Dale Jorgenson

Samuel W. Morris University Professor

PRODUCTIVITY: Volume One -- Preface

This volume is devoted to postwar growth in the United States and is the first of two volumes containing my empirical studies of economic growth. Although the level of U.S. per capita output was higher than that of any other industrialized economy at the end of World War II, U.S. output has increased by more than four times and per capita output has more than doubled. By either measure the postwar growth performance of the U.S. economy has been outstanding. This volume documents in detail that investment is the predominant source of postwar U.S. economic growth.

A second volume, International Comparisons of Economic Growth, focuses on comparisons between the U.S. and Japan . This volume shows that the Asian model of economic growth exemplified by Japan gives even greater weight to investment and that investment is especially critical during periods of exceptionally rapid growth. Although the U.S. and Japan are often portrayed as economic adversaries, postwar experiences in both countries support policies that give high priority to stimulating and rewarding capital formation.

Jan Tinbergen (1942) was the first to divide the sources of economic growth between the two broad categories of investment and productivity. The comparative significance of these two components has dominated the voluminous literature on U.S. economic growth created during the postwar period. During the much discussed growth slowdown after 1973 it became apparent that protracted and sometimes acrimonious professional debates had failed to produce a consensus that could serve as a guide to economic policy.

Profound differences in policy implications militate against any simple resolution of the debate on the relative importance of investment and productivity. Proponents of income redistribution will not easily abandon the search for a "silver bullet" that will generate economic growth without the necessity of providing incentives for investment in tangible assets and human capital. Advocates of growth strategies based on capital formation will not readily give credence to claims of the importance of external benefits that "spill over" to beneficiaries that are difficult or impossible to identify.

To avoid the semantic confusion that pervades popular discussions of economic growth it is essential to be precise in distinguishing between investment and productivity. Investment is the commitment of current resources in the expectation of future returns and can take a multiplicity of forms. The distinctive feature of investment as a source of economic growth is that the returns can be internalized by the investor. The most straightforward application of this definition is to investments that create property rights, including rights to transfer the resulting assets and benefit from incomes that accrue to the owners.

Empirical research has gradually broadened the meaning of sources of growth to include investments that do not create property rights. For example, a student enrolled in school or a worker participating in a training program can be viewed as an investor. Although these investments do not create assets that can be bought or sold, the returns to higher educational qualifications or better skills in the workplace can be internalized. The contribution of investments in education and training to economic growth can be identified in the same way as for tangible assets.

The mechanism by which tangible investments are translated into economic growth is well understood. For example, an investor in a new industrial facility adds to the supply of assets and generates a stream of rental income. The investment and the income are linked through markets for capital assets and capital services. The income stream can be divided between the increase in capital input and the marginal product of capital or rental price. The increase in capital contributes to output growth in proportion to the marginal product.

Similarly, an individual who completes a course of education or training adds to the supply of people with higher qualifications or skills. The resulting income stream can be decomposed into a rise in labor input and the marginal product of labor or wage rate. The increase in labor contributes to output growth in proportion to the marginal product. Although there are no asset markets for human capital, investments in human and nonhuman capital have the common feature that returns are internalized by the investor.

The defining characteristic of productivity as a source of economic growth is that the incomes generated by higher productivity are external to the economic activities that generate growth. These benefits "spill over" to income recipients not involved in these activities, severing the connection between the creation of growth and the incomes that result. Since the benefits of policies to create externalities cannot be appropriated, these policies typically involve government programs or activities supported through public subsidies.

Publicly supported research and development programs are a leading illustration of policies to stimulate productivity growth. These programs can be conducted by government laboratories or financed by public subsidies to private laboratories. The justification for public financing is most persuasive for aspects of technology that cannot be fully appropriated, such as basic science and generic technology. The benefits of the resulting innovations are external to the economic units conducting the research and development.

Similarly, public investments in infrastructure can be justified by appealing to externalities associated with the use of public facilities. Improvement of public roads may relieve traffic congestion, but the returns can be appropriated only by limiting access to public facilities or monitoring the sue of these facilities. This can be done for bridges, tunnels, and high density highways, but is more difficult for urban streets and low density highways. If access is unlimited, the benefits are external to the authority undertaking the investment.

