INTERNATIONAL ECONOMIC JOURNAL 93 Volume 12, Number 2, Summer 1998 PRIVATE AND GOVERNMENT INVESTMENT: A STUDY OF THREE OECD COUNTRIES MEHDI S. MONADJEMI AND HYEONSEUNG HUH* University of New South Wales This paper examines the relationship between private investment and government spending in Australia, Britain and the United States. Since all time series data are stationary in first difference and cointegrated, these series are represented by an error correction model. Variance decomposition and impulse response functions are employed to investigate the effects of government spending on private investment. Generally the empirical results provide limited support for crowding out effects of government investment on private investment. The rate of interest and the corporate profit ability showed significant effects on private investment in two out of three cases. [E62, E63] 1. INTRODUCTION The effect of public spending on private sector expenditure has received considerable attention in the economic literature. Bailey (1971) was the first study that argued in support of some degree of substitutability between the two types of spending. The focus of research in this area has ranged from those studies that were mainly concerned with the effectiveness of fiscal measures to those that examined substitutability and complementarity relationship between private and government spending. The former studies include Barro (1974), Kormendi (1983) and Feldstein (1982), while in the latter category Aschauer (1989), Bailey (1971), Barro (1981), Monadjemi (1993), Erenburg (1993), Karras (1994) and Erenburg and Wohar (1995) may be mentioned. The empirical results of these studies in both categories are highly controversial. The first group of studies generally examined the empirical implications of the Ricardian equivalence hypothesis (REH), with Kormendi (1983) providing strongest support for the REH. The theoretical arguments and the empirical findings of most of the studies in this group were summarised in Seater (1993), in which he argued that good empirical studies are supportive of the REH and those studies that refute it generally lack econometric precision. In the second group of studies Aschauer (1985), Barro (1981), Bailey (1971) and Monadjemi (1993) provided evidence in support of the substitutability hypothesis, whereas the empirical *This paper was presented in seminars at the University of Kent and the University of Southamption. The authors would like to thank Alan Carruh, Andy Dickerson, Lance Fisher, Owen O Donnell, Peter Sanfey, Tony Thirlwall and participants in both seminars for their valuable comments.
94 M. MONADJEMI AND H. HUH findings of Aschauer (1989), Erenburg (1993), Karras (1994) and Erenburg and Wohar (1995) were supportive of the complementarity nature of public and private spending. Aschauer (1989) argued that to examine the relationship between private and government spending, it is essential to distinguish between different types of expenditure. Some categories of government expenditure such as expenditure on roads, education, airports and research may increase private sector s productivity and hence may complement private investment expenditure. However, there are certain types of government consumption expenditure on foods and health which may substitute for private consumption expenditure on these items. Moreover, as Karras (1994) has shown, the relationship between government and private spending is influenced by the size of the government. Karras (1994) argued that, as the size of the government sector increases, it is likely that the relationship between private and public spending turns into substitutability rather than complementarity. Perhaps this change in the relationship is due to the provision of more public services rather than infrastructure as the economy matures and the size of the government sector expands. The purpose of this paper is to empirically examine the relationship between private investment and government spending in Australia, Britain and the United States. The distinguishing features of this paper from the other research in this area are: (1) the use of quarterly time series instead of annual data (2) the presentation of empirical results based on an error correction model (ECM) and (3) the use of data from three countries. The theoretical discussion of the paper is developed in Section 2. In Section 3 econometric methodology is discussed. The empirical results of the study are presented in Section 4. A comparison with the Aschauers study and summary and concluding remarks are offered in Sections 5 and 6 respectively. 