An empirical analysis of life insurance policyholder surrender activity

The non-forfeiture options of a cash value life insurance policy allow the policyholder to gain access to any cash value he or she may have built up over the life of the contract. One of these options is the policy surrender, which is a voluntary decision to terminate coverage and to receive the entire cash value of the policy. Since most life insurer assets are invested in fixed-income instruments of positive duration, insurer portfolios are susceptible to losses in the event of an increase in interest rates. As a withdrawal feature, the surrender option is a potential source of interest rate risk for life insurers. The surrender option may also be exercised by many policyholders in the event of economic duress. Because the surrender option is a legally mandated part of all cash value life insurance policies, insurers may be legally bound to bear a risk that may be highly correlated across their pools of cash value life insurance policies, violating a primary principle of insurance. This dissertation analyzes data on life insurance policyholder surrender activity to determine whether the risks posed by policy surrender are functions of certain macroeconomic variables and, as a result, are highly correlated across all policies. The results indicate that surrender activity is positively related to inflation, real interest rates, and unemployment, while surrender activity is directly related to real per capita income. Average policy size also has an inverse relationship with surrender activity while policy loans and company leverage have a positive but lagged influence on surrenders. Since many policies that are surrendered are replaced with new, and sometimes better, products, the incentives of the insurance agent in recommending surrender and replacement are potentially important factors in life insurance disintermediation. Specifically, the role of agent compensation is explored in a principal-agent context and legalized premium rebating is analyzed using data from the two states in which rebating has been legalized, California and Florida. Rebating is found to have a direct statistical relationship with surrender activity