On the Business Cycle Effects of Government Spending

Transcription

1 On the Business Cycle Effects of Government Spending Jang-Ting Guo University of California, Riverside June 30, 2002 Abstract We show that a one-sector real business cycle model with mild increasing returns-toscale and variable capital utilization is able to produce qualitatively realistic business cycles driven solely by aggregate demand shocks. In particular, a positive government spending shock can lead to simultaneous increases in output, consumption, investment, employment, labor productivity and the real wage. Our analysis illustrates a close relationship between this result and the recent literature that explores indeterminacy and sunspots in the real business cycle framework. Keywords: Business Cycles, Government Spending Shocks, Indeterminacy. JEL Classification: E30, E32, E62. I would like to thank Sharon Harrison, Kevin Lansing, Yi Wen, and seminar participants at UC Riverside and Academia Sinica, Taipei, Taiwan for helpful comments and suggestions. All remaining errors are mine. Department of Economics, 900 University Ave., University of California, Riverside, CA, , Phone: (909) , ext. 1588, Fax: (909) , guojt@mail.ucr.edu.

2 1 Introduction Recently, Devereux, Head and Lapham (DHL, 1996) haveexploredthemacroeconomiceffects of temporary and permanent changes in government spending in a one-sector real business cycle (RBC) model with increasing returns and monopolistic competition. In a parametric version of the DHL model with logarithmic utility, indivisible labor and large markup ratio of price over marginal cost (= 1.5), a positive government spending shock can lead to higher output, consumption, investment, employment and real wage. This result is driven by the endogenous response of total factor productivity to a change in government spending. DHL s finding is an important one since it shows that qualitatively realistic business cycles can be generated in a dynamic general equilibrium framework with only demand shocks. On the contrary, in a standard one-sector RBC model with perfect competition and constant returnsto-scale, aggregate demand shocks yield countercyclical consumption, labor productivity and real wage, which is not consistent with the U.S. data. This paper makes two points. First, to maintain comparability with previous studies that mostly adopt the standard RBC model, DHL restrict the analysis to (annual) parameterizations in which their economy exhibits saddle-path stability and equilibrium uniqueness. 1 However, we find that DHL s large-markup specification possesses an indeterminate steady state (a sink) and stationary sunspot equilibria at the quarterly frequency whereby the discount rate and the capital depreciation rate are lower than those for the annual formulation. The intuition for this result is straightforward. In order to generate positive correlations between government spending shocks and macroeconomic aggregates in the DHL economy, the markup must be sufficiently strong to imply that the equilibrium wage-hours locus is positively sloped and steeper than the labor supply curve. This turns out to be exactly the necessary condition for equilibrium indeterminacy in a laissez-faire one-sector RBC model (see Benhabib and Farmer, 1994). Moreover, it can be shown that the minimum level of increasing returns needed for multiple equilibria is positively related to the discount rate and the capital depreciation rate. Therefore, indeterminacy is more likely to occur when the DHL model is examined under a quarterly parameterization. Second, given recent empirical findings of Burnside (1996) and Basu and Fernald (1997), 1 The notable exception is Rotemberg and Woodford (1992) who obtain similar results as DHL do in an environment with oligopolistic pricing and increasing returns. While the price-to-cost markup in DHL is a constant, it is countercyclical in the Rotemberg-Woodford model due to implicit collusion. 1

