Click for next page ( 147


The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 146
7 Analyzing the / Sources of Growth A number of economists have done quantitative analyses of the sources of economic growth, or the growth of output, for the United States. Such analyses are closely related to the measurement of productivity since they usually try to divide the sources of growth into two major components: the increase in inputs and the increase in output per unit of input. The latter is what is called productivity growth. When a complex phenomenon has multiple causes, as the growth of output does, there is usually no unique way of decomposing it and attributing portions of the phenomenon to particular causes. Analysts must make certain assumptions in order to proceed; although most of these assumptions may seem reasonable, all can be questioned. Change in economic output is such a phenomenon. Growth of output can be studied at various levels: establishment, firm, industry, or sectors of the economy. The lower the level of aggregation, the simpler the analysis and the more firmly grounded in the theory discussed in Chapter 3. The discussion in this chapter refers mainly to the growth of output for large sectors. CONCEPTUAL FRAMEWORK A basic assumption generally made in analysis of economic growth is that change in output is caused first by change in inputs. If a product is produced by inputs of labor and capital and these inputs each grow by 3 percent while output grows by 5 percent, 3 percentage points of the output 146

OCR for page 146
Analyzing the Sources of Growth 147 growth are attributed to the use of more inputs and the remaining 2 percentage points are attributed to a rise in multi-factor productivity. If, as is usually the case, the inputs do not all increase at the same rate, a weighting scheme is needed to obtain a measure of the growth of total input. The weights most frequently used are the value of inputs in some base period (for a discussion of such weights, see Chapter 3~. The measure of growth of total input will be affected by the nature of the weighting scheme selected and by the choice of dates for the base period. In addition to the complications caused by unequal increase in inputs and choice of base period, difficulties arise when there has been a change in the quality of one or more of the inputs. For example, workers providing the labor input may have more education in a recent period than in the base period. Some analysts include this quality change in their input measure; others include it instead among the explanations of the productivity increase. In either case, one needs a method of measuring the contribution of increased education of workers to the growth of output. This measure is usually derived from cross-sectional differences in hourly wages at a point in time among workers with differing educational attainments. In a perfectly competitive labor market, such differences reflect the marginal productivity of schooling. To the extent that labor markets are not competitive, the resulting estimates of growth from this source are biased, although the size and direction of the bias are difficult to determine without further analysis. Changes in the age, sex, and industrial composition of the workers furnishing the labor input pose issues similar to those posed by changes in educational attainment. Analysts may adjust their input measure to reflect such changes, or they may include them as explanations of changes in a productivity measure that uses unweighted labor inputs. In either case, the contribution of the changing composition of labor input to the change in output must be inferred from cross-sectional differences in hourly wages on the assumption of competitive labor markets (see Chapter 6 and Kunze in this volume). One important source of productivity growth may be economies of scale, which refers to the fact that up to some limit, larger establishments generally are more efficient than smaller ones. Economies of scale are not limited to those internal to firms, but also include those that arise from expansion of markets that an industry serves. Estimates of the effect of increased economies of scale on the growth of output through time are made by inferences from two sets of data: estimates of the change in establishment size through time and estimates of differences at a point in time in output per unit of input for establishments of different sizes. Since economies of scale may differ substantially in different kinds of production,

OCR for page 146
148 REPORT OF THE PANEL some disaggregation is generally desirable in estimating their eject on aggregate output. A second source of productivity growth is shifts of resources among sectors of the economy. For example, for many years the value of output per hour of labor in American agriculture was substantially lower than that in the rest of the economy: a shift of workers from agricultural to nonagricultural employment therefore raised aggregate output because the value of nonagricultural output was raised by more than the value of agricultural output was reduced. To estimate the effect of such shifts, one needs estimates of the size of the shift and of the difference in output per unit of input between the relevant sectors. It should be noted that a shift of resources that raises output per unit of labor will not necessarily raise output per unit of all inputs. It is widely believed that one of the most important sources of productivity change is the growth of knowledge, including both the growth of scientific and technological knowledge and improvements in methods of organization and management. Unfortunately, it has not yet been possible to devise any direct aggregative measure of the growth of knowledge through time. Studies of growth therefore include the growth of knowledge in a residual, attributing to this source most of the change in output that cannot be accounted for otherwise. In interpreting studies of the sources of growth, it should be kept in mind that different sources of growth may interact. An interaction of possibly great importance is that between increases in capital and the growth of technical and scientific knowledge. Many technological ad vances are incorporated or "embodied" in new capital equipment of some kind. This embodiment means that a given rate of growth of knowledge will have a greater effect on output when gross additions to capital stock are large than when they are small. Because of this interaction, much of the effect on output of increases in capital may ultimately have been caused by the improved technology incorporated in new capital equip- ment. The presence of such interactions means that there must be some arbitrary assumptions in any decomposition scheme that does not or cannot explicitly estimate the effects of interaction. This very brief and simplified introduction to methods of analysis of growth, called growth accounting, is not intended to cover all of the problems encountered or the ways in which they have been handled. (For more technical discussion of these issues, see the works cited in this chapter.) To illustrate the greater complexity of the classification actually used in growth-accounting studies, Table 7-1 presents the classification used in one of the best-known studies, that of Denison (1972~.

