Estimating Pervasive Factors with Missing Observations

We suggest a technique for estimating pervasive economic factors which allows the use of all available security return data. The resulting factor estimates can be used in applications and tests of the Arbitrage Pricing Theory (APT). An obvious advantage of the technique is that more precise estimates of the factors are obtained while avoiding potential survivorship biases in factor construction. Empirically, the factor estimates using the entire data set outperform (in terms of asset pricing) estimates using only continuously traded assets.