive and inadequate capacity in the industry.2 According to this scenario, the underwriting cycle is most prominent in longtail lines (usually liability lines) because forecasting horizons are longer and anticipated investment income is more substantial for these lines.3 The other school of thought, which we call the rational expectations/ institutional intervention hypothesis, emphasizes insurance market rationality and considers whether underwriting cycles are caused by external events and market features not under the control of the insurer. Reactions to these phenomena make it appear that insurers behave irrationally. These external events include: (1) institutional, regulatory, and accounting characteristics (Cummins and Outreville, 1987; Witt and Miller, 1981; Outreville, 1990; and Tennyson, 1991); (2) exogenous shocks to surplus attributed to natural catastrophes, unexpected increases in claim costs, or shifts in loss distributions (Cummins and McDonald, 1992); (3) interest rate changes coupled with changes in equity values (Doherty and Garven, 1992; and Cummins and Danzon, 1997); and (4) uncertainties in the market environment (Berger and Cummins, 1992). Although the evidence is extensive regarding the existence of and causes for underwriting cycles in the United States, the examination of international underwriting cycles is limited to Cummins and Outreville (1987). The purpose of the present article is to explore further the presence of underwriting cycles in international property-liability insurance markets and to test the hypothesized explanations for these cycles. …
Read full abstract