OCR for page 146
Analyzing the Sources of Growth TABLE 7-1 Denison's Classification of Sources of Growth Z49 Components of Total Factor Input Changes in labor input including: Employment Average hours worked Age/sex composition Education level Effect of length of work week Health Other personal characteristics (effort, etc.) Changes in capital input Changes in non-residential land Components of Output per Unit of Input Improvements in resource allocation (at the industry or higher level of aggregation)a Economies of scale (at the level of the individual firm, the industry, or the economy wide level of aggregation)b Advances in knowledgeC Rate of diffusion of new knowledge (at the industry or higher level of aggregation)d Improvements in resource allocation within firmed Changes in cost of "business services to government"e Changes in aspects of the legal and human environment within which business must operate that affect costs of productionf Changes in the intensity with which employed resources are used that result from fluctuations in the pressure of demand" Changes in irregular factors such as weather and strikes h Changes in the extent to which the use of multiple labor shifts permits economizing in the use of capitals Changes in productive efficiency independent of changes in any of the other determinantsJ SOURCE: Denison, E. F. (1972) Classification of sources of growth. Review of Income and Wealth 1 8(1 ): 22-2 5. Reprinted by permission. aImprovements in Resource Allocation: Changes in the degree to which the actual allocation of employed human and property resources departs from the allocation that would maximize national income. It is con venient to distinguish two broad aspects of allocation, each of which can be further sub divided in detail. (a) The extent to which the allocation-among industries or products, or among firms categorized by size, degree of risk, or other significant characteristics-of each type of input in the aggregate departs from that which would maximize national income. (Each type of capital input distinguished is regarded as a separate input.) (b) The extent to which the allocation of individual workers among individual jobs de parts from that which would maximize national income. There is a less important counterpart for individual capital goods and parcels of land. A wide variety of changes (including aspects of tax, wage, and labor market policy) to increase correspondence between rewards and the value of contributions to output, to identify the abilities and potentialities of workers better, to eliminate discrimination- particularly in hiring and promotion of workers, or for or against risky investments-can affect this determinant. Since economic change itself causes misallocation as a result of lags in adjustment, an increase in mobility of resources or a reduction in the pace of change in demand patterns or technology also affects this determinant.

OCR for page 146
150 FOOTNOTES TO TABLE 7-1 (Continued) REPORT OF THE PANEL bEconomies of Scale: Changes in economies of scale resulting from changes in the size and product compo- sition of the national market for business output. The size of the national market is determined by all the other income determinants listed, and cannot be altered directly. Neither can composition of the market if free consumer choice is allowed. Changes in economies of scale resulting from changes in the geographic concentra- tion of domestic customers (for intermediate as well as final products, hence of pro- duction as well as of individuals). This determinant can be changed by policies that alter the geographic distribution. CAdvances in Knowledge: Changes in the state of knowledge-technical, managerial, or organizational-that govern the amount of output that business can obtain by use of a given quantity of resources. In general, changes in the state of knowledge occur on an international basis, but this determinant may vary from country to country because of differences in economic structure. This determinant is greatly affected by the mere passage of time because knowledge once obtained is rarely lost, but it may be affected by such factors as the amount of resources devoted to research, the level of education and training, and the character of patent laws. dRate of Diffusion of New Knowledge, and Improvements in Resource Allocation Within Firms: Changes in the amount by which output obtained with the average production tech nique actually used falls below what it would be if the best technique were used, because of: (a) Changes in obstacles imposed (usually by government or labor union regulation) against efficient utilization of resources in the uses to which they are actually put. This determinant can be altered by changing the restrictions. (b) Changes in the extent to which existing knowledge is available to those in a position to apply it. This determinant can be altered by improving channels of communica tion. (c) Changes in the time lag between the dates at which business structures and equip ment are installed (incorporating knowledge of design at that date) and the dates at which they are in use. This determinant can be altered by modernizing capital goods, and this may occur if investment in business structures and equipment is increased. eChanges in the Cost of "Business Services to Government": Changes in the cost of "business services to government," such as collecting taxes or filing statistical reports, and changes in the adequacy of "government services to busi- ness," such as provision of law courts or roads for business use. This determinant can be altered by transferring the costs of functions between busi ness and government. fChanges in Aspects of the Legal and Human Environment Within Which Business Must Operate That Affect Costs of Production: Changes in aspects of the legal and human environment within which business must operate that affect costs of production by business. One example is the honesty of the public in general and customers or suppliers in particular, which affects business costs of protection against robbery, fraud, etc., and may even govern the determination of whether certain types of business operation, such as self-service, are feasible. Another example, currently important in many countries, is changes in requirements imposed to limit polluting in the process of production. These determinants may be altered by policies that affect public behavior, or by changes in laws that affect the conditions under which production may take place, or in the distribution of costs of environmental protection and the like between business, on one hand, and consumers and government, on the other.