Tangible investments are the most important sources of postwar U.S. economic growth. These investments appear on the balance sheets of firms, industries, and the nation as a whole as buildings, equipment, and inventories. The benefits of these investments appear on the income statements of these same economic units as profits, rents, and royalties. Investments in tangle assets are the primary focus of chapters 1-5 and 9 below. My empirical studies of these investments and tax policies that affect them are presented in two companion volumes, Capital Theory and Investment Behavior and Tax Policy and the Cost of Capital.

Investments in human capital, especially through formal education, provide highly significant sources of postwar U.S. economic growth and are the focus of chapters 6-8. These investments do not appear on the balance sheets of individuals receiving the education or the institutions providing it. However, increases in labor incomes make it possible to measure the investments and assess their contributions to economic growth. Chapter 1 summarizes the evidence that investments in tangible assets and human capital, taken together, account for the great preponderance of U.S. economic growth.

The starting point for my research on postwar U.S. economic growth was the aggregate production function employed by Paul Douglas (1928, 1948) and his associates. Tinbergen (1942) took a critical step beyond Douglas 's original concept by adding a time trend interpreted as the level of "efficiency." Tinbergen analyzed the sources of U.S. economic growth over the period 1870-1914 and found that productivity accounted for slightly more than a quarter of U.S. economic growth during the period 1870-1914, while growth in capital and labor inputs accounted for about three-quarters. These proportions are very similar to those for postwar U.S. economic growth summarized in Chapter 1.

The notion of efficiency or total factor productivity was introduced independently by George Stigler (1947) and became the starting point for a major research program at the National Bureau of Economic Research. The National Bureau program resulted in important contributions by Moses Abramovitz (1956) and Solomon Fabricant (1959) and culminated in the monograph by John Kendrick (1961a), Productivity Trends in the United States . The conceptual framework of Douglas and Tinbergen was combined with data generated by Kendrick (1956) in Robert Solow's (1957) justly celebrated article, "Technical Change and the Aggregate Production Function."

The distinction between substitution and technical change emphasized by Solow parallels the distinction between investment and productivity as sources of economic growth. However, Solow's definition of investment was limited to tangible assets. He specifically excluded investments in human capital by omitting substitution among different types of labor inputs. Furthermore, Solow identified the contribution of tangible assets with increases in the stock, omitting substitution among different types of capital inputs. As a consequence, Solow attributed almost all of U.S. economic growth to "residual" growth in productivity.

My paper, "The Embodiment Hypothesis," published in 1966 and reprinted as chapter 2 of this volume, took an initial step beyond the aggregate production function. I introduced a production possibility frontier, allowing for joint production of consumption and investment goods from capital and labor services, and employed this framework in generalizing Solow's (1960) concept of embodied technical change. I showed that economic growth could be interpreted, equivalently, as "embodied" in investment or "disembodied" in productivity growth.

My paper with Zvi Griliches, "The Explanation of Productivity Change," published in 1967 and reprinted as Chapter 3, identified embodiment empirically be introducing constant quality price indices for investment goods. This removed the indeterminacy in Solow's (1960) model. As a natural extension of Solow's (1956) one sector neoclassical model of economic growth, his model of embodiment had only a single output and did not allow for the introduction of a separate price index for investment goods.

Constant quality price indices for capital goods of different ages or vintages were developed by Robert Hall (1971). This important innovation made it possible for Hulten and Frank Wykoff (1981) to estimate relative efficiencies by age for all types of tangible assets included in the U.S. National Income and Product Accounts. The Hulten-Wykoff relative efficiencies were incorporated into capital stocks that I employed in the introductory chapter to this volume, "Productivity and Postwar U.S. Economic Growth."

For each type of investment Griliches and I defined the price of capital input as a rental price rather than an asset price, using a model of capital as a factor of production I had introduced in 1963. This made is possible to measure the marginal products of capital inputs directly. We incorporated differences in returns due to the tax treatment of capital income into a constant quality index of capital input and modeled the flow of capital services, relative to the stock, as the consequence of substitution among capital inputs.

