Dale Jorgenson

Samuel W. Morris University Professor

Investment: Volume One -- Preface

This is the first of two volumes containing my empirical studies of investment behavior and is devoted to the cost of capital as a determinant of investment expenditures. The second volume, entitled Tax Policy and the Cost of Capital, focuses on the cost of capital approach to the analysis of tax policy. The model of capital as a factor of production for introduced in my 1963 paper, "Capital Theory and Investment Behavior," provides the unifying framework for the two volumes.

This volume contains the studies of investment spending cited by the American Economic Association in awarding me the John Bates Clark Medal in 1971. The citation reads in part:

Dale Jorgenson has left his mark with great distinction on pure economic theory (with, for example, his work on the growth of a dual economy): and equally on statistical method (with, for example, his development of estimation methods for rational distributed lags). But he is preeminently the master of the territory between economics and statistics where both have to be applied to the study of concrete problems. His prolonged exploration of the determinants of investment spending, whatever the ultimate lessons, will certainly long stand as one of the finest examples in the marriage of theory and practice in economics.

My 1963 paper, reprinted as chapter 1 below, introduced all the important features of the econometric models of investment behavior summarized in this volume. The key innovation of this paper was a concept of the cost of capital incorporating the tax treatment of income from capital. The paper also contained a derivation of investment demand from a model of capital as a factor of production and introduced new econometric methods for modeling gestation lags in the investment process.

I discussed statistical methods for modeling gestations lags in my 1966 paper, "Rational Distributed Lag Functions," reprinted as chapter 3 below. I analyzed relationships among capital as a factor of production, the cost of capital, and investment expenditures in my 1967 paper, "The Theory of Investment Behavior," reprinted as chapter 6. The link between tax policy and investment expenditures is the subject of the companion volume, Tax Policy and the Cost of Capital .

The cost of capital has become an indispensable analytical tool for studying the dynamics of investment behavior. Both macroeconometric models and intertemporal general equilibrium models have employed the cost of capital as a determinant of investment expenditures. Macroeconometric models are especially useful in capturing the short-run dynamics of investment in response to changes in the cost of capital, while general equilibrium models are essential for capturing long-run dynamics.

Widespread applications of the cost of capital are due to the fact that this concept summarizes information about future consequences of investment in a highly succinct form. The closely related concept of the marginal effective tax rate, discussed in detail in Tax Policy and the Cost of Capital, facilitates the representation of economically relevant features of complex tax statutes and greatly enhances the transparency of tax rules. The cost of capital and the marginal effective tax rate characterize both economic and tax consequences of investment decisions.

In chapter 1 I derived the demand for capital services from the neoclassical theory of optimal capital accumulation. As constraints the model included a production function with capital as a factor of production and an accumulation equation relating capital stock to past investments. I identified the accumulation equation and the cost of capital as the key elements in modeling the dynamics of investment.

I defined the user cost of capital as the rental price of capital services, representing this price as the product of the price of investment goods and the cost of capital. My formulation of the cost of capital and the rate of capital loss on investment goods. The price of capital services also incorporated several features of the U.S. tax structure for capital income -- the corporate tax rate, depreciation for tax purposes, and tax deductibility of interest.

An important feature of the econometric model of investment expenditures presented in chapter 1 was gestation lags between the lags among intermediate stages of the investment process corresponding to anticipations of investment expenditures by business firms. I estimated the gestation lags at each stage of the process by representing them as distributed lag functions, based on the new econometric methodology I had designed for this purpose.

I implemented the econometric model of investment behavior presented in my 1963 paper for quarterly data covering the period 1948-1960 for U.S. total manufacturing. I employed the fitted investment functions to characterize gestation lags in the investment process, including intermediate stages of this process. Finally, I presented estimates of short-run and long-run responses to changes in market conditions and tax policy changes. In 1982 my paper was identified as a Citation Classic, based on 195 citations in the Social Sciences Citation Index.

My 1965 paper, "Anticipations and Investment Behavior," reprinted in chapter 2, gave investment functions for four industries, covering the whole of the U.S. economy. I separated total manufacturing into durables and nondurables industries. I also modeled investment in regulated and all other industries. My paper gave additional details on each component of the econometric model of investment I had introduced in 1963. Although the model design and analysis of results were similar, important differences among industries emerged from the study.

Empirical implementation of the theory of investment derived in my 1963 and 1965 papers required estimates of rates of replacement and depreciation. This was the focus of my 1974 paper, "The Economic Theory of Replacement and Depreciation," reprinted in Tax Policy and the Cost of Capital . I presented a dual theory of replacement investment and economic depreciation and surveyed the empirical literature. I concluded that depreciation at a constant geometric rate provided a satisfactory approximation for modeling investment behavior.