OCR for page 146
Analyzing the Sources of Growth FOOTNOTES TO TABLE 7-1 (Continued) 151 "Changes in the Intensity With Which Employed Resources Are Used That Result from Fluctuations in the Pressure of Demand: Changes in the intensity with which employed resources are used that result from fluctuations in the pressure of demand. The chief reason that this determinant, as well as that of h, requires isolation is to make it possible to disentangle transient from continuing forces. However, this deter- minant may be affected by aggregate demand management policies. hChanges in Irregular Factors Such As Weather and Strikes: Changes in irregular factors that affect output per unit of input, particularly in the weather and/or in the impact of strikes. This determinant may be altered by measures to affect the weather, or to promote labor peace. "Changes in the Extent to Which the Use of Multiple Labor Shifts Permits Economizing in the Use of Capital: Changes in the extent to which the use of multiple labor shifts permits economizing in the use of capital in particular uses, apart from changes resulting from variations in the pressure of demand. This determinant may be affected by changing the use of multiple shifts. JChanges in Productive Efficiency Independent of Changes in Any of the Other De- terminants: Changes in productive efficiency that take place independently of changes in any of the other determinants. Economists are sometimes reluctant to admit existence of this determinant because it is inconvenient. I am convinced that efficiency, so Rae-;, d,f- fers among countries and surmise that it may vary over time within a country. One plausible explanation is that efficiency actually achieved is affected by the strength of competitive pressures upon firms to minimize costs. So closely linked that, for empirical studies, I include it here is the quality of man- agement. This is conceptually covered under the classifications of labor input and re- source allocation, but I do not think it can be comprehensively handled in this way at present. We would need to know more about this determinant to identify with any certainty the types of policy that would affect it, but if my surmises are correct they would in- clude intensification of competitive pressures, and policies to improve the selection and training of management and to stimulate the replacement of unsuccessful managers. EMPIRICAL EVIDENCE Empirical evidence on the sources of productivity growth has been generated by two types of research. One is the growth-accounting study, discussed in the preceding section, in which the investigator attempts to account for all the sources of the actual growth of output over some time period. The other type of research measures directly the ejects of one or more of the factors that usually remain in the residual or the growth- accounting studies. Studies of the relationship between research and development (R&D) expenditures and productivity advance are of this type. Such studies are usually carried out at the level of the detailed

OCR for page 146
152 REPORT OF THE PANEL industry or even the individual firm rather than at the more aggregate level considered in growth-accounting studies. GROWTH-ACCOUNTING STUDIES The main body of empirical work on growth accounting covers the penod since 1929. In Table 7-2 this penod is broken down into three subpenods: 1929-1948, 1948-1969, and 1969-1977. In the first penod, which includes the depression and World War II, the growth of measured productivity was significantly below that of the second penod. In the most recent penod the rate of productivity growth has fallen below that of the two earlier penods. Our review here is restricted to the penod since World War II.: The main difficulties of measurement and interpretation can be illustrated from a survey of the studies that cover this penod. For 1948-1969, our discussion is based largely on the work of Denison and of Jorgenson and Chnstensen; for the most recent penod we have drawn from the work of several researchers.2 1948-1969 Table 7-2 shows that during 1948-1969, productivity, measured either by the output per labor-hour or by output per unit of input, grew more rapidly than in the preceding penod. Tables 7-3 and 7-4 summarize Denison's analysis of the sources of this productivity growth. The TABLE 7-2 Growth Rates for Aggregate Productivity Measures, Selected Time Periods, 1929-1977 (average annual rates in percentage points) Productivity Measure 1929-1948 1948-1969 1969-~977 Output per unit of input (Denison)a 1.5 2.1 Not available Output per hourb 2.3 c 3.1 1.8 aDerived from output data in Table 6-1, p. 62, of Denison, E. F. (1974) Accounting for U.S. Economic Growth 1929-1969. Washington, D.C.: Brookings Institution. Reprinted by permission. Output covers the nonresidential business sector and refers to the Bureau of Economic Analysis's national income measure in 1958 prices. bDerived from the Bureau of Labor Statisticst output measures in Table B-38, p. 301, of U.S. President (1978) Economic Report of the President, 1978. Washington, D.C.: U.S. Government Printing Office. Output covers the private business sector and refers to the Bureau of Economic Analysis's GNP measure in 1972 prices. CEstimated from Table 6-1 of Denison, E. F. (1974) Accounting for U.S. Economic Growth, 1929-1969. Washington, D.C.: Brookings Institution. Reprinted by permission. See note on output in footnote a above.