2. THEORETICAL DISCUSSION Macroeconomists are interested in the relationship between private investment and public expenditure mainly because of the crowding-out effect of public spending. The crowding-out effect reduces the ability of the government to influence economic activity through fiscal measures. Using annual data for the United States over the period of 1953-1986, Aschauer (1989) attempted to examine the relationship between public capital and private investment in terms of the effect of public spending on private sector s marginal productivity of capital. Public expenditure may have a complementary relationship with private investment if that type of expenditure improves the productivity of private investment. In this case, an increase in public spending leads to an increase in private investment. However, public spending may crowd-out private investment if the relationship between them is based on substitutability. Hence, the relationship between public spending and private sector s productivity may provide additional information about the relationship between public spending and private investment. This issue was examined in Aschauer (1989), where
PRIVATE AND GOVERNMENT INVESTMENT 95 it was shown that government investment crowded-out private investment while increasing marginal productivity of private capital by a factor of one to one. Aschauer (1989: 186) argued that the empirical results show that while both channels appear to be operating, the latter comes to dominate, so the net effect of a rise in public investment expenditure is likely to raise private investment. This means that government investment had a positive effect on private investment and caused a crowding-in rather than a crowding-out. The theoretical relationship between private and public spending goes back as early as Bailey (1971) and Buiter (1977). These two studies were mainly concerned with the crowding-out effect of public expenditure and the degree of substitution and complementary relationship between private and public spending. In a recent study in this area Erenburg and Wohar (1995), in the context of accelerator, neoclassical and security valuation models, examined the causal linkages between private and public investment in equipment. The empirical analysis of Aschauer (1989) is based on the neoclassical model in which private non-residential investment was assumed to be a function of government investment, government consumption and the rate of return on private non-financial corporate capital. The theoretical framework of this study is based on the marginal efficiency theory of investment, the accelerator theory and the causal relationship between private and public investment. The marginal efficiency theory of investment maintains that there is an inverse relationship between private investment and the rate of interest. This is the oldest theory of investment in which the interest cost of financing has an adverse effect on private investment. The accelerator theory of investment assumes that change in sales is an important determinant of private investment. However, Erenburg and Wohar (1995), in the context of the neoclassical theory, assumed that the demand for capital is a function of cost of capital and other factors. The influence of other factors is taken into account by including a proxy for sales or output. This form of the neoclassical theory of investment is similar to the accelerator theory. In this study the rate of interest and the profit of non-financial corporations are included in the private investment function in order to capture the combined effects of cost of capital and sales on demand for investment. Erenburg and Wohar (1995) also considered the effect of profit rate by including the ratio of after tax profits (adjusted for inventory valuation and capital consumption allowances) to corporate income in their investment model. Taken together, the rate of interest and the level of profit may indicate the effects of changes in economic and financial conditions on private sector s demand for investment. The main purpose of this study is to analyse the effect of public spending on private investment. Generally investment studies do not include any measure of public spending unless the study is explicitly concerned with the crowding-out effect of public expenditure. This type of investment functions were theoretically discussed in Buiter (1977) and Mundell (1992) and were empirically examined in Aschauer (1989), Erenburg and Wohar (1995), Erenburg (1993) and Tatom (1991). Basically these studies argued that the influence of public investment should be considered since public