A constant rate of depreciation and replacement greatly simplifies the neoclassical theory of optimal capital accumulation. Economic depreciation is proportional to the price of investment goods, while replacement investment is proportional to the stock of capital. My introduction, "Capital as a Factor of Production," to Technology and Capital Formation, the 1989 book that I edited with Ralph Landau, surveyed more recent evidence on replacement and depreciation supporting this formulation of the theory.

Empirical implementation of the neoclassical theory of optimal capital accumulation also required an explicit representation of the production function. I assumed that the production function is Cobb-Douglas with elasticity of substitution equal to unity and constant returns to scale. In my 1973 paper, "Investment and Production," reprinted in Tax Policy and the Cost of Capital , I showed that both assumptions were consistent with a broad range of empirical evidence from econometric studies of production.

My 1966 paper, "Rational Distributed Lag Functions," reprinted as chapter 3, introduced a new econometric methodology particularly suitable for modeling gestation lags in the investment process. The key innovation was flexibility in representing both autoregressive and moving average components of distributed lag functions. This representation was later assimilated into the statistical literature through the "transfer function" approach of Box and Jenkins (1970). Box and Jenkins allowed for considerably greater generality in representing the error structure.

My 1967 paper, "The Theory of Investment Behavior," reprinted as chapter 6, compared the neoclassical theory of optimal capital accumulation with alternative theories of investment. These included the theory presented by John Maynard Keynes in his 1936 classic, The General Theory of Employment, Interest, and Money. Keynes' theory was based on the concept of the marginal efficiency of investment and had become the subject of a voluminous critical literature. Trygve Haavelmo (1960a) was particularly forceful in articulating the consensus that had emerged. This was that the Keynesian marginal efficiency schedule, expressing investment as a function of the rate of return, could not be derived by static profit maximization.

I sided with Haavelmo against Keynes, but argued that a dynamic theory of investment was implied by the accumulation equation of the neoclassical theory. Comparing alternative paths of optimal capital accumulation, I showed that investment could be expressed as a downward-sloping function of the rate of return. I also discussed an alternative approach advanced by Haavelmo , giving investment as a function of the rate of return in a market equilibrium balancing the demand and supply of investment goods. I argued that the missing element in h\this approach was an accumulation equation generating investment demand.

My 1973 paper, reprinted in Tax Policy and the Cost of Capital , presented a model of investment with internal adjustment costs and irreversibility, combining features of models originated by Kenneth J. Arrow (1962) and Robert E. Lucas (1967). Capital services were allocated between installation of capital and the production of marketable output. Internal adjustment costs for investment arose from the loss of output. The cost of capital was a shadow price that reflected both the market price of investment and the shadow value of installation.

In chapters 4, 5, and 9 James A. Stephenson and I gave investment functions for fifteen two-digit manufacturing industries in the U.S. -- eight durable goods and seven nondurable goods industries. This greatly disaggregated the econometric model of investment presented in my 1963 paper and added considerable detail to my 1965 models for durables and nondurables manufacturing. In chapter 12 Sidney S. Handel and I provided a parallel disaggregation of my 1965 model for regulated industries, communications, and other transportation industries, as well as all the regulated industries taken together.

In chapter 4 Stephenson and I analyzed gestation lags in the investment process for each of the fifteen manufacturing industries. In chapter 5 we tested the hypothesis that aggregation errors were absent in more highly aggregated models and rejected this hypothesis. In chapter 9 we tested the consistency of models of gestation lags for the investment process as a whole and intermediate stages and accepted the hypothesis that these models were consistent. My paper with Handel in chapter 12 gave similar results for the regulated industries.

The cost of capital that Stephenson and I analyzed gestation lags in the investment process for each of the fifteen manufacturing industries. In chapter 5 we tested the hypothesis that aggregation errors were absent in more highly aggregated models and rejected this hypothesis. In chapter 9 we tested the consistency of models of gestation lags for the investment process as a whole and intermediate stages and accepted the hypothesis that these models were consistent. My paper with Handel in chapter 12 gave similar results for the regulated industries.

The cost of capital that Stephenson and I employed in chapters 4 and 5 included a weighted average rate of return, like that proposed by Franco Modigliani and Merton Miller (1958), in place of the long-term bond rate used in chapters 1 and 2. This innovation established an important connection between the cost of capital and the theory of corporate finance. In the terminology of Modigliani and Miller, the "cost of capital" referred to this weighted average rate of return, while I reserved the term "cost of capital" for the sum of the rate of return the rate of depreciation, and the rate of capital loss, adjusted for the taxation of capital income.

