Multiple Indicators and the Causal Approach to Measurement Error

Since the practical utility of the kind of multiple indicators approach discussed in Costner's paper depends on the ways it can be broadened to include a diversity of causal situations, the purpose of the present paper is to extend the argument in three respects: (1) to show that it holds generally in any recursive system; (2) to note the circumstances under which a single indicator of one or more of the variables can be used; and (3) to point out that the use of multiple indicators can be combined with an instrumental-variables approach that has been discussed in the econometrics literature.