An Estimate of Statistical Variation in Development Factor Methods

The author provides variability estimates for development factor projections under the assumption that the age-to-age factors are lognormally distributed. Keywords: Confidence Estimates Abstract This paper explores some properties of the lognormal distribution. It is possible that these properties can provide information not only regarding the variability of age-to-age development factors but also regarding that of age-to-ultimate factors, if the actuary is willing to assume that these factors are lognormally distributed. Considered are problems of parameter estimation and uncertainty under the assumption of independence of the age-to-age factors. Some results are generalized by weakening the independence hypothesis, and a method of parameter estimation with missing observations is presented. This paper is intended as a starting point, indicating useful results if factors are assumed to be lognormally distributed. Still an open question is the specific situations where such is the case.