OCR for page 146
Analyzing the Sources of Growth 153 TABLE 7-3 Denison's Analysis of Sources of Growth in Output,a 1948-1969 Output Factor inputs Labor (quantity and composition) Physical capital Landc Output per unit of input Improved resource allocation Economies of scale Irregular factors Residual Contributions to Growth Rate (Percentage Points per Year) 3.72 1.58 1.00 0.58 0.00 2.14 0.37 0.51 -0.18 1.44 Percent of Total Growth Rate Accounted for 00.0 42.s 26.9 5.6 0.0 s7.s 9.9 3.7 - .8 38.7 aIn the nonresidential business sector. bIncludes age, sex, schooling, efficiency of shorter hours (Table 7-1). CDenison assumes that the amount of land in production does not change during this period. SOURCE: Denison, E. F. (1974) Accounting for U.S. Economic Growth, 1929-1969. Table 8-2, p. 111. Washington, D.C.: Brookings Institution. Reprinted by permission. TABLE 7-4 Denison's Analysis of Sources of Growth in Output per Lab or-Hour,a 1948- 1 969 Contributions to Growth Rate (Percentage Points) Percent of Total Growth Rate Accounted for Output per labor-hour 3.13 100.0 Inputs 1.00 31.9 Labor composition per labor-hour 0.54 17.3 Capital per labor-hour 0.48 15.3 Land per labor-hour -0.02 -0.6 Output per unit of input 2.14 68.1 Resource allocation 0.37 11.5 Economies of scale 0.5 1 16.3 Irregular factors -0.18 -5.7 Residual 1.44 46.0 Numbers may not total due to rounding. aIn the nonresidential business sector. SOURCE: Derived from Venison, E. F. (1974) Accounting for U.S. Economic Growth, 1929-1969. Table 8-4, p. 114. Washington, D.C.: Brookings Institution. Reprinted by . . permlsslon.

OCR for page 146
154 REPORT OF THE PANEL contributions of the input categories are shown to permit comparison of the contributions of input and non-input factors to total output growth. The contributions of the sources are shown for both types of productivity measure.3 The contribution of each of the inputs to output growth is equal to the product of the growth rate in the amount of that input and the share of that input's value in total cost. For example, the entry of 0.58 for physical capital in Table 7-3 is the product of capital's share (about 0.16) and the average annual growth rate in Denison's index of capital input (3.6 percent). For the contribution to growth in output per labor-hour, the entries for inputs are equal to the product of the income share and the difference in growth rates between that input and labor-hours.4 The difference between the growth rate in output (and output per labor- hour) and the contribution of all the inputs is equal to the growth rate in output per unit of input. It is clear from both tables that these increases in output per unit of input as defined by Denison were more important in contributing to output growth than even the fairly large set of input factors that Denison measures (see Table 7-1~. It is also clear that a sizable amount of output growth during this period had to be included in the residual. The residual was derived by subtracting from the growth rate in output per unit of input the contribution of those sources of productivity growth that Denison was able to measure. Although this residual could include the effects of many forces, it is Denison's judgment that, for this period, unmeasured sources other than advances in knowledge had no appreciable effect on measured changes in output per unit of input. Denison explains his concept of advances in knowledge (Denison 1974, p. 79~: The term "advances in knowledge" must be construed comprehensively. It includes what is usually defined as technological knowledge knowledge concern- ing the physical properties of things, and of how to make, combine, or use them in a physical sense. It also includes "managerial knowledge" knowledge of business organization and of management techniques construed in the broadest sense. Advances in knowledge comprise knowledge originating in this country and abroad, and knowledge obtained in any way: by organized research, by individual research workers and inventors and by simply observation and experience. Two investigators, Griliches and Kendrick, have attempted to measure how much of this aggregate residual could be explained by the observable investments in R&D activities. Griliches (1980) estimated that in 1966, growth in the stock of R&D capital was contributing about 0.34 percentage point to the annual growth rate in output per unit of input. Kendrick (1977) estimated that, over the entire period 1948-1966, growth

OCR for page 146
Analyzing the Sources of Growth 155 in R&D capital contributed about 0.85 percentage point to the annual growth rate in output per unit of input. (Terleckyj [1974] has made a similar estimate.) These two estimates imply that between 24 and 59 percent of Denison's residual during 1949-1969 might have reflected advances in knowledge generated by visible R&D activities. Griliches urges caution in using his estimates and specifically notes that they may be upwardly biased. Although the two estimates do not refer to the same time period, the great difference between them strongly suggests the need for further research in this area. Jorgenson and Christensen have estimated the importance of productivi- ty change in the growth of gross private domestic product for 1949-1969. They find that real product grew at the rate of 4.34 percent, real factor input at 2.79 percent, and multi-factor productivity at 1.55 percent. The change in productivity accounts for 36 percent of the increase in real product (see Jorgenson and Christensen 1973, p. 3084. This is a smaller fraction than in Denison's estimates for a period of one year longer, in which productivity accounts for more than 50 percent of growth of output. One of the principal differences between the two estimates is that Denison is accounting for net output while Jorgenson and Christensen are accounting for gross output.5 196~1977 The analysis for the 1966-1977 period is not as reliable as that for the previous period, partly because of time constraints; careful research using revised data takes time. In addition, the recent period has been marked by rapid inflation, which makes it more difficult to measure correctly the amount of physical capital (including inventories). Table 7-S presents data, by industry division, on the apparent slowdown in productivity growth that has been mentioned at several points in our report.6 It shows that the apparent slowdown that began about 1966 has occurred in almost every industry sector. From Table 7-5 and from Table 7-6 (which shows changes in the relative share of aggregate labor input by industry divisions over time) we estimate that shifts in industry mix since 1966, frown divisions with high growth rates to those with low ones (e.g., services), cannot account for more than 0.1 of a percentage point of the 1.5 percentage point slowdown.7 The slowdown in productivity growth since 1966 also appears in the estimates of multi-factor productivity by Jorgenson and his associates. As noted above, Jorgenson and Christensen estimate the rate of growth of multi-factor productivity for 1949-1969 at 1.55 percent a year. For 1966- 1973 a simple average of the corresponding estimated annual rates of

