A note on added information in the RAS procedure: Reexamination of some evidence

An example in Miernyk (1977) presented a rather counterintuitive result, namely, that introducing accurate exogenous information into an RAS matrix estimating procedure could lead to an estimate that was worse than one generated by RAS using no exogenous information at all. This became an oft-cited black mark against RAS. Miller and Blair (1985) included a different (and small) illustration of the same possibility. It was recently pointed out by one of us that the Miller-Blair numerical results are wrong. For that reason, we decided to reexamine all the empirical evidence we could find on the subject. While figures in both Miernyk and Miller-Blair appear to be wrong, more recent published examples seem to have it right. In short, it is possible to identify examples in which additional (correct) information leads to poorer RAS estimates, at least under several fairly common metrics for comparing "closeness" of matrices. However, the overwhelming majority of the evidence is to the contrary. As a general rule, introduction of accurate exogenous information into RAS improves the resulting estimates, and counterexamples should probably not be taken too seriously.