The Lighting Case Study Recast as a Decision Tree
Economic benefits in the case of improved solid state lighting arise primarily from savings in electricity times the marginal cost of electricity generation. They also arise potentially from reduced capital costs and reduced air conditioning costs. The report of the Panel on Benefits of Lighting R&D (see Appendix F) focused on the electricity savings benefits and in essence considered a simple decision tree with three outcomes: 150 lumens per watt (lpw), 125 lpw, and 100 lpw for each of two levels of program funding: $50 million and $25 million per year.
In its calculations, the panel assumed no benefits without the government-supported program, eliminating the second branch at the first node of the tree. At the second node (first chance node), the panel estimated the probabilities of technical success; for the case of the $50 million budget, the probabilities were 40 percent for 150 lpw, 65 percent at 125 lpw, and 90 percent at 100 lpw. The panel chose to display a range around the probability estimates and then averaged them for each outcome. For the $50 million budget, it then calculated the benefits, but for only the stretch goal of 150 lpw. (The panel also calculated the benefits for the 100 lpw outcome, but only for the $25 million funding level.) A value of 70 percent was assumed for the probability of market success. The economic benefits were obtained from a DOE National Energy Modeling System (NEMS) run, discounting them at 4 percent and multiplying them by the probabilities of technical and market success. The final calculations generated a conservative estimate of expected benefits, because the panel utilized only the top branch at each node of the decision tree, excluding the expected benefits of partial success. One effect of this simplification was that the panel calculated a larger benefit for the smaller funding level than it did for the larger investment, despite the fact that the benefits of the partial investment would presumably be realized along the way toward meeting the larger goal.
When the lighting case is reexamined in the future, a few simple changes may be useful. First, NEMS can be used to estimate two basic items: (1) the savings in electricity if the lighting research is successful in achieving the high, medium, and low outcomes and (2) the marginal cost of electricity generation and distribution. Rather than using NEMS to obtain the benefits per se, the electricity savings and marginal cost of electricity can be multiplied by each other to obtain a benefits estimate that approximates economic surplus, a closer approximation of total benefits than the consumer benefits estimate generated by NEMS. A second change might be to use the simple spreadsheet calculations to estimate benefits for the medium and low outcomes and to multiply these benefits by the probabilities so that a weighted total expected benefit can be calculated to avoid the problem with underestimation encountered when only the stretch goal (outcome) is used.
Third, a more explicit elicitation of market probabilities of success based in part on consideration of the next-best alternative would be useful.