OCR for page 146
156 REPORT OF THE PANEL TABLE 7-5 Growth Rates in Output per l~bor-Hour for Private Business Sector and Industry Divisions, Selected Time Periods (average annual rates in percent) Sector or Industrya1947-1966 b 1966-1977 Private business sector3.1 1.6 Farming5.6 4.6 Manufacturing3.1 2.2 Mining4.3 -1.0 Construction3.0 -1.5 Transportation3.1 2.0 Communications5.2 5.6 Utilities6.3 2.2 Wholesale and retail trade2.6 1.9 Finance, insurance, and real estate0.9 0.4 Services1.5 1.1 Government enterprises-0.7 -0.1 aIndustry rates derived from unpublished data provided by the Bureau of Labor Statistics, U.S. Department of Labor. bThe year 1966 was an all-time high period of productivity growth; breaking the 1947- 1977 trends at 1966 therefore emphasizes the slowdown. However, when the subperiods are broken at 1968 or later, the slowdown is still readily discernible. TABLE 7-6 Distribution of Hours in the Private Business Sector by Industry Division (percentage) Industry Division194819671977 Private business sector100.0100.0100.0 I arming18.37.04.S Manufacturing29.132.929.4 Mining1.81.11 .: Construction5.26.36.6 Transportation6.24.84.4 Communications1.41.61.7 Utilities1.01.11.1 Wholesale and retail trade21.624.225.7 I inancc, insurance, and real estate3.65.26.4 Services10.413.516.4 Government enterprises1.52.22.2 SOURCE: Derived from unpublished data provided by the Bureau of Labor Statistics. U.S. [department of Labor.

OCR for page 146
Analyzing the Sources of Growth TABLE 7-7 Growth Rates in Output per Labor-Hour for Aggregate Economy Measures, by Selected Periods (average annual percentage rates) 157 Sector 1 947-1966 1966-1 978 1966-1973 1973-1 978 Private business 3.2 1.6 2.1 1.S Private nonfarm business 2.6 1.5 l.9 0.9 SOU RC ~ : U.S. Presi den t ( 1 979 ) Economic R eport of the Presiden t, 1 9 79. Table B-3 8, p. 227. Washington, I).C.: U.S. Government Printing Office. growth of multi-factor productivity is only 0.8 percent a year, and the estimates for 3 of the 8 years are negative (see Christensen et al. 1980, Table 11.10). Table 7-7 shows the apparent slowdown in output per labor-hour for more detailed time periods, and separately for the private business sector and the private nonfarm business sector. We show the two sectors separately because one of the causes of the slowdown has been a deceleration of the flow of resources out of farming relative to earlier periods. As we noted above (see Table 7-4), improved resource allocation through the shift of resources from the farm to the nonfarm sector contributed about 0.3 of a percentage point to the average annual growth of output per labor-hour during 1948-1969. According to Denison (1978b): " . . . the gain from this source has completely disappeared." The effects of this source can be seen in Table 7-7 by looking at the size of the decline in growth rates (between 1947-1966 and 1966-1978) within each sector; the decline narrows from 1.6 to 1.1 percentage points as one moves from the private business sector to the private nonfarm business sector. This differential decline (0.5 percentage point) is a rough measure of the contribution from this source to the overall productivity slowdown.8 Comparing productivity growth during the two subperiods since 1966 (Table 7-7) suggests that the slowdown may be getting worse. However, cyclical factors were severe during the 1973-1974 downturn in the economy, and the recovery that began early in 1975 had not reached a peak by the end of 1977. It is possible that productivity will grow enough in the rest of the cyclical expansion to get the economy back on the growth path it was on from 1966 to 1973; however, a study by Peter Clark (1977) suggests that the old growth path may not be regained. His analysis indicates that the full benefits of cyclical recovery were achieved in 1977 and that the economy had only regained the same growth rate, not the same growth path, as in 1966-1973: this would mean that a once-and-for- all loss in output growth occurred in 1973-1974. The comparison between

