The Economic Method of Mixed Estimation - An Application to the Credibility of Trend

Actuaries are often confronted with conflicting data and information. In trend analysis, ordinary least squares regression techniques do not allow for the introduction of extraneous information, conflicting or not. Credibility methods have been proposed to solve this shortcoming but have not been widely accepted. This paper discusses the use of an econometric technique, known as mixed estimation, to incorporate prior information directly into the specification of the trend model. The resultant parameter estimates are credibility weighted estimates. Mixed estimation goes one step further by generating a test statistic to test the compatibility of the data and the complement information. An effort has been made to keep theory and notation to a minimum, emphasizing practical application of the technique.