The accuracy of the BLS productivity measures - Bureau of Labor Statistics

Monthly Labor Review, Feb, 1999 by Edwin R. Dean

Overall productivity growth may be somewhat underestimated; despite continued progress, measurement and conceptual barriers remain

The Bureau of Labor Statistics has made sustained efforts to improve its productivity measures. The goals of these efforts have been to enhance the reliability of the measures; to facilitate analysis of economic performance; and to provide useful information to the public.(1) The BLS clearly recognizes that, despite the beneficial results of its program, there is room for further improvement.

In the past several years, some have voiced concerns about the accuracy of the trends in the BLS productivity series, mainly to suggest that productivity growth for the business sector of the economy has been understated. The BLS has also been concerned about the accuracy of the major sector productivity trends and has devoted considerable effort to examining the accuracy of these trends.

The concerns about possible underestimation of productivity growth have been focused on data for the business sector of the economy, and especially the services components of that sector. "Services," broadly defined, include all producing activities outside the "goods" sector. The major services-producing activities are transportation, communications, utilities, retail and wholesale trade, finance and insurance, and various additional services rendered to persons and businesses.(2) Commentators have wondered why productivity in services has not grown nearly as rapidly as productivity in manufacturing, particularly in light of anecdotal indications of improvements in several types of services.

Concerns have also been expressed about several measurement techniques used by the BLS to compute productivity trends. These concerns are directed to such questions as whether the BLS productivity data fully reflect changes in the quality of goods and services; whether the best techniques are used to introduce new, advanced products into the data series; and whether the BLS methods capture the full impact of new information technology on economic performance. Accordingly, several points in this paper attend briefly to these issues.

Five key issues

This article addresses five important issues related to the BLS measure of productivity in the business sector of the economy. First, the article examines whether there is in fact mismeasurement of productivity growth in the services portion of the economy. The article concludes that there are important measurement problems in some service activities and these problems may be leading to underestimation of productivity growth rates.

Second, the article addresses specific sources of mismeasurement, the sectors of the economy where these problems are found, and the possibility of determining the extent of underestimation. It concludes that there is no basis for determining the extent of the underestimation. Although existing information sheds some light on the magnitude of the problem, there is no basis for a precise determination of its extent. The available information does not indicate that the published data understate productivity growth by a large order of magnitude.

Third, the article discusses what can be done to improve the measurement of productivity in services. It describes a number of recent improvements in the quality of Federal Government statistics that have led, in turn, to improvements in the productivity data. In addition, it discusses steps that can be taken to improve further the productivity measures, especially within the services sector.

Two other significant issues cut across discussion of these three matters. One is the impact of alleged biases in the consumer Price Index on productivity data. The other is the intrinsic difficulty of defining output for a number of service activities, activities that have been labeled "hard-to-measure" services. The discussion of all of these questions draws heavily on the articles by William Gullickson and Michael Harper (pp. 47-67), and by Lucy Eldridge (pp. 35-46) in this issue.

How the trends differ

The most effective way to begin a discussion of these measurement issues is to compare the labor productivity trends for two major sectors, the business sector and manufacturing, one of its components.(3) Following are the average annual percentage rates of growth of labor productivity, or output per hour of work, for these two sectors:

              Business     Manufacturing

 1949-73        3.3             2.6
 1973-90        1.2             2.4
 1990-98        1.4             3.7

In the business sector, the growth rate of output per hour declined from the robust rate of 3.3 percent prior to 1973 to slightly more than 1 percent in the years following 1973. This decline is often called the "productivity slowdown." Another noteworthy result of these data is the contrast, after 1973, between continued robust growth in manufacturing productivity and the sharp deceleration of growth in overall business sector productivity.

The pattern of changes in the trends in the BLS series on multifactor productivity resembles that of the changes in labor productivity trends. Multifactor productivity is output per unit of all inputs combined, including labor, capital, and other inputs.(4) Following are the average annual percentage rates of growth of multifactor productivity for the private business sector and manufacturing: