U.S. federal energy efficiency standards are, by law, set at the maximum level of energy efficiency that is
technically feasible and economically justified, as gauged through regulatory impact analysis (RIA). RIA has
been an integral part of the government policy-making process for over 20 years in the U.S., but in February
2011, the main government oversight agency issued guidance to all agency heads for the first time that the “bestavailable
techniques” for RIA include those that identify “changing future compliance costs that might result
from technological innovation or anticipated behavioral changes.” This paper explores the question of what
techniques are currently in use by regulators to account for innovation in RIA, and what makes those available
techniques “best.” The paper focuses on two major examples of the use of learning curves in regulatory cost
estimates. The first was pioneered by the Environmental Protection Agency (EPA) in vehicle emissions control
at least as early as the late 1970s, and later adopted for vehicle fuel economy regulation by the National Highway
Safety Administration (NHTSA) in the mid-2000s. The second was implemented by the Department of Energy
(DOE) in 2011 to help set appliance energy efficiency standards. The paper: (1) provides an overview of some of
the major findings of the academic literature on learning curves in order to inform an assessment of a “best”
approach to a learning curve-based RIA cost adjustment technique; (2) describes the EPA-NHTSA and DOE
approaches to this technique; and (3) assesses these approaches against the criteria of alignment with economic
theory and of administrative sustainability (i.e., fit with existing laws and institutional arrangements, including
relationships between regulators and regulated industries).