Finally, Griliches and I combined different types of labor inputs, reflecting wage rates for workers with different levels of educational attainment, into a constant quality index of labor input. The methodology was based on a similar index that Griliches (1960) had developed for U.S. agriculture. Our index of labor input became the point of departure for a far more elaborate set of constant quality indices of labor input that I subsequently constructed with Frank Gollop (1980, 1983). These included characteristics of workers such as age, sex, occupation, class of employment and industry, as well as educational attainment.

In measuring productivity growth Griliches and I allowed for substitution between investment and consumption goods as outputs and capital and labor services as inputs. Our model of capital as a factor of production provided the conceptual framework for modeling substitution among capital inputs and permitted a detailed treatment of differences in returns on tangible assets. Finally, the parallel model of labor input encompassed substitution among labor inputs from workers with different levels of educational attainment.

The conclusion of the first phase of my research on postwar U.S. economic growth was that investments in human and nonhuman capital greatly predominated as sources of growth, while productivity played a clearly subordinate role. These findings were confirmed in my subsequent research with Laurits Christensen, described below, my 1987 book with Gollop and Fraumeni, Productivity and U.S. Economic Growth, summarized in Chapter 1, and my 1992 paper with Gollop, reprinted as chapter 9.

The second stage of my research on postwar U.S. growth began with two papers I published in 1969 and 1970 with Laurits Christensen. These studies provided a much more detailed implementation of the concept of capital as a factor of production. We utilized a model of the tax structure for corporate capital income that I had developed in a series of papers with Hall (1967, 1971), reprinted in the volume Tax Policy and the Cost of Capital . Christensen and I extended this model to noncorporate and household capital incomes in order to capture differences in returns due to taxation.

In 1973 Christensen and I incorporated estimates of the sources of economic growth into a complete system of U.S. national accounts in our paper, "Measuring Economic Performance in the Private Sector," reprinted as Chapter 5 of this volume. The critical innovation in this accounting system was the construction of internally consistent income, product, and wealth accounts. The integration of income and product accounts is the major achievement of the United Nations System of National Accounts; however, this system does not include wealth accounts that are consistent with the income and product accounts.

Christensen and I distinguished two approaches to the analysis of economic growth. We identified the production possibility frontier with the production account. We identified a social welfare function with the income and expenditure account. We utilized data on inputs and outputs of production to allocate sources of economic growth between investment and productivity. We divided the uses of economic growth between current consumption and future consumption through saving by means of data on incomes and expenditures.

Saving is linked to the asset side of the wealth account through capital accumulation equations for each type of asset. These equations provide a perpetual inventory of assets accumulated at different points of time or different vintages. Prices for different vintages are linked to rental prices of capital inputs through a parallel set of capital asset pricing equations. The complete system of vintage accounts gives stocks of assets of each vintage and their prices. The stocks are cumulated to obtain asset quantities, while the prices are used to evaluate the stocks and derive rental prices.

Christensen and I implemented our vintage accounting system for the U.S. on an annual basis for the period 1929-1969. We constructed an internally consistent system of income, product, and wealth accounts, paralleling the U.S. National Income and product Accounts. This generated the information required to implement both production and welfare approaches to the analysis of economic growth. Our system of vintage accounts also provides the conceptual underpinnings for measurements of investment in human capital described in more detail below.

An important objective of the Christensen-Jorgenson accounting system was to provide the data for econometric modeling of producer and consumer behavior. In collaboration with Lawrence Lau, Christensen and I introduced an econometric model of producer behavior based on the translog production possibility frontier in a paper published in 1973 and reprinted in the companion volume, Econometrics and Producer Behavior . We modeled joint production of consumption and investment goods from inputs of capital and labor services, utilizing data on these outputs and inputs from the production account.

Subsequently, Christensen, Lau, and I constructed an econometric model of consumer behavior based on the translog indirect utility function. We estimated this model on the basis of data from the income and expenditure account of the Christensen-Jorgenson accounting system. Diewert (1976) showed that the Tornqvist (1936) index numbers employed in this system are "exact" for the translog production possibility frontier and indirect utility function.

Yun and I constructed a complete econometric model for postwar U.S. economic growth in a paper published in 1986 and reprinted in the companion volume, Tax Policy and the Cost of Capital . We based our model of producer behavior on the translog production possibility frontier and our model of consumer behavior on the translog indirect utility function. Finally, we derived a social welfare function for evaluating alternative tax policies and analyzed the consequences of the Tax Reform Act of 1986 on U.S. economic growth in a paper published in 1990 and also reprinted in Tax Policy and the Cost of Capital.

