Public spending, prívate investment, and growth in O.E.C.D. countries: an empirical investigation

P.S.P.l/10-2-96 I I I ENCUENTRO DE ECONOMIA PUBLICA Departamento de Teoría Económica y Economía Política Universidad de Sevilla Sevilla 9, 10 de febrero de 1995 PONENCIA Public spending, prívate investment, and growth in O.E.C.D. countries: an empirical investigation. Isabel ARGIMON MAZA Banco de España José Manuel GONZÁLEZ-PÁRAMO Banco de España. Departamento de Hacienda Pública Universidad Complutense José María ROLDAN European Monetary Institute Summary: This paper is concerned with the empirical relationship between government spending and prívate investment. A panel of 14 OECD countries is used. We present evidence which suggests the existence of a significant crowding-in effect of prívate investment by public investment, through the positive impact of infrastructure on prívate investment productivity. Moreover, government consumption appears to crowd out prívate investment. The implications of these resulte are of foremost importance when it comes to fiscal consolidation. Déficit reductions engineered through cute in public investment could severely impinge upon prívate capital accumulation and growth prospecte. Resumen: Este trabajo analiza la relación empírica existente entre el gasto público y la inversión privada en un panel de 14 países pertenecientes a la OCDE. La evidencia presentada sugiere la existencia de un significativo efecto "crowding-in" de la inversión pública, que opera a través de la complementariedad entre las infraestructuras y la inversión privada. Por otra parte, el consumo público parece ejercer un "efecto expulsión" de la inversión privada. La implicaciones de estos resultados son de gran trascendencia en materia de consolidación fiscal. Reducciones del déficit conseguidas mediante recortes en la inversión pública podrían dañar severamente la acumulación de capital y las perspectivas de crecimiento. 1. INTRODUCTION: FISCAL POLICY, GROWTH, AND CROWDING OUT In this paper we investígate the relationship between prívate investment and government spending. Empirical evidence is brought to bear upon two closely related questions which lie at the core of the debate on the macroeconomic effects of fiscal policy: 1) Is public spending productive? 2) To what extent, if any, does government spending substitute for prívate investment? The first question has stimulated a considerable amount of research sin ce Rubinson (1977) and Ram (1986) found a positive empirical relationship between government size and GDP growth. Although this association has al so been obtained in more recent work (see Lin, 1994a), a number of papers have identified an inverse association between government spending and output growth (e.g. Grossman, 1988, Mallow, 1986, Peden and Bradley, 1989, and Grier and Tullock, 1989). In a careful and comprehensive analysis of this issue, Dowrick (1993) concludes that government size does not appear to have a systematic and significant effect on growth. This literature has been criticized on three main grounds. First, as shown by Hsieh and Lai (1994) and Dowrick (1993), the sign of the association is quite sensitive to the choice of the sample of countries, the period under study, the econometric techniques and the existence of measurement problems. Second, as different categories of spending are well documented to have diverse economic effects, analyses ignoring this fact are not easy to interpret. Easterly and Rebelo (1993) and Lin (1994b), for example, show that spending categories that promote human or physical capital accumulation are positively associated with growth, while other spending items have negative or neutral effects. Finally, the single equation findings on the government - size/GDP- growth link could be invalid when this relationship runs through indirect channels, i.e. prívate investment. These empirical shortcomings have a direct bearing on the second question. Since capital accumulation is the engine of output growth, any crowding-out effects of prívate investment by public spending impinge upon production expansión and welfare prospecte. The bulk of the empirical literature finds a significantly negative effect of public consumption on growth while the effects of public investment are found to be positive although less robust (Barro, 1991; Grier and Tullock, 1989; Easterly and Rebelo, 1993). To what extent can these resulte be traced to a crowding out of prívate investment? Aschauer (1989b) finds that the direct crowding-out effect of public investment is outweighed by a direct crowding-in effect associated with the role of public capital as a productive input and its complementante0 with prívate capital (see Aschauer, 1989a). The evidence gathered in Erenburg (1993), who estimates a simple macromodel with rational expectations, and in Erenburg and Wohar (1995), reinforces this conclusión, while the resulte in Bairam and Ward (1993) support the crowding-out hypothesis. As to the impact of government consumption, Aschauer (1989b) concludes that its crowding-out effect on prívate investment carries only a marginal explanatory power, a result that could be interpreted as an indication that public consumption is a cióse substitute for prívate consumption. However, recent evidence gathered by Karras (1994) forcefully suggeste that prívate and government consumption are best described as complementary (or unrelated) goods(2), a feature that reinforces the crowding-out effect of prívate investment. The rest of the paper is organized as follows. Section II outlines the theoretical argumente behind the crowding-out hypothesis. For expositional purposes, we use a simple overlapping-generations model in which public and prívate capital are complemente, whereas public and prívate consumption are independent Section III evaluates the impact on prívate investment and prívate productivity of public spending. The empirical resulte obtained with a panel of 14 OECD countries are presented and briefly discussed. Finally, Section IV draws the main conclusions. 