3 the large markup used by DHL is too high to be empirically plausible for the U.S. economy. To resolve this inconsistency, we incorporate variable capital utilization into the DHL model. Specifically, a more intensive utilization of capital is assumed to accelerate its rate of depreciation. In a symmetric equilibrium, the social technology displays a larger labor elasticity of output and a higher level of aggregate returns-to-scale (with respect to capital and labor inputs) than that in the original DHL economy. Consequently, varying capital utilization amplifies the effects of government spending shocks since it enriches the model s endogenous propagation mechanism by providing an additional margin to change output. It follows that the required magnitude of increasing returns in a one-sector RBC model to produce procyclical responses of output, consumption, investment, employment, labor productivity and real wage to changes in government purchases can be substantially reduced. In particular, we show that, ceteris paribus, DHL s results can be obtained at a level of increasing returns equal to 1.1, which lies within an empirically realistic range. The remainder of this paper is organized as follows. Section 2 presents a simplified version of the DHL model and investigates its local dynamics. Section 3 analyzes the effects of government spending shocks in the DHL economy with variable capital utilization. Section 4 concludes. 2 The Model This section reproduces the DHL model with one simplifying modification regarding the market structure. In the DHL economy, there is an intermediate-good sector in which monopolistically competitive firms operate using capital and labor inputs. The number of these intermediate firms is endogenously determined under the condition of free entry and exit. 2 Asinglefinal good is then produced from the set of available intermediates in a perfectly competitive environment. Since the degree of aggregate increasing returns in production is equal to the price-to-cost markup ratio in this economy, we present the specification with perfect competition and productive externalities. 3 This simplification streamlines our exposition without affecting any result of the paper. 2 As a result, the productivity of an intermediate firm depends positively on the number of intermediate goodsinuse.thisisreferredtoas returnstospecialization intheliterature. 3 See Benhabib and Farmer (1994) for another alternative formulation that incorporates internal increasing returns at the intermediate-firm level in an imperfectly competitive market structure without free entry. Each of these three settings yields a reduced-form aggregate production function with the same form as (3). 2

4 2.1 Firms There is a continuum of identical competitive firms in the economy, with the total number normalized to one. Each firm produces output y t according to a constant returns-to-scale technology y t = x t k α t h 1 α t, 0 < α < 1, (1) where k t and h t are capital and labor inputs, and x t represents productive externalities that are taken as given by the individual firm. We postulate that externalities take the form x t = αχ k t h (1 α)χ t, χ 0, (2) where k t and h t denote the economy-wide average levels of capital and labor services. In a symmetric equilibrium, all firms take the same actions such that k t = k t and h t = h t, for all t. As a result, (2) can be substituted into (1) to obtain the following aggregate production function that may display increasing returns: y t = h i kt α ht 1 α (1+χ). (3) When χ =0, the model collapses to a standard RBC formulation with aggregate constant returns-to-scale (zero markup). 4 As in DHL, we restrict our analysis to the case of α(1+χ) < 1 whereby the degree of increasing returns is not strong enough to generate endogenous growth. Under the assumption that factor markets are perfectly competitive, the first-order conditions for the firms profit maximization problem are given by r t = α y t k t, (4) w t =(1 α) y t, (5) h t where r t is the capital rental rate and w t is the real wage. Notice that the parameters α and (1 α) represent the capital and labor share of national income, respectively. 4 Notice that the level of aggregate returns-to-scale in our model, 1+χ, is equal to the markup ratio of price over the marginal cost, which is denoted as 1/ρ in DHL. Therefore, in addition to representing the degree of productive externalities, χ can also be interpreted as the markup. 3

5 2.2 Households The economy is populated by a unit measure of identical infinitely-lived households, each endowed with one unit of time. The representative household maximizes a discounted stream of expected utilities over its lifetime: " # X max E 0 β t log c t η h1+σ t, 0 < β < 1, σ 0 and η > 0, (6) 1+σ t=0 where E is the conditional expectations operator, β is the discount factor, c t is consumption and σ denotes the inverse of the intertemporal elasticity of substitution for labor supply. 5 Households derive income from providing capital and labor services to firms. The only fundamental uncertainty present in the economy is an exogenous shock to government purchases. The budget constraint faced by the representative household is c t + i t + τ t = y t = w t h t + r t k t, (7) where i t is investment and τ t is a lump-sum tax. The law of motion for the capital stock is k t+1 =(1 δ)k t + i t, k 0 given, (8) where δ (0, 1) is the capital depreciation rate. The first-order conditions for the household s optimization problem are given by ηc t h σ t = w t, (9) 1 1 = βe t [ (1 δ + r t+1 )], (10) c t c t+1 lim k t βt t+1 =0, (11) c t where (9) is an intra-temporal condition that equates the household s marginal rate of substitution between consumption and leisure to the real wage. Equation (10) is the standard Euler equation for intertemporal consumption choices and (11) is the transversality condition. 5 The period utility function in DHL is given by log c t + V (L h t), wherel represents the time endowment and V ( ) is a concave function. Here, for expositional simplicity, ³ we normalize the time endowment to one unit, V and specify a particular function form for V with σ 00 (L h), whereh denotes the steady-state labor V 0 hours. Notice that none of the results is sensitive to this modification. 4

