Tax policy and investment.
Modeling Investment: Back to the Future
Policymakers in the United States and other industrial economies evidently believe that business fixed investment responds strongly to tax changes (given the frequency with which governments manipulate tax policy parameters). Hence it is disturbing that models emphasizing the net return to investing - the "neoclassical" family of dynamic models emphasizing the role of the user cost of capital, or Q, or estimating the Euler equation for the choice of the capital stock - are defeated by ad hoc models in forecasting "horse races," and that structural variables frequently are found to be economically or statistically insignificant.
The problem is seen easily in aggregate data. Movements of aggregate variables, including investment, over the business cycle are determined simultaneously; disentangling the marginal impact of a single forcing variable is difficult. For example, an exogenous increase in aggregate demand might lead firms to be more optimistic about their sales prospects and to purchase more investment goods; it also might be expected at least in the short run to lead to higher interest rates. If we examine the correlation between investment and the interest rate, we might even find that the sign is the opposite of that predicted by the neoclassical theory. While an instrumental variables procedure might allow us to overcome this simultaneity problem, the estimator is only as good as the instruments, and good instruments are in short supply. Microeconomic data, however, provide a rich additional source of variation; my own work has focused on tests using firm-level data.
Conventional empirical tests of neoclassical models assume convex costs of adjusting the capital stock and attempt to estimate a parameter related to marginal adjustment costs.(1) Jason Cummins, Kevin Hassett, and I note that conventional estimated coefficients on fundamental variables in firm-level panel data for the United States and other countries are very small, implying implausibly large marginal costs of adjustment.(2) Such estimates imply very small effects of permanent investment incentives on investment.
In my research, I have focused on two general explanations of the failure to estimate significant tax effects on investment: measurement error in fundamental variables and misspecification of costs of adjusting the capital stock.
A major problem in using investment models based on Q or the user cost of capital to estimate effects of tax changes on investment is that measurement error in Q or the user cost of capital may bias downward the estimated coefficient. On a statistical level, Cummins, Hassett, and I estimate neoclassical models in firm-level data using first differences and longer differences (as opposed to the usual fixed-effects, within-group estimator) to address measurement error problems. We find lower adjustment costs and a greater response of investment to fundamentals) In other work, I depart from the strategy of using proxies for marginal Q and rely on the firms's Euler equation to model the investment decision. Using Compustat data for the United States, Anil Kashyap, Toni Whited, and I could not reject the frictionless neoclassical model for most firms, and the estimated adjustment cost parameters are more reasonable than those found in estimates of Q models. Very similar results are reported for European manufacturing firms by Cummins, Trevor Harris, and Hassett and for investment in overseas subsidiaries of U.S. multinational corporations by Cummins and myself.(4)
Again, one reason the data may not appear to favor neoclassical models over accelerator models is a simultaneous equations problem. To the extent that data incorporate exogenous changes in both the real interest rate and the intercept of the investment function, aggregate demand shocks may dominate the hypothesized negative relationship between investment and the user cost of capital. In this case, the estimated coefficient on the user cost of capital (or Q) will be "too small," leading to adjustment costs that are "too large." Such simultaneity increases apparent accelerator effects, because positive shifts of the investment function raise both investment and output. Conventional instrumental variables have not proved very helpful in addressing the simultaneity problem.