In chapter 7 Calvin D. Siebert and I compared alternative econometric models of investment for fifteen major U.S. manufacturing corporations. We concluded that neoclassical models provided better explanations of corporate investment than models based on accelerator, liquidity, and expected profits principles. In chapter 8 we analyzed two alternative specifications of the neoclassical model in greater detail. Both included a weighted average cost of capital, but differed in the treatment of inflation in the prices of investment goods. Our comparisons showed that the assumption of perfect foresight or rational expectations of inflation was the most appropriate formulation of the neoclassical model.

Expected profits subsequently emerged as a component of Tobin's q , where expected profits were identified with the market value of the firm's outstanding securities. Following William Brainard and James Tobin (1968) and Tobin (1969), Tobin's q is defined as the ratio of expected profits to the market value of the firm's assets. Fumio Hayashi (1982) provided a neoclassical interpretation of Tobin's q by showing how to identify internal costs of adjustment from the q-ratio . Costly adjustment is an alternative to gestation lags in representing the short-fun dynamics of investment behavior.

In chapters 10 and 11 Jerald Hunter, M. Ishaq Nadiri and I compared the models of investment I had presented in chapters 4 and 5 with Stephenson with alternative models for the same fifteen manufacturing industries. These models has been constructed by Locke Anderson (1964), Robert Eisner (1965), and John Meyer and Robert Glauber (1964). Eisner's model was based on the accelerator principle and was consistent with a production function characterized by an elasticity of substitution equal to zero, rather than unity, as Stephenson and I had assumed.

Anderson and Meyer and Glauber focused on sources of finance, rather than the specifications of technology. Both had assumed that retained earnings were the least costly course of finance. This assumption re-emerged in the "new view" of taxation and corporate finance introduced by Mervyn King (1977). By contrast the neoclassical model incorporating a weighted average rate of return was later identified with "traditional view" of taxation and corporate finance by my 1991 book with Kun-Young Yun , Tax Reform and the Cost of Capital .

My book with Yun presented a complete theory of corporate finance and taxation, including corporate and personal taxes on income from capital. We extended this theory to cover the taxation of income from capital in noncorporate and household sectors, which are not subject to the corporate income tax. We showed that Tobin's q is identically equal to one in the traditional view of corporate finance, so that the expected profits model reduces to the neoclassical model.

The comparisons of alternative models presented in chapter 10 showed that the neoclassical models I had constructed with Stephenson were clearly superior to the three alternative models. However, the neoclassical models had been constructed from the data set used in the comparisons. In chapter 11 Hunter, Nadiri , and I assessed the potential role of "data mining" in these comparisons by carrying out predictive tests for all four models, using four additional years of quarterly data covering the period 1961-1964. These tests showed no evidence of data mining for the neoclassical model or the accelerator model of Eisner, but revealed considerable evidence of data mining for the Anderson and Meyer and Glauber models.

I summarized research on the cost of capital as a determinant of investment expenditures in my 1971 survey article, "Econometric Studies of Investment Behavior," reprinted in chapter 13. This paper surveyed the extensive body of research on investment behavior for industrial sectors of the U.S. economy. This included my papers with Stephenson, reprinted as chapters 4, 5, and 9, and Handel, reprinted as chapter 12. I also surveyed research on investment by individual firms, including my papers with Siebert, reprinted as chapters 7 and 8.

In chapter 13 I compared alternative specifications of models of investment at the industry level. The comparison included determinants of the demand for capital, the role of gestation lags in the investment process, and modeling of replacement investment. I concluded that a flexible representation of distributed lags was essential for describing the short-run dynamics of investment. I also identified output and the cost of capital as key determinants of the demand for capital; there was little support for internal funds or liquidity. Finally, I concluded that replacement investment was proportional to capital stock.

I had proposed the investment functions presented in my 1965 paper for the Brookings quarterly macroeconometric model of the United States . My 1971 paper and my 1976 paper co-authored with Roger H. Gordon, both reprinted in Tax Policy and the Cost of Capital , presented detailed simulations of the economic impacts of tax policy on investment. These papers employed the cost of capital as a vehicle for introducing changed in tax policy into the DRI quarterly macroeconometric model of the U.S. economy. An important issue in this type of application, emphasized by Lucas (1976), is modeling expectations about future tax policies and prices of investment goods.

