Elusive Efficiency and the X-Factor in Incentive Regulation: the Törnqvist v. DEA/Malmquist Dispute

Thu Sep 06 16:24:00 EDT 2007
By Dr. Jeff D. Makholm

Incentive-based regulation is practiced worldwide, and all applications of it require some form of efficiency or productivity measurement—the X-factor. Including this factor in a multi-year regulatory formula allows the formula to survive intact for several years, and this longer regulatory lag between tariff reviews strengthens the incentives on firm performance. The factor, an index number, is intended to permit prices to move between tariff reviews according to an objective and reliable pattern. Differing opinions have arisen, however, on which index number to use. NERA Senior Vice President Dr. Jeff D. Makholm examines these opinions in this chapter from The Line in the Sand: The Shifting Boundary Between Markets and Regulation in Network Industries.

One index number, the Malmquist Index, has generated considerable interest in some regions (particularly in Australia and Europe) because of its ostensible ability, when used in conjunction with data envelopment analysis (DEA), to distinguish readily between technical change for an industry (which the X-factor is generally held to measure) and efficiency for a particular firm. However, the DEA/Malmquist procedure for separating individual firm efficiency from technical change is inherently unreliable for identifying how inefficient a firm is. Neither the quality of data for regulated firms, nor the essentially idiosyncratic nature of such firms, supports an analysis of the level of efficiency of individual utilities. To the extent that regulators attempt to use the DEA/Malmquist procedure to set tariffs to reflect "efficient firm" standards, they inject unsupportable subjectivity and an unreliable methodology into a tariffmaking process. The only reliable alternative is to estimate the X-factor directly by measuring long-run rates of change in efficiency indices. The Törnqvist index is best suited to this process, but other similar indices offer similar results.