Cummins, Hassett, and I argue that major tax reforms offer periods in which there is exogenous cross-sectional variation in the user cost of capital or tax-adjusted Q. We demonstrate that major tax reforms also are associated with significant firm- and asset-level variation in key tax parameters (such as the effective rate of investment tax credit and the present value of depreciation allowances). Hence tax variables are likely to be good instruments for the user cost of Q during tax reforms. We estimate significantly greater responses of investment to the user cost of Q following U.S. tax reforms in 1962, 1971, 1981, and 1986 than in other periods; we also find significantly greater responsiveness of investment to fundamentals following tax reforms in 14 countries than that detected using firm-level panel data for those countries.(5)
Complications: Capital-Market Imperfections and Lumpy Investment
Not all firms face the frictionless capital markets I described for the neoclassical models. Therefore tests may not be able to ascertain whether the observed sensitivity of investment to financial variables differs across firms and whether these differences in sensitivity explain the weak apparent relationship between the measured user cost (or Q) and investment. My research in this area has integrated information and incentive problems in the investment process by moving beyond the assumption of representative firms by examining firm-level panel data in which firms can be grouped into high-net-worth and low-net-worth categories. For the latter category, changes in net worth or internal funds affect investment, holding constant underlying investment opportunities (desired investment). Following my work with Steven Fazzori and Bruce Petersen,(6) many empirical researchers have placed firms into groups as a priori financially constrained or financially unconstrained.
Two aspects of the conclusions of this research are noteworthy in the context of measuring incentives to invest. First, numerous empirical studies have found that proxies for internal funds have explanatory power for investment, holding constant Q, the user cost, or accelerator variables.(7) This suggests that tax policy may have effects on investment by constrained firms beyond those predicted by neoclassical approaches. (Indeed, returning to the accelerator analogy, Ben Bernanke, Mark Gertler, and Simon Gilchrist argue that this literature describes a "financial accelerator."(8) In particular, the quantity of internal funds available for investment is affected by the average tax on earnings from existing projects. In this sense, average as well as marginal tax rates faced by a firm can affect its investment decisions.(9)
Second, empirical studies by me and by others generally find that the frictionless neoclassical model is rejected only for the groups of firms that a priori are financially constrained.(10) Hence, while the shadow value of internal funds may not be well captured for some firms in standard representations of the neoclassical approach, the neoclassical model with convex adjustment costs yields reasonable estimates for most firms of the response of investment to fundamentals and to tax parameters.
The small estimated sensitivity of investment to fundamentals and tax variables in conventional empirical approaches led some researchers to suggest that adjustment costs may be nonconvex (one example being "irreversible" investment).(11) Hassett and I argue, however, that much of U.S. investment is in the form of capital goods with well-operating secondary markets.(12) Cummins, Hassett, and I also use firm-level data to investigate whether there was evidence of bunching of investment around tax reforms. We estimate transition probabilities among various ranges of investment rates over the year prior to, of, and after the tax reform and find no evidence that firms with large investment were likely to have lower investment in prior or subsequent years. Indeed, only a very tiny fraction of the sample was ever found to transit from high-investment to low-investment states.(13)
In part, the conclusions of such studies may differ because of differences in the level of aggregation. At a sufficiently fine level of disaggregation, all investment looks lumpy. The plant-level evidence suggests that investment appears lumpy, but the firm-level evidence does not corroborate this. However, there may be interesting differences between the investment behavior of plants and firms, as might be the case if, for example, managerial attention is limited and only a fraction of a firm's plants adjust their capital in a given year. Clearly, reconciling the plant-level and firm-level results is an important topic for future research.
Rethinking Tax Policy
My research suggests that tax incentives for investment are important components of the net return to investing and that the short-term and long-term responses of investment to permanent tax incentives are large. The deeper policy question remains: Would permanent investment incentives to increase the stock of business fixed capital raise economic welfare?
A scenario under which investment incentives might have an especially large impact on the quantity of investment without dissipation in the prices of investment goods is one in which firms' demand for capital is responsive to changes in the user cost of capital and in which capital goods are supplied perfectly elastically. While it is implausible that the supply function for most individual capital goods manufacturers is perfectly elastic, the effective supply of capital goods to a given domestic market might well be highly elastic in the long run if the world market for capital goods is open. Investment incentives would raise prices of capital goods in the short run if the supply of capital goods is highly inelastic.
Using data for the United States and ten other countries, Hassett and I find that local investment tax credits have a negligible effect on prices paid for capital goods - indeed, we find that the capital goods prices for most countries are very highly correlated and that the movements of these over time are consistent with the "law of one price." In addition, using disaggregated data on asset-specific investment good prices and tax variables for the United States, we find that tax parameters have no effect on capital goods prices.(14) Taken together, these tests suggest that the effects of investment tax policy have not been muted in a significant way by upward-sloping supply schedules for capital goods.