96 M. MONADJEMI AND H. HUH capital may compete or may complement private capital. Based on the foregone analysis the following vector autoregressive (VAR) model is proposed for investigating the relationship between real private and government investment: Y t = A(L)Y t + V t (1) where Y t is a vector of four time series, (P I t, CP t, R t, GI t )', A(L) is an (p x p) polynomial matrix in the lag operator L and V t is a vector of random disturbances with V t ~ N(0,Σ). P I t C P t, R t, and G I t are real private investment, real corporate profit, interest rate and real government investment respectively. 3. ECONOMETRIC METHODOLOGY Many studies in this area, such as Aschauer (1989) have used conventional regression techniques for estimating the effect of fiscal variables on private consumption or investment expenditure. Unfortunately, most macroeconomic time series are non-stationary and conventional regression techniques based on nonstationary data tend to produce spurious results. However, non-stationary time series may be c o i n t e g r a t e d if some linear combinations of the series become stationary. That is, the series may wander around, but in the long run there are economic forces that tend to push them to an equilibrium. Hence, cointegrated series will not move far away from each other and are linked in the long-run. Johansen (1988 and 1991) suggested tests for determining numbers of cointegrating vectors among a group of variables. Consider the following p variable VAR model: k Y t = u + φ i Y t-i + ε t (2) i=1 where u is the constant term, Y t is (p 1) vector of the variables under study, and the disturbance vector ε t of dimension (p 1) is distributed as an i.i.d. Gaussian process with zero mean and variance Ω. Assuming that series are cointegrated, equation 2 may be re-parameterised to give the following ECM representation: Y t = u + k-1 Γ i Y t-i + Y t-k + ε t (3) i=1
PRIVATE AND GOVERNMENT INVESTMENT 97 Γ j = -I - j i=1 φ i (4) = - I - k i=1 φ i (5) where I is the identity matrix. [See Johansen (1991) for details.] The long-run relationship between series is determined by the rank of. If time series are non stationary and cointegrate, then is not full rank, that is 0 < rank ( ) = r < p, where r is the number of cointegrating vectors. Johansen (1991) proposed two likelihood testing procedures in order to estimate the rank of. The first tests the hypothesis that the number of cointegrating vector is at most equal to r (Trace test). The second tests the hypothesis that the number of cointegrating vectors is equal to r (Max Eigenvalue test). When the series are found to be cointegrated, Johansen further demonstrated that can be factored as: = αβ' (6) where β is matrix of r cointegrating vectors and α is the matrix of weights attached to each cointegrating vectors in equation 3. Both α and β are p r matrices. 4. EMPIRICAL RESULTS The results of Dickey-Fuller and Phillips-Perron s test for unit-roots indicated that the null hypothesis of unit root cannot be rejected for levels of variables, but is rejected for first differences of all variables. 1 The results of Johansen s test for cointegration are reported in Table 1. These results show that the hypothesis of the existence of one cointegrating vector among Australian, British and US series cannot be rejected at the 5 percent significance level. Normalized coefficients of cointegrating vectors on private investment are reported in the lower part of Table 1 for all three countries. The dynamic interactions among the variables and the relative importance of various shocks can be evaluated by impulse response functions (IRF) and variance decompositions (VDC). To do so, the estimated ECMs are transformed into models 1 These results are available upon request from the authors.
98 M. MONADJEMI AND H. HUH in levels of the series and inverted to (cointegrated) vector moving average forms. [See Engle and Granger (1987) for details]. However, since these models are basically a reduced form relationship, they are difficult to interpret and may provide Table 1. Johansen s Test for Cointegration Hypothesis Trace 95% CV Alternative λ Max 95% CV Australia r 3 1.20 3.76 r = 3 1.20 3.76 r 2 8.38 15.41 r = 2 7.19 14.90 r 1 22.16 29.68 r = 1 13.78 21.07 r = 0 60.22 * 47.21 r = 0 38.06 * 27.14 Coefficients of cointegrating vectors PI GI CP R t C 1.00 1.20-0.87-0.04 10.56 (0.18) (0.05) (0.004) Britain r 3 2.57 3.76 r = 3 2.57 3.76 r 2 6.15 15.41 r = 2 3.58 14.90 r 1 20.93 29.68 r = 1 14.78 21.07 r = 0 55.33 * 47.21 r = 0 34.40 * 27.14 Coefficients of cointegrating vectors PI GI CP R t C 1.00 1.82 1.62 0.39 45.73 (1.46) (1.96) (0.32) United States r 3 0.67 3.76 r = 3 0.67 3.76 r 2 6.18 15.41 r = 2 5.51 14.90 r 1 21.85 29.68 r = 1 15.66 21.07 r = 0 48.02 * 47.21 r = 0 28.17 * 27.14 Coefficients of cointegrating vectors PI GI CP R t C 1.00 5.63 7.03 0.34 53.39 (6.28) (9.18) (0.40) Note: Sample periods are 1970.1-1991.4. * indicates significant cases at 5 percent level. Standard errors of cointegrating coefficients are reported in brackets. The results were produced using micro TSP Eviews package. Critical values for the Johansen cointegration test are obtained from Osterwald-Lenum (1992). Four lags are allowed in generating the above results.
