Abstract

This paper examines the use of changing weight indexes of inputs and outputs in productivity measurement. 1 The U.S. Bureau of Labor Statistics (BLS) productivity measurement program has for many years preferred changing-weight indexes to fixed-weight indexes. The paper examines the reasons for this preference and how it has been implemented over time. The paper examines this broad problem as follows. The next section of the paper, section 1, discusses weighting schemes for input indexes, from the viewpoint of the general logic for selecting an input index method and also by drawing on the theoretical literature relating to input measurement. Section 2 treats output indexes in a parallel fashion. In both sections, it is concluded that changing-weight indexes are preferable to fixed-weight indexes. The third section explores several important practical issues that a statistical organization must resolve before implementing one of the changing-weight index forms and explores the tension between the ideal methods favored in the theoretical literature and the practical requirements of an on-going statistical program. This section also examines the issue of which output concept—for example, value added or gross output—is appropriate for industries and sectors of the economy. Section 4 describes how BLS has implemented its choice of changing-weight indexes for productivity data. Section 5 is an analysis of the effects of the BLS changingweight indexes on trends in selected input, output, and productivity series. The final section of this paper, section 6, summarizes the main results of the paper. It also offers a few observations on international comparisons of productivity trends. 1

Citations