This ambiguity is a central reason for turning to variance components estimation; unlike regression, variance components procedures can assess the independent importance of nested business-unit, corporate and industry effects. Indeed, as Table 1 shows, corporate effects are essentially nil (in Sample A). Thus, the apparent significance attached to the introduction of corporate effects in the fixed-effect estimation is wholly due to the dispersion among corporate returns induced by the deeper business-unit effects. This case is an excellent illustration of the problem associated with using R2 or importance. In Sample A (Table 2) the marginal R2 of the industry and corporate effects are