Introduction

Ragnar Frisch worked with errors-in-variables (EIV) models. Later researchers in econometrics moved the field in the direction of errors in equations. That situation was partially rectified in the early 1970s by the contributions of Goldberger (1972) and Griliches (1974) and later surveys by Griliches (1985) and Chamberlain and Goldberger (1990), but EIV models still occupy a back seat in econometrics. When I was revising my Introduction to Econometrics (Maddala, 1992), reviewers unanimously suggested that I drop the chapter on “Errors in Variables” (it is “never used”) and add more interesting and useful topics like unit roots and cointegration.

Empirical researchers, however, have to face the problems of errors in variables all the time. This essay discusses that problem in the context of financial models in which proxies are used for unobservables almost all the time. In the following sections we shall discuss this problem with reference to the following topics:

1. tests of the capital-asset pricing model

2. tests of the arbitrage pricing theory, using observed macroeconomic variables as proxies for unobserved factors

3. measuring market responses to corporate pronouncements (dividends, stock splits, etc.), also known as testing signaling models

4. portfolio performance measures