OCR for page 146
158 REPORT OF THE PANEL TABLE 7-8 Possible Sources of the Slowdown in Productivity Growth Between 1948-1965 and 1965-1976, Output per Hour of All Workers in the Private Nonfarm Business Sector Adjusted for Adjusted for Change in Adjusted for EfTects of En Average Annual Age-Sex- Effects of vironmental and Growth Rate Education Capital/Labor Health Regula Time Period Percentagea Mix b GrowthC lions and Crimed 1948-1965 2.6 2.2 1.8 1.8 1965-1976 1.6 1.1 0.7 0.9 Difference 1.0 1.1 1.1 0.9 SOURCES: aU.S. President (1979) Economic Report of the President, 1979. Table B-38, p. 227. Washington, D.C.: U.S. Government Printing Office. bClark, Peter (1977) Capital formation and the recent productivity slowdown. Table 3. Journal of Finance (December); Denison, E. F. (1974) Accountcngfor U.S. Economic Growth, 1929-1969. P. 32. Washington, D.C.: Brookings Institution; and Denison, E. F. (1978) Where has productivity gone? Basis Point 3(1). CVarious sources; see text discussion. dDenison, E. F. (1978) Effects of selected changes in the institutional and human en vironment upon output per unit of input. Table 15, p. 41. Survey of CurrentBusiness (January). 1973-1977 and 196~1973 is probably not as important as finding out whether the decline in growth rate between 196~1977 and 1947-1966 (1.1 percentage points) will eventually represent a one-time-only loss in output growth or a permanent shift to a lower growth rate.9 Recent growth accounting analyses by Denison (1978a, 1978b), Clark (1977), Kendrick (1977), and Norsworthy and Fulco (1977) attempt to account for the apparent slowdown. Table 7-8 summarizes the findings of those studies: the differential in trend rates between the two periods for the private nonfarm business sector is shown to be 1.0 percentage points. This differential is unaffected by changes in the shift of resources from farming in the two periods. The average rates for each period after both have been adjusted for the effects of changes in age-sex-education composition of the labor input are also shown in the table. On balance, these factors had a positive effect on output growth in both periods, with the effect being larger in the latter period: thus, changes in labor force composition make it harder to explain the slowdown. On the basis of this factor alone, one would have expected the growth rate to have accelerated by 0.1 of a percentage point. Table 7-8 also presents an estimate of the effect of changes in the rate of

OCR for page 146
Analyzing the Sources of Growth 159 growth of the capital/labor ratio during the two periods. This ratio is the proper way to account for changes in capital input when one is seeking to explain changes in output per unit of labor input, rather than the growth of output itself. The estimate is based on an average annual growth rate in the capital/labor ratio of 2.2 percent in the 1948-1965 period and 2.0 percent in the 1965-1977 period. These rates were derived from BEA time- series data on the capital stock in constant dollars of firms in the private business sector as defined in Bus productivity estimates (from Norsworthy and Harper 1979~. No adjustment is made for possible differences in the effects of capacity utilization on growth rates.~ The observed share of capital in total factor incomes (0.2) was used to convert those input growth rates into contributions to growth rates in output per labor-hour. This calculation indicates that none (or at most less than 0.1 of a percentage point) of the slowdown can be accounted for by deceleration in the growth rate of the capital/labor ratio in the nonfarm business sector. ii The last column in Table 7-8 shows the growth rates adjusted for the estimated effects on measured productivity of increased costs to businesses of complying with government environmental and health and safety regulations and of combating crime (e.g., shoplifting, vandalism). These estimates were taken from Denison's (1978a) study of the period 1957- 1975. For 1957-1965, Denison estimates an effect of only -0.02 percentage point, while for 1965-1975 he estimates an effect of-0.2 percentage point. Thus, this factor could account for about 0.2 percentage point of the decline. Overall, this analysis indicates that only about one third of the apparent slowdown can be explained in the way described above in terms of both input factors (capital/labor ratio, labor composition) and other measured factors (improved resource allocation and environmental and health regulations and crime). This suggests that factors such as advances in knowledge and the rate of diffusion and intensity of the use of new knowledge may have decelerated during the recent period. It is not clear why the rate of accumulation and diffusion of new knowledge should have begun slowing after 1966. Kendrick stresses the possible role of R&D investment, estimating that the slowdown in both the growth of the measured R&D stock and of the rate of diffusion of new technical knowledge could account for a decline of 0.3 percentage point in productivity growth between the two periods. Denison (1978b) speculates that increased government regulation may have diverted the attention of top management and diminished its effectiveness in seeking out and using new techniques. Fabricant stresses that the recent rapid inflation may have reduced the willingness of investors to undertake risky ventures, which are the kind most likely to produce above-average productivity gains.

OCR for page 146
160 REPORT OF THE PANEL MICRO-LEVEL STUDIES OF THE DETERMINANTS OF PRODUCTIVITY Studies at the micro-level have concentrated on the relation between R&D expenditures and productivity change. There have been relatively few studies of other determinants of productivity, such as economies of scale, managerial and organizational innovations, and the speed of diffusion of knowledge. This section gives two examples of the evidence on R&D as a determinant of productivity growth. Griliches (1980) attempted to estimate the overall contribution of R&D to the growth in aggregate productivity. To do this, he had to estimate two quantities, the growth rate of the aggregate stock of R&D input and the marginal product of that stock. To estimate the marginal product, he turned to the findings of micro-level studies of the effects of R&D on productivity. Such studies have been going on since the early 1960s, and although they have become increasingly more comprehensive over time (see Mansfield 1978), they cannot trace all the productivity effects generated by a successful innovation. If one is to estimate the overall (or social) marginal product of the R&D inputs of a single firm, one must look not only at the productivity of this firm, but also at other firms in the same industry. And when innovations lead to new and better products, one must look at firms in other industries as well. (Innovations that do not result in new products or services are called "process innovations" those that lower the cost of producing an existing product or service.) Research based on individual firms and using a detailed case study approach with each firm has been most successful in tracing some secondary effects. However, those cases may not be typical of all R&D inputs in the economy. Conversely, the more comprehensive studies that cover a large part of all R&D inputs have had to restrict themselves to estimating the private, as opposed to the social, marginal product. The study by Griliches (1980) measures the private return earned by each firm on its own R&D investment. Griliches worked with a data file containing observations on 883 individual manufacturing firms, which accounted for more than 91 percent of all formal industrial R&D conducted in the United States in 1963. The file merges each firm's records from the Census Bureau's economic censuses and annual surveys with data collected by the Census Bureau for the National Science Foundation's survey of R&D expenditures and employees. The economic censuses and surveys give data on the firm's value of shipments, purchases of intermediate goods, investment in durable equipment, total employees, and hours of production workers. Griliches develops a model relating the firm's output to inputs of labor and capital, R&D expenditure, the stock of R&D input, and the rate of