Griliches and I utilized the production estimates from my papers with Christensen for the period 1950-1962 in our reply to a critique of our 1967 paper by Denison (1969). This reply was published in 1972 and is reprinted as chapter 4 of this volume. We showed that Denison had confounded the production and welfare approaches to economic growth. He employed the concept of income, constructed from the welfare approach, as a measure of output. This concept is appropriate for analyzing the uses of economic growth, but inappropriate for analyzing the sources of growth.

The most important innovation of the second phase of my research on postwar U.S. economic growth was the development of two separate perspectives on the analysis of economic growth. The production approach requires data on outputs and inputs and allocates the sources of growth of output between investment and productivity. The welfare approach requires data on income and expenditure and allocates the uses of growth of income between consumption and saving. Once these two approaches are distinguished, the data to implement them can be generated within the accounting system Christensen and I have presented in Chapter 4.

The implementation of the welfare and production approaches to economic growth required the solution of a long-standing problem in national accounting. This is the integration of wealth accounts with income and product accounts. The solution of this problem was achieved by means of the vintage accounting system presented in my paper with Christensen. This system provided the data on assets and asset prices that underlie production, income and expenditure, and wealth accounts.

The third stage in my research on postwar growth was to disaggregate the sources of U.S. economic growth to the level of individual industries. This was accomplished in my papers with Gollop (1980, 1983) and Fraumeni (1980, 1986). These papers dispensed with the aggregate production function, giving value added as a function of capital and labor inputs as well as productivity. We defined industry output as a function of productivity and capital, labor, and intermediate inputs, avoiding the artificial construct of value added at the industry level employed, for example, by Kendrick (1961a).

Gollop and I (1980) presented production accounts for fifty-one industrial sectors of the U.S. economy, including output for each sector and capital, labor, and intermediate inputs in both current and constant prices. Intermediate input consists of goods and services produced by other sectors, while capital and labor inputs are defined in the same way as at the aggregate level. We disaggregated intermediate input by industry of origin, integrating the U.S. inter-industry transactions accounts with the U.S. national product accounts.

Gollop and I disaggregated the model of capital services employed in my work with Christensen to include twenty categories of producers' durable equipment and fourteen categories of nonresidential structures and generated this information for individual industries. We also disaggregated labor input by two sexes, eight age groups, five educational attainment levels, two employment classes, and ten occupational categories within each sector. Our data on labor input combined the establishment surveys underlying the U.S. National Income and Product Accounts with household surveys from the Census of Population and the Current Population Survey.

My 1980 paper on "Accounting for Capital" extended the vintage accounting system I had developed with Christensen to include both sectoral and aggregate production accounts. Fraumeni and I implemented this new accounting system for the U.S., covering the period 1948-1973. In 1987 we published updated data on sectoral and aggregate production accounts in our book with Gollop, Productivity and U.S. Economic Growth This book contains by far the most detailed data ever compiled on productivity in the U.S. economy. The results are summarized in Chapter 1.

Given the relative importance of intermediate input in the value of output in most industrial sectors, it is not surprising that this is the predominant source of growth at the sectoral level, exceeding both productivity growth and the contributions of capital and labor inputs. The contributions of these two inputs are substantially more important than produc-tivity growth for almost every sector. The data from Productivity and U.S. Economic Growth have been updated through 1985 in my 1992 paper with Gollop, reprinted as Chapter 9.

My 1992 paper with Gollop is devoted to comparisons between sources of growth in the outputs of U.S. agriculture and nonagricultural industries over the period 1947-1985. This comparison showed that agriculture is an important exception to the findings presented in Chapter 1. Productivity clearly predominates among the sources of growth of this industry, accounting for eighty-two percent of postwar growth. By contrast productivity accounts for only thirteen percent of growth in the private nonfarm sector of the U.S. economy.

My 1980 paper also presented a methodology for aggregating over sectors. The existence of an aggregate production function imposes very stringent conditions on production patterns at the industry level. In addition to value added function for each sector, an aggregate production function posits that these functions must be identical. Furthermore, the functions relating sectoral capital and labor inputs to their components must be identical and each component must receive the same price in all sectors. Aggregation over sectors has also been discussed by Evsey Domar (1961) and Hulten (1978).

