Catastrophic losses and insurer profitability: evidence from 9/11

Journal of Risk and Insurance, March, 2008 by Xuanjuan Chen, Helen Doerpinghaus, Bing-Xuan Lin, Tong Yu

ABSTRACT

We examine the effects of 9/11 on the insurance industry, hypothesizing a short-run claim effect, resulting from insufficient premium ex ante for catastrophic losses, and a long-run growth effect, resulting from ex post insurance supply reductions and risk updating. Following Yoon and Starks (1995) we use short- and long-run abnormal forecast revisions to measure both effects, analyzing them as a function of firm-specific characteristics. We find that firm type, loss estimates, reinsurance use, and tax position are important determinants of the short-run position. Firm type, loss estimates, financial strength, underwriting risk, and reinsurance are key determinants of the firm's long-run position.

INTRODUCTION

The terrorist attack on the World Trade Center (WTC) on September 11, 2001, had an enormous impact on the insurance industry. Swiss Reinsurance estimates WTC losses at US$21 billion (Swiss Re, 2004), making it one of the most costly natural or manmade catastrophes in history. The WTC attack resulted in immediate and unexpected increases in insurer claim payments as well as a reduction in their capacity to supply catastrophe insurance. In theory, the WTC attack may have led to risk updating by insurers, resulting in a subsequent increase in insurance demand. Prior work suggests that both supply reduction by insurers and risk updating by insurers positively affect insurer long-run growth and profitability (Gron, 1994; Winter, 1994; Cummins and Danzon, 1997; Froot and O'Connell, 1999). The purpose of this study is to investigate the effect on insurers of both the large unexpected losses experienced in the short term (a "claim effect") and the long-run potential growth opportunities for insurers due to supply reduction and potential risk updating (a "growth effect") following 9/11.

The prior literature using event study methodology provides mixed evidence on the effect of catastrophic events on insurers. Shelor, Anderson, and Cross (1992) find positive abnormal returns for property and liability insurers after the Loma Prieta earthquake, positing that this is due to increased demand for insurance coverage, consistent with a growth effect. Conversely, Lamb (1995) finds a negative stock price reaction for insurers with direct business exposure following Hurricane Andrew, consistent with a short-run insurance claim effect. Similarly, Cummins and Lewis (2003) examine the WTC attack, the Northridge earthquake, and Hurricane Andrew, and find a strong negative impact (i.e., a claim effect) on insurer stock prices immediately following the catastrophe.

The conventional event study methodology, however, does not allow separation of short--and long-run effects that may occur with a mega-catastrophe like the WTC attack. Consequently, we adopt the Yoon and Starks (1995) approach, using short-run abnormal forecast revisions to measure changes in firms' existing assets and long-run abnormal forecast revisions to measure firms' future growth opportunities. We use 1-year earnings forecasts for short-run analysis and 5-year earnings growth forecasts for long-run analysis. Instead of focusing on the industry-level response to 9/11 we test short- and long-run abnormal forecast revisions as a function of individual firm characteristics. Prior studies provide some insight on firm characteristics that may explain insurer performance following a catastrophe. Kleffner and Doherty (1996) show that the supply of earthquake coverage is a function of insurer leverage, level of diversification, and profitability. Doherty, Lamm-Tennant, and Starks (2003) argue that leverage, level of new capital, expected growth, and losses may affect post-WTC stock price performance. Gron and Winton (2001) find that insurer product concentration, specifically in riskier or longer-tail lines of business, affects firm performance. Cummins and Lewis (2003) suggest insurer financial ratings to be an important predictor of postloss stock performance.

Using data from the Institutional Brokerage Estimate System (IBES) database, we obtain analysts' forecasts of earnings per share for a sample of public U.S. insurers in order to test short- and long-run abnormal forecast revisions as a function of firm characteristics. We match the IBES data with Compustat, Lexis Nexis, National Association of Insurance Commission (NAIC), and A.M. Best's data on firm financial and other variables. Our results are consistent with a short-run claim effect and a long-run growth effect. Specifically, there is a statistically significant relationship between the short-run abnormal forecast revision and firm type (property-liability vs. non-property-liability insurer), estimated loss, use of reinsurance, and the insurer's tax position. We find evidence consistent with a long-run growth effect as well. There is a statistically significant relationship between the long-run abnormal forecast revision and firm type, estimated losses, financial strength, underwriting risk, and use of reinsurance.