2. PRIVATE INVESTMENT AND PUBLIC EXPENDITURE The equilibrium approach developed in Aschauer (1988) and Aschauer and Greenwood (1985), among others, assumes a competitive economy populated by rational, identical, infinitely lived individuáis. In this context, the general equilibrium relationship between public spending and prívate investment may be expressed in the following two equations: (1 Public and prívate capital are "complementary" when the marginal productivity of the prívate capital increases as the quantity of public capital increases. (2) If public consumption increases the marginal utility of prívate consumption, both are said to be "complementary", and viceversa. 3 i = i(fk,ig,cg), i,>0, z2<0, z3^0 (1) fk =A(Up, fu<Q.f¿& (2) where i is prívate spending, fk is the marginal product of capital, ig and cg are public investment and public consumption, respectively, k is prívate capital and kg is public capital. Along neoclassical lines, changes in investment -which are described by the partial derivatives of the above functionsare the result of intertemporal smoothing of consumption by prívate agents. Aschauefs discussion of the crowding-out issue is heiuistic, due to the complexity of his analytic framework. Nonetheless, his qualitative conclusions carry over in much simpler neoclassical models. Suppose that our economy is populated by overlapping generations of equal size. Each household lives for two periods. Households earn labour income only in the first period. Labour supply is fixed. Consumption of a representative household when young in time t is: c,1 = w, - t, - s, , (3) whereas consumption when oíd at time t+1 can be written as: <Á=í,(K+1), (4) where w, is the wage rate, st is savings, rt+I is the interest rate in the second period and t, is a tax levied on the young. The government uses tax revenues to finance public consumption of a public good nature -which enters household's utility function- and public investment -which is a productive input in prívate production. Suppose that the utility function of a representative agent is: U = hLnc¡ + (l-tyMCf +_L[5I/ic,2+1 + (1-8)1^, (5) where cg is public consumption, 6 represente relative preference for prívate consumption and p is the rate of time preference. Maximization of (5) subject to (3) and (4) yields the savings function: st = o(w, - t), a=_L . (6) For simplicity, assume that both prívate and public capital fully deprecíate in every period. The production function in per worker terms is: yt = Ai,ai¡ , a+p<l , (7) where i is prívate capital (and prívate investment) and ig is public capital (and public investment). Competition in factor markete implies: l4rf =CL4/;%p -<¿ (8) wt -(l-a)Ai?i¡ -(l-cty . (9) Capital available for production in t+1 equals savings of the young in t. This gives the momentary equilibrium condition: ^ = 0(1 -a)M?i¡ ~ ot, . (10) This equation, together with the government budget constraint + i, • CID define the equilibrium of the economy. A permanent tax-financed increase in public consumption reduces both firstperiod prívate consumption and savings, and thus investment in the next period. In the long-run, after dropping time subscripte, we obtain: 5 di_ de l-oa(l-cO¿/a-1// < O , (12) g provided that the economy is not too far away ffom the golden rule (note that when r=0, by equation 8 the denominator in 12 reduces to l-o(l-a)). As this crowding-out effect generated by public consumption is associated with the effect of taxes upon savings, a tax-fínanced increase in public investment must also produce a direct crowding-out effect of prívate investment However, public investment may have an additional effect prívate investment is crowded-in whenever public capital raises the productivity of prívate capital. In fact, if the initial level of public capital were well below its optimal (i.e. output-maximizing) level, the crowding-in effect would outweigh the negative influence of taxes upon savings. More generally, the steady-state effect of an increase in public investment upon prívate capital accumulation is: In the more elabórate Aschauer model this ambiguity emerges as well. The differences are just a matter of detail. When public and prívate capital are equally productive, an increase in public investment crowds out an equivalent amount of prívate investment, with no effects upon wealth and consumption. However, an additional crowding-out effect would arise if public investment were more productive: as lifetime wealth increases so does consumption today. These two short-run effects could be