The data for sectoral production accounts can be generated in a way that avoids the highly restrictive assumptions that underlie the aggregate production function. These data can then be compared with those from the aggregate production account to test for the existence of an aggregate production function. In chapter 1 I show that this hypothesis is inconsistent with empirical evidence. Gollop, Fraumeni, and I have also presented statistical tests of the much weaker hypothesis that value added functions exist for industrial sectors, but this hypothesis is also rejected.

An important objective of my book with Gollop and Fraumeni was to generate data for econometric modeling of producer behavior at the sectoral level. In 1981 I developed a general equilibrium model of production with Fraumeni, including econometric models of production for thirty-five industrial groups. These models represented the rate of productivity growth in each sector as well as the technical coefficients as functions of relative prices. Production models with these features were incorporated into an econometric model of the U.S. economy I constructed with Peter Wilcoxen and presented in papers reprinted in the accompanying volume, Econometric General Equilibrium Modeling .

The conclusion of my research on production at the sectoral level is that specifications of technology such as the aggregate production function and sectoral value added functions result in substantial oversimplifications of the empirical evidence. However, these specifications are useful for particular but limited purposes. The aggregate production function is a worthwhile simplification in modeling long-run growth at the aggregate level and sectoral value added functions are essential for aggregating over sectors.

A satisfactory analysis of the sources of postwar U.S. economic growth at the sectoral level requires a flexible approach. Rather than imposing the assumptions required for an aggregate production function, capital, labor, and intermediate inputs must be treated symmetrically. This framework is also the most appropriate starting point for econometric modeling of producer behavior at the sectoral level. Studies presenting models of this type are summarized in my 1986 survey paper, "Econometric Methods for Modeling Producer Behavior," reprinted as Chapter 1 in the companion volume, Econometrics and Producer Behavior." reprinted as Chapter 1 in the companion volume, Econometrics and Producer Behavior.

An important limitation of the framework of Abramovitz, Kendrick, and Solow is that the implied definition of investment specifically excludes investments in human capital. Increases in labor inputs resulting from investments by individuals in their own education and training are allocated to productivity growth. This has some intuitive appeal within a framework based on welfare rather than production, since effort is related to hours worked at the individual level. However, a constant quality measure of labor input is essential for the production approach to economic growth.

Constant quality indices of labor input were developed by Griliches (1960) for U.S. agriculture and Denison (1962) for the U.S. economy as a whole. My 1967 paper with Griliches presented a constant quality index for labor input focusing on differences in educational attainment among workers. My subsequent papers with Gollop (1980, 1983) gave constant quality indices of labor input for fifty-one industrial sectors of the U.S. economy. We disaggregated labor input for each industry by age, sex, educational attainment, class of employment, and occupation. This required the creation of an extensive data base giving hours worked and hourly compensation for individual categories of labor input.

Constant quality indices of labor input are an essential first step in incorporating investments in human capital into empirical studies of economic growth. However, completion of this task requires the measurement of these investments as an output of the U.S. economy. In 1989 Fraumeni and I extended the vintage accounting system developed in my work with Christensen to incorporate investments in human capital in a paper reprinted as Chapter 6 below. Our essential idea was to treat individual members of the U.S. population as human assets with "prices" given by their lifetime labor incomes.

The starting point for our measurements of lifetime labor incomes was the data base on market labor activities assembled in my papers with Gollop (1980, 1983). We first integrated our labor data base with a system of demographic accounts for the U.S. population constructed by Marilyn McMillen and Kenneth Land (1980). Our system of demographic accounts included the enrollment status for all individuals between five and thirty-four years of age and employment status for individuals between fourteen and seventy-four years of age.

We derived annual estimates of hours worked and hourly labor compensation classified by two sexes, individual years of age, and individual years of educational attainment. For each of these categories we constructed estimates of lifetime incomes from market labor activities. We then estimated hours devoted to nonmarket activities by allocating the total time available among work, schooling, household production and leisure, and maintenance. Finally, we imputed hourly labor compensation for nonmarket activities by reducing market wage rates by marginal tax rates on labor incomes and estimated lifetime incomes for nonmarket activities as well.

