Utilizing the Kelly Criterion to Select the Best Projects When Capital is Temporarily Constrained

摘要

The Kelly Criterion, developed in 1956 by John Kelly at the Bell Laboratories, provides a method to allocate capital to a project with the intent of maximizing the return on the capital employed and limiting exposure to a critical shortfall in the total capital available for other projects. This shortfall can occur when projects that are funded early in the funding cycle are subjected to a run of bad luck and both the corporate success rate and value added from exploration falls significantly lower than expected. This disappointment could cause a tactical revision to the budget and diminish the pool of capital available for the remaining projects. Even when this criterion has already been applied to balance the portfolio with the corporate risk attitude and the capital available, the budget may be subjected to a sudden reduction in the remaining funds available due to reasons beyond their control. This constraint may possibly be due to temporary cash flow shortages, another corporate division with a sudden need for capital or as we have seen in the last six months the need to pay down debt. Because the constrained budget is not a change in corporate attitude regarding money to be placed at risk, but rather a temporary economic remedy to a shortage of cash currently available, the company may prefer to reduce the budget year allocation but maintain the corporate risk attitude. To do this the company must determine which projects in the portfolio best meet the corporate objectives for maximizing long term return at an appropriate level of risk and either reduce equity or postpone some projects to meet the cash flow constraints. This paper will suggest one method to make the required adjustments based on a linear programming model. A linear program solution is similar to a marble dropped into a tilted box. The marble will come to rest at the intersection of the two sides that form the lowest location in the box. It will not find a solution if, for instance, one side is perfectly aligned with the low point such that all the points on that edge are equally low or if there are baffles that prevent the marble from continuing to roll to the lowest point. Other more robust models such as non-linear or integer programming might find a solution in these more complex situations.