Prediction of the U.S. Employment Links: An Application of an Empirical Bayes Procedure.

An empirical Bayes procedure is used to adaptively predict monthly employment links (a link being the ratio of all employee counts in a particular month to the corresponding figure in the previous month) for the metropolitan statistical areas throughout the United States. By comparing with the true link of a month which is available only nine to thirteen months after the month has passed, the author's prediction is substantially and uniformly superior in MSAs, months, and states to existing estimators in terms of the average of squared deviations or the average of the absolute relative errors.