As did Schmalensee (1985: 348), I rely on Searle's (1971: ch. 9-11) treatment of the theory and practice of variance component estimation. The basic method draws on the fact that any quadratic form in observations is a linear combination of the variance components. To see this, let r be the vector of N observed returns, let i' = E(r) and let V = var(r). From (2) it should be clear that every element of i' is µ and that each element of V is a linear combination of the seven unknown variances and covariances. A theorem by Searle (1971: 54) establishes that for any symmetric matrix Q,