Overview and Introduction
From 1949 to 1973, the Bureau of Labor Statistics (BLS) estimates that U.S. non-farm multifactor productivity grew at 1.9% per year. After 1973, multifactor productivity grew only 0.2% per year (table A). Despite a 20-year intensive research effort to find the cause, no convincing explanation of the post-1973 productivity slowdown exists.
Whatever the ultimate cause, circumstantial evidence suggests that services industries play some important role in the slowdown. In the first place, the aggregate numbers indicate that the slowdown is greater in the non-goods producing portions of the economy. While no official estimate of productivity in services is published by the BLS, nonfarm multifactor productivity slowed by 1.7 percentage points (from 1.9% per year to 0.2%), and manufacturing productivity fell by 0.6 percentage points (from 1.5% per year to 0.9%). Because manufacturing accounts for about 22% of non-farm business, this implies a 2 percentage point slowdown in the non-manufacturing sector.
If the data are right, one might infer, as did Baumol (1967) many years ago, that productivity improvements in services are harder to achieve than in goods producing industries. If so, the shift of the economy toward a larger share of services implies a reduction in the national rate of productivity improvement.
But this view of manufacturing and services is undoubtedly too simple. Substantial disparities exist among productivity growth rates within the manufacturing sector and also within the non-manufacturing sector. It simply is not true that all individual services industries have productivity growth rates that are lower than all individual manufacturing industries, or even below the average for manufacturing industries.
But more importantly, perhaps, the data may not be right. One popular hypothesis about the productivity slowdown is that it is a product of mismeasurement. According to this hypothesis, the mismeasurement of output contributes to the productivity slowdown because an increasing portion of output is not captured in the basic statistics.
Again, circumstantial evidence points to the services industries. Griliches (1994) pointed out that some of the services industries whose productivity growth rates in the 1947-1973 era were as high or higher than productivity growth in manufacturing industries had, since 1973, much lower productivity improvements. Additionally, the productivity slowdown has been particularly intense in services industries where output is hard to measure—health services, for example, have the greatest labor productivity slowdown of any industry in table 1, and both banking and health services have large multifactor productivity slowdowns (see table 2). This points again to possible mismeasurement.
Another puzzle involves computers. The 1992 capital flow table shows the purchases, by industry, of computer equipment (Bonds and Aylor, 1996). The five industries that are the largest purchases of computers are all services—in order, financial services, wholesale trade, business services, insurance, and communications. Those five services industries account for more than 50 per cent of US investment in computers. Within these industries computers have created new forms of service output that may not be fully captured in the statistics. An example is the growth of ATM machines in banks that reduce the time spent waiting in line for teller transactions, make the transactions available on weekends, and have, with computer-assisted verification systems for credit card purchases, virtually eliminated the need to carry traveler’s checks on foreign travel to many countries. Prior to the 1999 revisions to GDP, ATM usage was not reflected in the measure of banking output in the national accounts.
In all of these services industries, conceptual and empirical problems in measuring output and prices are notorious: For example, an economic consulting firm is part of the business services industry. How do we measure the output of an economic consulting firm? How would we construct a price index for economic consulting? And how would we compute the productivity of economists? The science of economics is no closer to developing methods for measuring the output of economists’ own activities than it is for measuring the output of banks, law firms, and insurance agents. All of these services poses difficult problems for constructing price indexes and real output measures and therefore for measuring productivity.
This paper gives a progress report of a project we are conducting, with collaborators, on service sector output and productivity. Its major message is that there is no central theme to the problem of services measurement. Each industry we have examined contains unique problems. If quality change is, as Shapiro and Wilcox (1996) put it, the “house-to-house fighting” of price indexes, measuring service output requires, at least, a hedgerow-by-hedgerow assault.
In the next section, we present some measures of the growth in labor and multifactor productivity within the services industries in a form that is consistent with the published measures for the aggregate economy. This allows us to document the wide dispersion of productivity growth rates across industries and the pervasiveness of the post-1973 slowdown. Section III summarizes recent research on individual sectors that have been subjects of Brookings workshops on measurement issues in services industries.
Paper will be published as Chapter Two in the forthcoming book “Services in the International Economy” edited by Robert M. Stern.