PRIVATE AND GOVERNMENT INVESTMENT 99 little information. To draw a meaningful structural interpretation, appropriate identifying assumptions are to be imposed on the basis of economic theory. In their paper, King et al. (1991) used a set of long-run zero identifying restrictions for an ECM in which the cointegration restrictions are taken into account. On the other hand, Orden and Fisher (1993) employed a standard Choleski-type of contemporaneous identifying restrictions in their ECM. The model that is employed in this study does not lend itself to the application of long-run zero identifying restrictions like those used in King et al. (1991). This is because there is no strong theoretical long run relationship between private and government investment that is widely supported in the literature. Without a strong economic rationale, imposing long-run zero restrictions may be too restrictive. Accordingly, in this paper, similar to Orden and Fisher (1993), contemporaneous identifying restrictions are employed, which are less restrictive. In VAR models with Choleski-type of identifying restrictions, the ordering of variables may alter the results. That is, choosing a certain order of variables imposes a particular recursive structure on the model. The recursive structure assumes that variables appearing first contemporaneously influence the latter variables but not vice versa. Therefore, it is important to list the most exogenouslooking variables earlier than the most endogenous-looking variables. This type of ordering allows the exogenous variables to contemporaneously influence endogenous variables but not to be influenced by them. To this end, the ordering of GI CP R and PI was chosen for the implementation of IRF and VDC analyses. In this order, all variables have contemporaneous influence on PI. Table 2 reports the results of VDC for 1, 5, 10, 15 and 20 quarters ahead in order to assess the relative importance of various shocks in accounting for the forecast error variance of private investment. 2 As indicated by VDC results, with the exception of Britain, changes in nonfinancial corporate profit appear to be relatively important variables in explaining forecast error variance of private investment. Government investment is important in Australia but not in the other two countries. Changes in the rate of interest have an important influence on private investment in Britain and the United States but not in Australia. VDC coefficients indicate the importance of certain variables but fail to give information about the direction of response of variables to certain shocks. That is, one cannot deduce any information about complementarity or substitutability of private and public expenditure from VDC coefficients. However, IRF functions may be used to examine the positive or negative responses of private investment to changes in different variables in the model. The IRF derived from the ECM are presented in Figure 1. These functions show the dynamic responses of real private investment to a one standard deviation shock of each series except for its own. One standard deviation confidence interval bands are also constructed to test the significance of the response to a particular shock. The 2 The VDC and IRF results were produced using Rats software package.
100 M. MONADJEMI AND H. HUH response is considered significant if confidence intervals do not pass through the zero line. For Australia and the United States the responses of private investment to one standard deviation shock of corporate profit are positive and significant. The response to government investment shock is negative and significant in the Australian case. Interest rate changes influence private investment in Britain and the United States but not in Australia. With the exception of Australia, these results i ndicate that changes in government investment have no effect on private investment. The effects of corporate profitability (in the Australian and the U.S. cases) and the rate of interest (in Britain and the United States) are consistent with the theoretical discussion presented earlier. 3 Table 2. Variance Decomposition of Real Private Investment Australia Quarters Ahead GI CP R PI 1 11.99 4.46 0.28 83.27 5 31.15 33.42 1.59 33.84 10 36.90 40.46 1.93 20.72 15 37.89 42.25 1.80 18.06 20 38.37 43.00 1.77 16.85 Britain Quarters Ahead GI CP R PI 1 6.87 1.98 1.62 89.54 5 12.44 5.13 34.97 47.83 10 13.54 7.81 56.26 22.15 15 12.06 9.41 59.00 18.04 20 13.98 9.90 59.54 16.58 United States Quarters Ahead GI CP R PI 1 6.46 18.98 4.15 70.40 5 10.99 48.61 15.98 24.41 10 10.83 49.62 19.55 19.99 15 10.12 48.97 21.81 19.11 20 9.50 48.19 23.48 18.83 Note: Sample periods are 1970.1-1991.4. Variables are in log levels as specified in the ECM, with one cointegrating vector in each case. Lag length of variables is four. 3 The VDC and IRF for different ordering of variables such as (PI R CP GI) or (R CP GO PI) produced similar results.