OCR for page 146
Analyzing the Sources of Growth I6I return on R&D investment. He estimates the parameters of his model by fitting multiple regression equations (the observations are on the individual firm) to measures of output and of inputs. He estimates the private rate of return to R&D investment at 27 percenti3 and concludes (Griliches 1980, p. VI-3: "In general, this paper can be viewed as another supporting link in a chain of a rather limited number of investigations for the argument that R&D investments have yielded a rather high rate of return in the recent past." A recent study by Mansfield et al. (1977) illustrates the nature of case study evidence. The authors contacted a number of business firms in the Northeast and tried to persuade them to provide data bearing on the social and private returns from their innovations. Many firms would not participate, feeling that the information requested was too confidential to release to anyone. Although these refusals probably bias the sample, the results nevertheless suggest a possible range of outcomes. TABLE 7-9 Characteristics of Selected Innovations Approximate Industry Producing Type of Type of Date of the Innovation Innovation Nature of Innovation User Innovation Primary metals New product New type of metal that reduced Firms Late 1930s cost of appliances Machine tools Control systems Construction Drilling Industrial equipment Paper Thread New product New product New product New product New process New product New product Industrial controls New product Electronics New product Chemicalsa New product New computer controls New type of component New material that cut cost of building Reduced cost of drilling wells New type of drafting New paper product that cut costs of users New type of thread that cut costs of garment makers New mechanism for doors New device that reduced costs of certain video tape operations New product that reduced costs Firms Firms Late 1960s Firms Mid 1960s Firms Firms ~- r lrms Firms Firms Firms F. 1rms Mid 1960s Mid 1960s Early 1960s Early 1970s Early 1 970s Late 1960s of users Chemicalsa New process Reduced costs of production Firms Mid 1960s Chemicalsa New process Reduced cost of certain aromatic Firms Late 1960s chemicals Chemicalsa New process Major process improvement Firms Early 1960s Household cleaners New product New product that reduced cost Households Early 1960s of cleaning floors Stain removers New product New stain remover Households Mid 1960s Dishwashingliquids New product New product that cut costs of Households Early 1960s operating dishwashers aChemicals are defined to include petroleum refining. SOURCE: Mansfield, E., et al. (1977) Social and private rates of return from industrial innovations. Table I. Quarterly Joumal of Economics: 221~240. Reprinted by permission.

OCR for page 146
162 REPORT OF THE PANEL TABLE 7-10 Estimated Social and Private Rates of Return from Investment in 17 Innovations Rate of Return (percentage) Innovation Social Private Primary metals innovation 17 18 Machine tool innovation 83 35 Component for control system 29 7 Construction material 96 9 Drilling material 54 16 Drafting innovation 92 47 Paper innovation 82 42 Thread ir~novation 307 27 Door control innovation 27 37 New electronic device Negative Negative Chemical product innovation 71 9 Chemical process innovation 32 25 Chemical process innovation 13 4 Major chemical process innovation 56a 31 Household cleaning device 209 214 Stain remover 416 4 Dishwashingliquid 45 46 Median 5 6 25 abased on investment of entire industry. SOURCE: Mansfield, E., et al. (1977) Social and private rates of return from industrial innovations. Table I. Quarterly.Ioumal of Economics: 221-240. Reprinted by permission The authors (Mansfield et al. 1977, p.222) asked each firm to pick at random one or more of its recent innovations. [Then] . . . many manweeks were spent gathering data concerning each innova- tion and its ejects from the innovating firm, from firms using the innovation (if it was used by firms), and from other sources. These innovations occurred in a wide variety of industries (described below), and in firms of quite different sizes. Most of them are of average or routine importance, not major breakthroughs. Although the sample cannot be regarded as randomly selected, there is no obvious indication that it is biased toward very profitable innovations (socially or privately) or relatively unprofitable ones. The sample is described in Table I [Table 7-9 of this report]. The authors developed methods for estimating the private and social rate of return for each of the three types of innovation shown in Table 7-9: new product with firms as user; new product with household as user; and new process. Table 7-10 reproduces the findings of the study on the private and social rate of return for the 17 innovations in the sample.

