Indeed, further evidence suggests that the failure to achieve 100 per cent explanatory power can be blamed largely on the restrictiveness of the above assumptions, as opposed to a failure to forecast earnings over an infinite horizon. Consider a regression that includes the forecast of price to-book premium at the end of four years in addition to the regressors in equation (4). Because that premium should reflect abnormal earnings beyond four years, that regression should achieve 100 percent explanato ry power under assumptions (1) through (4). In fact, such a regression (not reported in the table) produces an average R squared of 80 percent. That we can approach this level of explanatory power so closely with only accounting data on the right-hand side is a strong testimony to the power of accounting data to reflect value well even over relatively short hori zons. There is evidently little to be gained by forecasting earnings and book value beyond four years!