Date of Award

Fall 1994

Project Type


Program or Major

Natural Resources

Degree Name

Doctor of Philosophy

First Advisor

Robert C Harriss


Previous studies which have generated estimates of the potential for cost-effective gains in energy efficiency have generally acknowledged the uncertainty in their inputs and conclusions, but none have gone beyond simple scenario analysis to quantify the uncertainties in their inputs or conclusions. This research develops and demonstrates methods for the explicit treatment of uncertainty in energy conservation potential analysis. New methods and critical data requirements are highlighted through application to the study of current weatherization potential.

Sensitivity analysis finds that, contrary to claims appearing in the literature, estimates of cost-effective conservation potential are more sensitive to uncertainties in empirical inputs than to variations in assumed discount rates. A taxonomy of the input uncertainties affecting estimates of current weatherization potential is developed. The availability of data to support estimates of each input uncertainty is found to be minimal. Estimates of annual energy savings are the most uncertain input to the analysis of current weatherization potential. This input's uncertainty is also significantly more complex to analyze and characterize than that of either installed cost or measure lifetime (for weatherization measures), because of the number of separate factors contributing to it.

Methods are demonstrated for translating probabilistic descriptions of input uncertainties into probabilistic measure-specific conclusions. Methods for aggregating and plotting these results in modified conservation supply curves are presented. Ninety percent confidence intervals for population mean cost of conserved energy per weatherization measure are estimated to range from roughly 60% to nearly 400% of typical point estimates. Ninety percent confidence intervals for population mean annual energy savings per weatherization measure are estimated to be less wide but still significant, ranging from roughly 35% to 160% of "typical" point estimates. The most significant contributor to uncertainty in both of these outputs is uncertainty in estimates of annual energy savings per measure installation. Probabilistic conclusions about the supply curve as a whole, as well as confidence intervals for such summary results as the total technical potential and the total cost-effective conservation potential given a threshold price, all require specification of the statistical dependence of each uncertainty's influence upon separate measures.