compensated by the crowding-in effect which operates through an increase in prívate capital productivity. As to the consequences of an increase in public consumption, Aschauer admits the possibility that its effect could be nil. If public consumption is a perfect substitute for prívate consumption or the marginal propensity to consume out of wealth remains constant over time for every agent, prívate investment changes would not be needed to smooth out consumption. None of these assumptions are made in our simple expositional model(3). (' These models are not directly comparable. Note that we are assuming that public and prívate consumption are neither complementary ñor substitutes. On the other hand, with income accruing only in the first period and a fixed propensity to save, crowding out obtains even if the marginal propensity to consume remains constant. In such case, which implies r = p, prívate investment is crowded-out by [l-a(l-p) ]"' units per di a[l-P(l-aMí°í/"'] < (13) l-oaO-cOA--1// > 6 3. DATA, METHODOLOGY AND EMPIRICAL RESULTS The empirical analysis is carried out using the annual time-series-cross-section data from Summers and Heston (1991). Only the fourteen OECD countries with available data on infrastructure capital were included for the period 1979-88(4), which resulted in an unbalanced panel data set Data definitions are given in the Appendix. Our analysis of the crowding-out hypothesis is based on the estimation of the following pair of equations: +^to**+Vln** +4* (15) where subscripts t refer to time and j to country, i and ig are, respectively, prívate and public investment in relation to prívate productive capital stock (prívate and public investment rates), cg is public consumption, in relation to prívate productive capital stock, fk is marginal prívate capital productivity (which is proxied by the ratio of gross operating surplus to prívate productive capital stock), k is the stock of prívate productive capital and kg is infrastructure capital, so that no residential investment is involved(5). We would expect 2^ to be positive, and both a3 and a4 to be non-positive under the crowding-out hypothesis. On the other hand, given the standard assumption of decreasing marginal productivity, b2 is expected to be negative, and b3 positive if infrastructures and prívate productive capital are complemente. unit of additional government consumption spending. (4) The countries and the periods are the following ones: Australia (1979-87), Austria (1979-87), Belgium (1979-87), Canadá (1979-88), Denmark (1979-88), Federal Republic of Germany (1979-88), Finland (1979-87), France (1979-88), Ireland (1979-87), Norway (1980-86), Spain (1979-86), Sweden (1979-87), United Kingdom (1979-87) and United States (1980-87). (5) Unfortunately, our data do not allow private and public infrastructure to be separated out. It may be argued that both types of infrastructure should have roughly the same effects on productivity. In any case, this fact has to be borne in mind when interpreting the results. 7 The econometric analysis of the crowding-out issue requires the use of techniques well suited to deal with country heterogeneity. Let us first refer to the most rectrictive versión of (14) and (15). Assume that a^, = am and b^, = bm, Vjf Vt: that is, the response of prívate investment and prívate capital productivity to the right-hand side variables does not change over time ñor across countries. The resulte of such a restrictive model are presented in column (1) of Table 1. In columns (2) to (6) we relax these restrictions so as to take into account the heterogeneity of the data. In column (2) we report the resulte of letting a, and b, change across countries while taking the remaining coefííciente to be the same across countries. The estimates correspond to a fíxed effects model with country dummies, and are therefore a within estímate. The dummy coefficiente are not reported. In order to choose between the most restrictive model and the within estimates, an F test is carried out, whose resulte are reported in the lines headed by F^. In column (3) we report the estimates of the model in first differences. Such an approach not only takes into account fíxed effects, which drop out when the model is specified in first differences, but may be regarded as a way of dealing with non-stationarity when this problem is presentí Column (4) reporte the resulte obtained from a random effects model. The main difference with the within estimator is that the country specific effects are now treated as random and uncorrelated with the regressors. A Hausman test is constructed to check for this restriction, with the resulte shown in the row labeled %2 in table l^. (6) This test is formulated as: SSRR - SSRU/(ku - kr) _ p SSRU/(T _ to) ~ (te -M. (T-tu) where U and R stand for the unrestricted and the restricted model, respectively, SSR is the sum of squared residuals, k is the number of estimated coefficients and T is the total number of observations. ^ The Hausman test is formulated as: where pw is the vector of the k within-group estimates, PGLS is the corresponding generalized least squares estímate of the random effects model, and Vw and VGLS are their respective estimated variance-covariance matrices. Under the nuil hypothesis of no córrelation (the random effects 8 B. INSTRUMENTAL VARIABLES ESTIMATES TABLE Al. CROWDING-OUT EFFECTS INSTRUMENTAL VARIABLE ESTIMATION (1979-88) (1) (2) (3) li 0.80 (4.92)*' 0.74 (4.68)" 0.80 (5.00)** eq(l) -0.07 (0.13) -0.62 (1.36) -1.02 (2.28)** -1.05 (2 66V* eq (z) ln(k/l)jt -0.56 (A A<C\** (4.