of Return in Empirical Research," Accounting and Business Research (Summer 1979), pp. 201-208. L.CL. Skerratt and G. Whittington, "On the Use of the
In the likely case of zero covariance between ~i and €i (the error with which ARR proxies IRR is independent of the error with which size can explain IRR) then Var(ei) > Var(£i). This means that the precision with which b can be estimated from equation (3) is likely to be less than estimating directly from equation (2). Consequently, the chances of a type II error are increased.