By the early 1980s the cost of capital had been incorporated into investment functions for all major forecasting models of the U.S. economy. And simulations of the short-run economic impacts of alternative tax policies had become the staple fare of debates over specific proposals. (1) In chapters 7 and 8 with Siebert I had treated these expectations as rational. However, macroeconometric models, including the DRI model, did not incorporate this treatment of expectations into simulations of alternative policies. (2)

Simulations of the long-run impacts of tax policy based on the cost of capital made their appearance in tax policy debates by the end of the 1980s. (3) The cost of capital with rational expectations was incorporated into an intertemporal general equilibrium model in my 1986 papers with Kun-Young Yun , reprinted in Tax Policy and the Cost of Capital . Equilibrium in this model was characterized by an intertemporal price system that clears markets for capital and labor services and consumption and investment goods. Asset prices were set equal to present values of future services, incorporating rational expectations of future prices.

An important innovation in the models I constructed with Yun was external costs of adjusting the rate of investment. Yun and I represented technology by a production possibility frontier rather than a production function. We implemented this model of production econometrically, using the translog production possibility frontier introduced in my 1973 paper with Laurits Christensen and Lawrence Lau, reprinted in the companion volume, Econometrics and Producer Behavior . Outputs included consumption and investment goods, while inputs consisted of capital and labor services. (4)

Since production of investment goods required foregoing consumption goods output, adjusting the rate of investment was costly. However, the opportunity costs of adjustment were fully reflected in the market price of investment and were external to the adjustment process. Equilibrium required a balance between demands and supplies of investment goods -- precisely the formulation proposed by Haavelmo . However, the demand for investment goods was generated by the accumulation equation relating capital stock to past investment, as in the neoclassical theory.

Equilibrium also required a balance between demands and supplies of capital services, where the rental price of these services included the cost of capital. The markets for investment goods and capital services linked the past and the future through the backward-looking accumulation equation and the forward-looking asset pricing equation. This formulation encompassed all the elements of the neoclassical theory of optimal capital accumulation discussed in my 1967 paper and, in addition, external costs of adjustment of the rate of investment.

With costly adjustment the short-run dynamics of investment can be included among the constraints imposed in the optimization process. Long-run dynamics represented by the accumulation and asset-pricing equations then arise as a natural generalization of the neoclassical theory presented in my 1967 paper. With external adjustment costs the cost of capital depends only on the market price of investment goods. In models with internal costs of adjustment the cost of capital depends on Tobin's q , which also includes the market value of securities. This simplification is an important advantage of models with external costs.

A very comprehensive and detailed survey of empirical research on investment has been presented by Robert S. Chirinko in his 1993 paper, "Business Fixed Investment Spending: Modeling Strategies, Empirical Results, and Policy Implications." Chirinko distinguishes "models with implicit dynamics" summarized in chapter 13 from "models with explicit dynamics" constructed subsequently. This distinction parallels that between models with gestation lags developed within the framework presented in chapter 1 and models based on costly adjustment, like the intertemporal equilibrium model I constructed with Yun .

The neoclassical theory of optimal capital accumulation that provided the framework for my empirical studies of investment behavior was substantially extended in subsequent research surveyed by Chirinko . However, the dynamics of investment behavior has continued to be the central focus of this research. The dynamic perspective I introduced has been successfully integrated into tax policy analysis in the work summarized in the companion volume, Tax Policy and the Cost of Capital . This perspective has also been incorporated into corporate finance in the work reviewed in my 1991 volume with Yun , Tax Reform and the Cost of Capital .

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 Division of Applied Sciences at Harvard edited the manuscripts and proofread the machine-readable versions. Warren Hrung , then a senior at Harvard College , checked the references and proofread successive versions of the typescript. Gary Bisbee of Chiron Inc. Prepared the manuscripts for typesetting, typeset the volume, 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 Economy Policy of the Kennedy School of Government, Harvard University , is gratefully acknowledged. As always, the author retains sole responsibility for any remaining deficiencies in this volume.

Endnotes:

1. See, for example, Robert S. Chirinko and Eisner (1983) and Jane G. Gravelle (1984).

2. Techniques for perfect foresight or rational expectations were subsequently introduced by David Lipton, James Poterba , Jeffrey Sachs, and Lawrence Summers (1982) and Ray C. Fair and John B. Taylor (1983).

3. See, for example, the survey of general equilibrium modeling of the impacts of the Tax Reform Act of 1986 by Yolanda K. Henderson (1991).

4. Additional details are given in my 1986 survey, "Econometric Methods for Modeling Producer Behavior," also reprinted in Econometrics and Producer Behavior .