OCR for page 146
Analyzing the Sources of Growth 163 This study, like the Griliches study, gives the overall impression of a high return to resources invested in R&D. It also suggests that industrial R&D has probably been an important source of productivity advance for the economy as a whole. However, the data underlying these studies are imprecise. In the Griliches study, the impreciseness relates to the familiar limitations of survey data for measuring output and input. For the Mansfield study, the impreciseness results both from the non-randomness and limited size of the sample of firms and from the dependence on firms' records and on subjective estimates by executives for measures of prices, quantities, and costs. RECOMMENDATIONS This chapter has briefly reviewed the concepts, findings, and some of the limitations of recent growth-accounting studies at a more detailed level of some of the sources of productivity growth. These studies have been valuable in suggesting, at least in general terms, the relative importance of forces contributing to output growth, within the limitations imposed by the assumptions made to derive the estimates. Further research should contribute to reducing those limitations. Because all of the growth-accounting studies have been done outside the government, the literature has been brought up to date only at long and irregular intervals. Recent results are generally not available when they might be of most interest for current economic analysis. Recommendation 16. The Panel recommends that BES and BEA take joint responsibility for developing and maintaining measures of some of the sources of growth (such as physical capital and work-force composition) so that policy makers can have timely and accurate information on at least the more easily measurable sources of productivity change. Although the existing research on the sources of productivity change provides useful knowledge about the broad patterns of contributions to growth, it tells us relatively little about more detailed aspects and patterns. How do the various inputs and non-input sources interact with one another (for example, the rate of capital investment and the rate of embodied technical change)? What is the relative importance of new product and service innovation, whose output erects are much harder to measure, in total innovation? How representative is the existing evidence on the social rate of return to R&D investments?

OCR for page 146
164 REPORT OF THE PANEL Recommendation 17. The Panel recommends that government agen- cies support research aimed at improving knowledge about the sources of productivity change. These agencies should be especially attentive to research that focuses on measuring technical and organizational change and new product and service innovation. The Panel also recommends funding micro-level studies of the returns to research and development. Micro-level studies of innovations in personnel management and other organizational behavior should also be encouraged. NOTES 1. For a detailed analysis of the 1929-1948 period, see Denison (1974~. 2. Different researchers break the postwar period into different subperiods: Denison uses 1948-1969 and 1969-1977; BES (see next section) uses 1947-1966 and 1966-1977; Christensen, Cummings, and Jorgenson (see Chapter 9) use a variety of starting dates. 3. The data on output per labor-hour were derived from output per person employed in the nonresidential business sector in Denison (1974, Table 8-2, p. 111~. The growth rate of output per hour was derived by subtracting from the growth rate of output per person employed the contribution of "average hours per person" ~-0.38 percentage point) plus a small amount (0.1 percentage point) for the contribution of the greater growth in capital per labor-hour and the lesser decline in land per labor-hour than for the corresponding rates per employed person. This analysis may not apply precisely to the BES measure, which is for the private business sector. 4. For the assumptions underlying this method of calculating input contribu- tions, see Chapter 3. 5. For detailed discussion of the differences in methods, see the series of articles by Jorgenson and Griliches and by Denison in the special issue of Survey of Current Business (1972) 52~5) Part II. 6. For most of the industries in Table 7-5, BES maintains but does not publish productivity series because it believes that the available data do not measure output reliably. 7. Shifts of industry mix made a much larger negative contribution to growth in the 1947-1966 period-about 0.4 percentage point. If anything, this makes it harder to explain the slowdown. Note, however, that the other important type of shift in industry mix, from industries with low levels of productivity to industries with high levels, does play a significant role in explaining the slowdown (see below). 8. This measure may overstate the effects from that source because the growth rate in the capital/labor ratio decelerated more in agriculture than elsewhere so that part of the 0.5 differential is attributable to the capital/labor factor, which we treat separately within the nonfarm business sector. Probably, 0.3 percent is a better estimate of the contribution of the deceleration in improved resource allocation to the slowdown.

OCR for page 146
Analyzing the Sources of Growth 165 9. The impact of a one-percentage-point lower growth rate becomes quite significant even in a relatively short time period. For example, $10,000 grows to $18,061 after 20 years at a 3-percent rate, but only to $14,859 at a 2-percent rate. 10. Peter Clark (1977) estimates that differential trends in capacity utilization during the two periods bias the estimates based on the unadjusted stock data. Although there is probably some effect of differential capacity utilization on the slowdown, it is very difficult to measure, especially in nonmanufacturing sectors. 11. Very recently (since 1974), the growth rate in the capital/labor ratio has decelerated sharply, so this measure will become important in explaining future developments. 12. Kendrick's and Fabricant's analyses were submitted as testimony before the Joint Economic Committee (U.S. Congress 1978~. 13. The rate of return is equal to the dollar value of the marginal product per dollar of R&D input. Thus, 27 percent means that $100 of R&D investment will yield a permanent increase in output per year of $27. In this form, the productivity effects of R&D can easily be compared with those of other forms of investment for increasing output, such as investment in human capital and physical capital.