Description: This paper uses a Bayesian approach to persistency to explore retirement cost projection variability. The analysis is divided into two parts. In the first part a subjective population projection model is developed, based on a beta-binomial distribution for the number of participants persisting through a given age. Probability statements regarding such items as the number of participants at a given age and the distribution of retirees are developed. The second part shows how the model might be used to generate confidence intervals for retirement cost projections. Both the single-entry-age case and the multiple-entry-age case are considered. From Transactions of Society of Actuaries 1979, Vol. 31.