An issue of theoretical and practical importance to R&D managers is the proper allocation of an organization's resources between the periodic selection and budgeting of R&D projects and the implementation of these projects. One possibility is to consider previous allocations as an established base and to restrict the attention of R&D managers to an examination of any subsequent deviations from the base. The advantage of this strategy, often called incremental budgeting, is that it focuses managerial and analytical efforts on a few current decisions. A second possibility, called zero-base budgeting, is to ignore previous allocations and to reexamine all feasible alternatives at the start of each budget cycle. This paper presents three models, a static model and two dynamic models, in which incremental and zero-base allocation procedures are viewed as extremes between which there may be a point of diminishing returns. The models are risk/return models in which a total budget is allocated between the budgeting effort, which reduces risk, and the budgeted activities, which increase expected return. The budget for each activity during each time stage is partitioned into two components---a base and an increment. The expected return for each activity is a concave increasing function of its budget, and the risk depends quadratically on the base and in a geometrically decreasing fashion, on the bases in previous stages. The cost of evaluation is proportional to the increment, and the total cost of all activities, when added to the budgets for the activities, cannot exceed a fixed total budget. Zero-base budgeting means setting the base to zero so that the budget equals the increment plus the cost of evaluation. It is shown that a policy of zero-base budgeting during all time stages is nonoptimal, but an instance of zero-base budgeting during a single stage may be optimal.