PRIVATE AND GOVERNMENT INVESTMENT 101 N o t e : Sample period is 1970:1-1991:4. Variables are log levels except for the interest rate. Four quarter lags are allowed to produce these results. Bands shown with dotted lines indicate one standard deviation above or below the mean. The means of IRF and their c o r r e s p o n d i n g standard deviations were computed using Monte Carlo simulation with 100 replications.
102 M. MONADJEMI AND H. HUH 5. A COMPARISON WITH ASCHAUER S STUDY It is interesting to compare and contrast the empirical results of this study with those of Aschauer s (1989). Aschauer used U.S. annual time series data on various types of investment and consumption expenditures and deflated them by net private non-residential fixed capital stock. The sample period in Aschauer s study included the period from 1953 to 1986. In contrast, all series in this study are quarterly time series during the period 1960s to 1991 and all variables are logarithms of real values. Aschauer s analysis is, in principle, based on a singleequation estimation whereas the empirical results of this study are based on IRF and VDC derived from the error correction model. Aschauer did not examine his time series data for stationarity and cointegration while in this study both unit root test and the Johansen procedure were employed to determine the appropriate form of data. Aschauer found a strong crowding-out effect in relation to government investment expenditure but not for government consumption and military expenditures. Because of the positive effect of government non-military capital on the rate of return on private capital, Aschauer concluded that the relationship between private and public capital is complementary. The results of this study show some crowding-out effects over a limited range only in the Australian case. The positive effect of corporate profitability on private investment in two cases, agrees with the positive and significant effect of the rate of return on private capital in Aschauer s study. The negative and significant effect of interest rate on private investment is not included in Aschauer s study but is considered here and the results are consistent with the marginal efficiency theory of investment. 6. CONCLUDING REMARKS An attempt was made in this paper to examine the effect of government spending on real private investment in Australia, Britain and the United States. Since all of the time series are stationary in first difference and cointegrated, the ECM is the appropriate model for examining the responses of private investment to fiscal shocks. The empirical results of the study show that corporate profitability and in some cases interest rates, are the most important variables affecting private investment. Private investment showed significant responses to a shock of government investment only in Australia but not in the other two countries. The responses of private investment in Britain to a shock of government investment was negative over the entire 20 quarters. APPENDIX Australian Data (Seasonally adjusted real values in millions of Australian dollars.)
PRIVATE AND GOVERNMENT INVESTMENT 103 PI = Real value of private investment expenditure on equipment and non dwelling construction, millions of 1989-90. Source: Economic Series, DX database V2.1, School of Economics, University of New South Wales (DXDB). GI = Real value of government investment expenditure, millions of 1989-90 dollars. Source: DXDB. CP = Real value of corporate trading profit, nominal values divided by price deflator of GDP. Source: DXDB. R t = Nominal 90 days bank accepted bills rates. Source: Reserve Bank of Australia Bulletin, Table J1, various issues. British Data (Seasonally adjusted real values in millions of British pounds.) PI = Real private, gross fixed capital formation in 1990 prices. Source: E c o n o m i c Trends Annual 1995 Edition (ET), Table 1.8, column 1. GI = Real general government and public corporations gross fixed capital formation in 1990 prices. Source: ET 1995, Table 1.8, columns 2 and 3. CP = Real value of gross trading profits of companies, nominal values divided by the implicit price deflator of GDP. Source: ET 1995, Table, 1.4, column 2. R t = Nominal call money rates. Source: DXDB. United States Data (Seasonally adjusted real values in billions of dollars.) PI = Real private non-residential gross private domestic investment expenditure, billions of 1982 dollars. Source: Survey of Current Business (S C B) January 1976, Table 1.2 lines 4 and 5 and SCB September 1990 and S C B July 1991, Table 1.2 lines 8. GI = Real government non-defence expenditure, government purchase of goods and services less federal government defence expenditure, billions of 1982 dollars. Source: SCB January 1976, Table 3.2, and SCB September 1990 and SCB July 1992, Table 3.8B. CP = Real value of non-financial corporate profit, nominal values divided by the implicit price deflator of GDP. Source: SCB various July issues. R t = Nominal federal fund rate. Source: DXDB. REFERENCES Aschauer, D. A., Fiscal Policy and Aggregate Demand, American Economic Review, March, 1985, 117-128. Aschauer, D. A., Does Public Capital Crowd Out Private Capital?, Journal of Monetary Economics, September 1989, 171-188. Bailey, M., National Income and the Price Level, McGraw-Hill, 1971.
