Managers are usually responsible for the control of the level of several process variables, such as cost, quality, rate of output, and so on. The levels of these process variables are known as the states of the system, and they may be represented by the values either of a continuous variable or by a discrete variable. It is assumed that these states can be ordered in terms of their desirability. Some processes may move only from a more desirable to a less desirable state while others may shift in either direction. These shifts may occur with or without the intervention of the manager. The control system is a plan formulated by management to indicate when intervention should take place. Three types of control systems are generally possible. The first involves no intervention by the manager until a breakdown of the process occurs. This approach is to be favored when the continued operation of the control system is less costly than the benefits to be derived from intervention. The second type of control system consists of periodic intervention by the manager for purposes of adjusting or otherwise influencing the process. The most compelling reason for this approach lies in its ease of application. The third approach bases the intervention decision on information obtained from the process. This information is typically obtained by sampling. The present paper will concentrate on this third type of control system. Several components are incorporated into any control system. First, the manager must establish the variable or variables to be controlled. He must

Cost variance, Cost control, Capital investment decision

T HAS long been recognized that a good economic forecast should consist of several components. In addition to predicting turning points, a forecast should theoretically include some statements about the timing of the and the amplitude and duration of the subsequent movement. If the forecaster makes quarterly quantitative estimates of GNP, he is, in fact, estimating all of the aforementioned components. However, when forecasters use other types of predictions, they usually do not provide estimates of the timing or amplitude. This is especially true when analysts speak of the forecasting behavior of the leading series and/or the rate of change methods. While has been considerable discussion about the success of these methods in forecasting turning points, I very little is known about other aspects of their forecasting behavior. It has generally been concluded, that, in practice, the leading series and rate of change methods predict every turning point of a predictand such as the Federal Reserve Board's Index of Industrial Production, but they also display a large number of false leads. As for their other forecasting characteristics, Moore2 has concluded that some information about the amplitude of a recession could be obtained about six months after the movement first began. Finally Wright and Okun have attempted to estimate the dates of turning points, I but no attempt has been made to attach probabilities to the predicted dates of turning points. Our paper will present one method for attaching probabilities to the turning point forecasts which are obtained from using the leading series and diffusion indexes. The probability must refer to the likelihood of the predictand's occurring in a given time interval. Statements such as there will be a turn or there is an X per cent chance of a turn are tautologies, for sooner or later will be a turn. Thus to have any usefulness, the probabilities must be attached to specific time periods. In the following section we shall outline the methodology of the study, discuss the data to which this methodology was applied, and finally present the results.