ÜO) 1.07 (6.29)'* o eq (1) eq (2) 0.024 0.041 0.024 0.024 txj 112 112 112 a: standard error of the regression The instruments used are ftaA, c,., and in equation (1) and ln (k/1),., and ln (kg/1), in equation (2). Country dummies where included in both equations as instruments. 15 v • :;v¿ r: ." i wf BIBLIOTECA C. PRODUCTION FUNCTION EQUATION The estimated equation is given by ln (yll) = a + (1 -a-p) \n[kll\ + p ln[£//] where so that again the restriction of constant retums to scale has been imposed(10). The resulte are presented in table A2. The inclusión of a time trend, t, aiming at capturing technological change, resulte in the estimates shown in columns (2), (4), and (6). The random effects model seems to be the one that fite better the data. The positive and statistically signifícant sign of the infrastructure variable seems to be present in all models, except under column (4). However, in that case, the trend variable is not statistically signifícant either, so that the specification seems to be rejected by the data. The fact that the random effects model cannot be rejected statistically may be a reflection of the fact that the technological heterogeneity among the different countries is not directly related to the different initial states of the technology. In fact, it could be identical for all countries, and the individual effects could be related to other aspecto of the economic environment. Looking at columns (5) and (6) it may be concluded that, although the output elasticity with respect to inftastructure is smaller that the elasticity with respect to private capital, the resulte imply again that there is crowding in of public investment, through its positive impact on private productivity. This result reinforces those presented in Table 1. The tests of the hypothesis of constant retums to scale are not completely satisfactory, even when the within estimator is considered. However, given the specification of the productivity function it seems more reasonable to have constant returns to scale. 16 REFERENCES ASCHAUER, D.A. (1988). "The equilibrium approach to fiscal policy"; Journal of Money, Credit and Banking, 20, pp. 41-62. ASCHAUER, D.A. (1989a). "Is Public Expenditure Productive?"; Journal of Monetary Economics, 23, pp. 177-200. ASCHAUER, D.A. (1989b). "Does public capital crowd-out private capital?"; Journal of Monetary Economics, 24, pp. 171-188. ASCHAUER, DA. and GREENWOOD, J. (1985). "Macroeconomic effects of fiscal policy"; Carnegie-Rochester conference series on public policy, 23, pp. 91-138. BAIRAM, E. and B. WARD (1993). "The externality effect of government expenditure on investment in OECD countries"; Applied Economics, 25, pp. 711-716. BARRO, R. (1991). "A cross-country study of growth, saving and government"; en D. Bernheim y J. Shoven (eds.): National saving and economic performance, NBER, University of Chicago Press, Chicago, Illinois, pp. 271-301. DOWRICK, S. (1993). "Government consumption: Its effects on productivity growth and investment"; in N. Gemmell (ed.): The growth of the public sector. Theories and evidence, Edward Elgar, Aldershot, pp. 136-152. EASTERLY, W. and REBELO, S. (1993). "Fiscal policy and economic growth: An empirical investigation"; 32, Journal of Monetary Economics, pp. 458-493. ERENBURG, S.J. (1993). "The real effects of public investment on private investment"; Applied Economics, 25, pp. 831-837. ERENBURG, S.J. and M.E. WOHAR (1995). "Public and private investment: are there causal linkages?"; Journal of Macroeconomics, Vol. 17, 1. GRIER, K.G. and TULLOCK, G. (1989). "An empirical analysis of cross-national economic growth, 1951-1980"; Journal of Monetary Economics, 24, pp. 259-279. 17 GROSSMAN, P.J. (1988). "Government and economic growth: a non linear relationship"; Public Choice, pp. 193-200. HSIEH, E. and LAI, K.S. (1994): "Government spending and economic growth: the G-7 experience"; Applied Economics, 26, pp. 535-542. KARRAS, G. (1994). "Government spending and private consumption: Some international evidence"; Journal of Money, Credit, and Banking, 26, pp. 9-22. LIN, S. (1994a). "Government spending and economic growth"; Applied Economics, 26, pp. 83-94. LIN, S. (1994b). "Allocation of government spending and economic growth"; Economic Notes by Monte dei Paschi di Siena, Vol. 23, 1. pp. 130-141. MALLOW, M.L. (1986). "Private sector shrinkage and the growth of industrialized countries"; Public Choice, 49, pp. 143-154. PEDEN, E. and BRADLEY, M. (1989). "Government size, productivity, and economic growth"; Public Choice, 61, pp. 229-245. RAM, R. (1986). "Government size and economic growth: A new framework and some evidence ffom cross-section and time-series data"; American Economic Review, 76, pp. 191-203. RUBINSON, R. (1977). "Dependency, Government revenue, and economic growth"; Studies in Comparative International Development, 12, pp. 3-28. SUMMERS, R. and HESTON, A. (1991). "The Penn World Table (mark 5): an expanded set of international comparisons, 1950-1988"; Quarterly Journal of Economics, May. 18