Knowledge that Transforms

Internal control systems, Individual judgement, Group judgment, Decision making

Allocation of indirect costs among products sometimes yields a paradoxical result that unit cost for each product may increase under one method of allocation and may decrease for each product under another method. The Stalcup Paper Company case illustrate such behaviour of costs at the same, time, provides an accounting example of Simpon’s Reversal Paradox Simpson (1951), which Blyth (1972) discussed in the statistics literature. As with other paradoxes, this cost allocation paradox disappears upon closer scrutiny. This paper examines the properties of allocated costs in order to arrive at an intuitive understanding of the results. The relationship of the cost allocation problem to Simpson’s Paradox and the implications of the analysis for cost control are briefly discussed. Necessary and sufficient condition for occurrence of the Paradox is also given. (This abstract was borrowed from another version of this item.)

Accounting changes, Analyst forecasts, Forecast accuracy, Earnings announcement

Corporate failure, Reporting delays, Information dissemination, Annual reports

In this paper, I postulate a causal, or structural, model of corporate earnings and present a theoretical analysis of various empirical issues related to the identification and estimation of time-series earnings models.' The postulated model describing corporate earnings is derived endogenously by specifying a model of the firm's production and invest ment structure and its inventory accounting rule. The main characteristic of this model is that the firm uses linear stochastic decision rules to determine its production, inventory, and capital investment levels. Such decision rules are frequently postulated in the economics literature. Using the theoretical properties of the derived earnings model, I then address estimation and forecasting issues in a general manner without actually doing the estimation. Starting with Dopuch and Watts [1972], much of the empirical research in accounting on the structural (i.e., time-series) properties of earnings numbers has first postulated a linear, stochastic, time-series model as the underlying earnings-generating process, and then identified and estimated the model using a sample of realized earnings numbers. These accounting studies usually have the goal of estimating an expectation model to generate expected or forecasted earnings. This type of end