While it is instructive to ask how effective investment incentives are at increasing the fixed capital stock, a still more important question remains: What is the social value of the increase in the fixed capital stock?
Hassett and I review comparisons of "golden rule" levels of the capital stock or net investment relative to output to their actual values over the period from 1980 to 1994. For benchmark parameter values, equipment investment and capital stocks are below their "golden rule levels" (assuming 1980-94 is sufficiently long to characterize a steady state), while residential investment and the residential capital stock, which received significant tax subsidies over this time period, are near or above their golden rule levels.(15) Such findings suggest that, by raising the stock of equipment capital, investment incentives have positive social returns.
The finding of substantial short-term and long-term effects of the user cost of capital on business investment has applications for current policy debates. In particular, I have focused on the consequences of a reduction in inflation and a switch from an income tax to a broad-based consumption tax for user cost of capital and investment.
Many economists have argued that, under fairly general assumptions, a reduction in the rate of inflation provides a relatively costless stimulus to business fixed investment by reducing the user cost of capital. Darrel Cohen, Hassett, and I derive the effect of inflation on the user cost and investment under various assumptions about sources of financing and about the openness of capital markets. We estimate that, if the United States were a closed economy, a single-percentage point decline in inflation from its current level lowers the user cost by about 0.5 percentage points. The effect is smaller, but still economically significant, if one assumes that the United States is a price-taker on international debt markets.(16) All else being equal, this "tax cut" would, provide a stimulus to investment.
Under the income tax, the user cost of capital is influenced by the corporate tax rate, investment incentives, and the present value of depreciation allowances. Under a broad-based consumption tax, taxes do not distort business investment decisions. In addition, given current U.S. tax policy, the user cost is lower under a consumption tax than under an income tax. Hassett and I estimate that, all else being equal, a move to a consumption tax would stimulate the demand for equipment investment significantly.(17)
Of course, other aggregate variables are also likely to change if such a large change to the tax code were adopted. For example, nominal interest rates and the supply of savings are likely to change. While it is difficult to say how large the net stimulus to investment would be, the consensus of the recent investment literature suggests that the partial equilibrium impact on investment may be quite large.
My current work on investment focuses on two areas: studying how managers choose "hurdle rates" in evaluating investment projects and examining links between plant-level and firm-level investment decisions. The study of "investment" offers a lens through which to learn more about organizational decisionmaking and links between "financial" and "real" decisions.
1 See the review in K.A. Hassett and R. G. Hubbard, "Tax Policy and Investment," NBER Working Paper No. 5683, July 1996.
2 J.G. Cummins, K.A. Hassett, and R.G. Hubbard, "A Reconsideration of Investment Behavior Using Tax Reforms as National Experiments," NBER Reprint No. 1946, February 1995, and Brookings Papers on Economic Activity (1994:2), pp. 1-74; and "Tax Reforms and Investment: A Cross-Country Comparison," NBER Reprint No. 2102, December 1996, and Journal of Public Economics 62 (1996), pp. 237-73.
3 J.G. Cummins, K.A. Hassett, and R.G. Hubbard, '71 Reconsideration of Investment Behavior..."
4 R.G. Hubbard, A.K. Kashyap, and T.M. Whited, "Internal Finance and Firm Investment," NBER Reprint No. 2004, September 1995, and Journal of Money, Credit, and Banking 27 (August 1995), pp. 683-701; J.G. Cummins, T.S. Harris, and K.A. Hassett, "Accounting Standards, Information Flows, and Firm Investment Behavior," and J. G. Cummins and R.G. Hubbard, "The Tax Sensitivity of Foreign Direct Investment: Evidence from Firm-Level Panel Data," in The Effects of Taxation on Multinational Corporations, M. Feldstein, J.R. Hines, and R.G. Hubbard, eds. Chicago: University of Chicago Press, 1995.
