The consistent estimation of future cash flow and future earnings: a predictive model with accounting double entry constraint

In empirical financial accounting research, there continues to be a debate as to what the best predictors of future earnings and future cash flows might be. Past accruals, earnings and cash flows are the most common predictors, but there is no consensus over their relative contributions, and little attention to the underlying accounting identities that link the components of these three prominent variables. The aim of this thesis is to investigate this controversy further, and to apply an innovative method which yields consistent estimations of future earnings and cash flows, with higher precision and greater efficiency than is the case in published results to date. The estimation imposes constraints based on financial statement articulation, using a system of structural regressions and a framework of simultaneous linear equations, which allows for the most basic property of accounting - double entry book-keeping - to be incorporated as a set of constraints within the model. In predicting future cash flows, the results imply that the constrained model which observes the double entry condition is superior to the models that are not constrained in this way, producing (a) rational signs consistent with expectations, not only in the entire sample but also in each industry, (b) evidence that double entry holds, based on the Wald test that the estimated marginal responses sum to zero, and (c) confirmation of model improvement by way of a higher likelihood and greater precision attached to predictor variables. Furthermore, by then using an appropriately specified model that observes the double entry constraint in order to predict earnings, the thesis reports statistically significant results, across all industries, that cash flows are superior to accruals in explaining future earnings, indicating also that accruals with a lower level of reliability tend to be more relevant in this respect.