The consequences of using ordinary Single Equation Least Squares in estimating regression parameters are examined in a case where the parameters relate to a household expenditure study in which the statistical model underlying the study is really a set of simultaneous equations.