5 J.G. Cummins, K.A. Hassett, and R.G. Hubbard, "A Reconsideration of Investment Behavior..." and "Tax Reforms and Investment..." and "Have Tax Reforms Affected Investment?", in Tax Policy and the Economy, Vol. 9, J.M. Poterba, ed. Cambridge: MIT Press, 1995. For an early emphasis on the importance of focusing on tax, rather than nontax, variation in the user cost of capital, see M. Feldstein and J. Flemming, "Tax Policy, Corporate Saving, and Investment Behavior in Britain," Review of Economic Studies, 38, October 1971.
6 S.M. Fazzari, R.G. Hubbard, and B.C. Petersen, "Financing Constraints and Corporate Investment," Brookings Paper on Economic Activity (1988:1), pp. 141-95; C.W. Calomiris and R.G. Hubbard, "Internal Finance and Investment: Evidence from the Undistributed Profits Tax of 193 7-1938," Journal of Business 68 (October 1995), pp. 443-82; M. Gertler and R.G. Hubbard, "Financial Factors in Business Fluctuations," in Financial Market Volatility: Causes and Consequences, Federal Reserve Bank of Kansas City, 1988; R.G. Hubbard, "Introduction," in Asymmetric Information, Corporate Finance, and Investment, R.G. Hubbard, ed., Chicago: University of Chicago Press, 1990; R.G. Hubbard and A.K. Kashyap, "Internal Net Worth and the Investment Process: An Application to the U.S. Agriculture,"Journal of Political Economy 100 (June 1992), pp. 506-34; and R.G. Hubbard, A.K. Kashyap, and T.M. Whited, op. cit.
7 See the review of studies in R.G. Hubbard, "Capital-Market Imperfections and Investment," NBER Working Paper No. 5996, April 199 7.
8 B.S. Bernanke, M. Gertler, and S. Gilchrist, "The Financial Accelerator and the Flight to Quality," Review of Economics and Statistics 78 (February 1996), pp. 1-15.
9 S.M. Fazzari, R.G. Hubbard, and B.C. Petersen, "Investment, Financing Decisions, and Tax Policy," American Economic Review 78 (May 1988), pp. 200-5.
10 See the review and examination of studies in R. G. Hubbard, "Capital-Market Imperfections and Investments,"Journal of Economic Literature, forthcoming, 1997.
11 I explore this linkage between the literatures in R.G. Hubbard, "Investment Under Uncertainty: Keeping One's Options Open,"Journal of Economic Literature 32 (December 1994), pp. 1816-31.
12 K.A. Hassett and R.G. Hubbard, "Tax Policy and Investment," in Fiscal Policy: Lessons from Economic Research, A.J. Auerbach, ed. Cambridge: MIT Press, 1997.
13 J.G. Cummins, K.A. Hassett, and R.G. Hubbard, "A Reconsideration of Investment Behavior..."
14 K.A. Hassett and R.G. Hubbard, "The World Market for Capital Goods: Does Local Policy Affect Prices?", mimeo, Columbia University, 1996.
15 K.A. Hassett and R.G. Hubbard, "Tax Policy and Investment..."
16 D. Cohen, K.A. Hassett, and R.G. Hubbard, "Inflation and the User Cost of Capital: Does Inflation Still Matter?", mimeo, Columbia University, 1997.
17 J.G. Cummins, K.A. Hassett, and R.G. Hubbard, "A Reconsideration of Investment Behavior..."
Glenn Hubbard is the Russell L. Carson Professor of Economics and Finance at Columbia University and a research associate in the NBER's Programs in Public Economics, Corporate Finance, Monetary Economics, and Economic Fluctuations and Growth. A brief biographical story about him appears in the Profiles section of this issue.
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||evaluation of 'neoclassical' investment models|
|Author:||Hubbard, R. Glenn|
|Date:||Jun 22, 1997|
|Previous Article:||Environmental tax policy using a two-part instrument.|
|Next Article:||Evaluating age discrimination laws.|