Fraumeni and I implemented our vintage accounting system for both human and nonhuman capital for the U.S. on an annual basis for the period 1948-1984. Asset prices for tangible assets can be observed directly from market transactions in investment goods; capital asset pricing equations are used to derive rental prices for capital services. For human capital wage rates correspond to rental prices and can be observed directly from transactions in the labor market. Lifetime labor incomes are derived by applying asset pricing equations to these wage rates. These incomes are analogous to the asset prices used in accounting for tangible assets.

Given our vintage accounts for human and nonhuman capital, Fraumeni and I constructed a system of income, product, and wealth accounts, paralleling the system I had developed with Christensen. In these accounts the value of human wealth was more than ten times the value of nonhuman wealth, while investment in human capital was five times investment in tangible assets. We defined "full" investment in the U.S. economy as the sum of these two types of investment. Similarly, we added the value of nonmarket labor activities to personal consumption expenditures to obtain "full" consumption. Our product measure included these new measures of investment and consumption.

Since our complete accounting system included a production account with "full" measures of capital and labor inputs, we were able to generate a new set of accounts for the sources of U.S. economic growth. Our system also included an income and expenditure account with income from labor services in both market and nonmarket activities. We combined this with income from capital services and allocated "full" income between consumption and saving. This provided a new set of accounts for the uses of U.S. economic growth. Our system was completed by a wealth account containing human wealth and tangible assets.

In 1992 Fraumeni and I developed a measure of the output of the U.S. education sector in a paper reprinted as Chapter 7. Our point of departure was that while education is a service industry, its output is investment in human capital. We estimated investment in education from the impact of increases in educational attainment on the lifetime incomes of all individuals enrolled in school. We found that investment in education, measured in this way, is similar in magnitude to the value of working time for all individuals in the labor force. Furthermore, the growth of investment in education during the postwar period exceeded the growth of market labor activities.

Second, we measured the inputs of the education sector, beginning with the purchased inputs recorded in the outlays of educational institutions, in a paper reprinted as chapter 8. A major part of the value of the output of educational institutions accrues to students in the form of increases in their lifetime incomes. Treating these increases as compensation for student time, we evaluated this time as an input into the educational process. Given the outlays of educational institutions and the value of student time, we allocated the growth of the education sector to its sources.

We aggregated the growth of education and noneducation sectors of the U.S. economy to obtain a new measure of U.S. economic growth. Combining this with measures of input growth, we obtained a new set of accounts for sources of growth of the U.S. economy. Productivity contributes almost nothing to the growth of the education sector and only a modest proportion to output growth for the economy as a whole. Our overall conclusion is that investment in human and nonhuman capital accounts for an overwhelming proportion of postwar U.S. economic growth.

The accumulation of empirical evidence presented in this volume has gradually shifted the terms of the professional debate over the important of investment and productivity as sources of postwar U.S. economic growth. My paper, "Productivity and Postwar U.S. Economic Growth," reprinted as chapter 1 of this volume, summarizes the findings in my 1987 book with Frank Gollop and Barbara Fraumeni, Productivity and U.S. Economic Growth . This chapter shows that productivity accounts for less than a quarter of the growth of U.S. economy during the period 1948-1979, while growth of capital and labor inputs accounts for more than three-quarters.

I would like to thank June Wynn of the Department of Economics at Harvard University for her excellent work in assembling the manuscripts for this volume in machine-readable form. Renate d'Arcangelo of the Editorial Office of the Division of Applied Sciences at Harvard edited the manuscripts, proofread the machine-readable versions and prepared them for typesetting. Warren Hrung, a senior at Harvard College, checked the references and proofread successive versions of the typescript. Gary Bisbee of Chiron Inc. Typeset the manuscript and provided the camera-ready copy for publication. The staff of the MIT Press, especially Terry Vaughn, Ann Sochi, and Michael Sims, has been very helpful at every stage of the project. I am also grateful to William Richardson and his associates for providing the index. Financial support from the Program on Technology and Economic Policy at the Kennedy School of Government, Harvard University, is gratefully acknowledged. As always, the author retains sole responsibility for any remaining deficiencies in the volume.