104 M. MONADJEMI AND H. HUH Barro, R., Are Government Debt Net Wealth?, Journal of Political Economy, November/December 1974, 1095-1117. Barro, R., Output Effects of Government Purchases, Journal of Political Economy, December 1981, 1086-1111. Buiter, W. H., Crowding Out and the Effectiveness of Fiscal Policy, Journal Public Economics, June 1977, 309-328. Engle, R. F., and Granger, C. W. J., Cointegration and Error Correction: Representation, Estimation and Testing, Econometrica, March 1987, 251-276. Erenburg, S., The Real Effects of Public Investment on Private Investment, Applied Economics, February 1993, 831-837. Erenburg, S., and Wohar, M., Public and Private Investment: Are There Causal Linkages?, Journal of Macroeconomics, Winter 1995, 1-30. Feldstein, M., Government Deficits and Aggregate Demand, Journal of Monetary Economics, January 1982, 1-20. Fuller, W. A., Introduction to Statistical Time Series, New York: Wiley and Sons, 1976. Johansen, S., Statistical Analysis of Cointegration Vectors, Journal of Economic Dynamics and Control, June-September 1988, 231-254. Johansen, S., Estimation and Hypothesis Testing of Cointegration Vectors in Gaussian Vector Autoregressive Models, Econometrica, November 1991, 1551-1580. Karras, G. Government Spending and Private Consumption: Some International Evidence, Journal of Money, Credit and Banking, February, 1994, 9-22. King, R., Plosser C. Stock, J. and Watson, M., Stochastic Trends and Economic Fluctuations, American Economic Review, September 1991, 819-840. Kormendi, R. C., Government Debt, Government Spending and Private Sector Behaviour, American Economic Review, December 1983, 994-1010. Monadjemi, M. Fiscal Policy and Private Investment Applied Economics, February 1993, 143-145. Mundell, A., Policy Watch: Infrastructure Investment and Economic Growth, Journal of Economic Perspectives, Fall 1992, 189-198. Orden, D. and Fisher, L. Financial Deregulation and the Dynamics of Money, Prices and Output in New Zealand and Australia, Journal of Money, Credit, and Banking, May 1993, 273-292. Osterwald-Lenum, M., Practitioner s Corner: A Note with Quantiles of the Asymptotic Distribution of the Maximum Likelihood Cointegration Rank Test Statistics, Oxford Bulletin of Economics and Statistics, 1992, 461-472. Schwarz, G., Estimating the Dimension of a Model, The Annals of Statistics, March 1978, 461-467. Seater, J. J., Ricardian Equivalence, Journal of Economic Literature, March 1993, 142-190. Tatom, J., Public Capital and Private Sector Performance, Federal Reserve Bank of St Louis Review, May/June 1991, 3-15.
PRIVATE AND GOVERNMENT INVESTMENT 105 Mailing Address: Dr. Mehdi S. Monadjemi, School of Economics, University of New South Wales, Sydney 2052 Australia. Tel: 612-93853494, Fax: 612-93136337, e - mail: m.monadjemi@unsw.edu.au Mailing Address: Dr. Hyeonseung Huh, School of Economics, University of New South Wales, Sydney 2052 Australia. Tel: 612-93855797, Fax: 612-93136337