6 2.3 Government The government purchases goods and services, and balances its budget each period. Therefore, the period government budget constraint is g t = τ t,whereg t represents government spending that does not contribute to either production or household utility. In addition, g t evolves according to g t+1 =(1 γ) g + γg t + ε t+1, g 0 given and 0 < γ < 1, (12) where g is the steady-state level of government purchases and ε is an i.i.d. shock with mean zero and variance σ 2 ε. Finally, the aggregate resource constraint for the economy is given by c t + k t+1 (1 δ)k t + g t = y t. (13) 2.4 Local Dynamics We focus on symmetric competitive equilibria which consist of a set of prices {r t,w t } t=0 and quantities {c t,h t,k t+1,g t+1 } t=0 that satisfy the household s and firms first-order conditions and the government budget constraint. To analyze the effects of changes in government spending, we begin by deriving the unique interior steady state of the model. Let θ be the exogenously given ratio of government spending to output at the steady state (g/y). With this, it is straightforward to derive the steady-state expressions of all other variables. We then take log-linear approximations to the equilibrium conditions in a neighborhood of this steady state to obtain the following dynamical system: ˆk t+1 ĉ t+1 ĝ t+1 = J ³ˆkt+1 ˆk t ˆk t+1 E t + ĉ t+1 E t (ĉ t+1 ) ε t+1 /g ĉ t ĝ t, ˆk0 and ĝ 0 given, (14) where hat variables denote percent deviations from their steady-state values, and J is the Jacobian matrix of partial derivatives of the transformed dynamical system. For the purpose of this paper, we are mainly interested in investigating local stability of equilibria for the large-markup parameterization that DHL have postulated. Each period in the model is taken to be one year (as in DHL) or one quarter. Following DHL, the capital shareofnationalincome,α, is chosen to be 0.29; the labor supply elasticity parameter, σ, is set to be 0 (indivisible labor); the degree of productive externalities (or markup), χ, is selected as 0.5; the steady-state government spending to output ratio, θ, is fixed at 0.2; and 5

7 the preference parameter η is equal to 3.78 so that the household spends about 30% of its time endowment in working. For DHL s annual specification, we set the discount factor, β, to be 0.96 and the capital deprecation rate, δ, tobe0.1. On the other hand, β =0.99 and δ =0.025 in the quarterly formulation. The local stability of the steady state is determined by comparing the number of eigenvalues of J (one of which is γ) located inside the unit circle with the number of initial conditions (= 2). It turns out that for the annual configuration that DHL have studied, the model exhibits saddle-path stability and equilibrium uniqueness whereby one eigenvalue of J lies outside and the others inside the unit circle. By contrast, all three eigenvalues are inside the unit circle for the quarterly parameterization, which implies that the steady state is indeterminate and becomes a sink. 6 In this case, there is a continuum of stationary rational expectations equilibria in which agents self-fulfilling beliefs (i.e., sunspots) can be an independent source of business cycle fluctuations (see Farmer and Guo, 1994). The intuition for the above result is straightforward. In order to generate positive relationships between government spending shocks and macroeconomic aggregates in the DHL economy, the magnitude of increasing returns (markup) must be large enough to imply that the equilibrium wage-hours locus is positively sloped and cuts the labor supply curve from below. This turns out to be exactly the necessary condition for equilibrium indeterminacy in a laissez-faire one-sector RBC model that Benhabib and Farmer (1994) havederived. 7 In addition, the minimum level of increasing returns needed for the existence of stationary sunspot equilibria is positively related to the discount rate and the capital depreciation rate because higher β and lower δ raise the effective labor demand elasticity (see Wen, 1998). 8 As the time step becomes smaller, labor is drawn more easily out of leisure to help fulfill agents optimistic expectations. Therefore, indeterminacy is more likely to occur when the DHL model is analyzed under a quarterly parameterization. 6 Saddle path stability remains at the quarterly frequency for the other two parameterizations that DHL consider: zero markup (χ =0) and small markup (χ =0.2). 7 See DHL s Figure 1d for an illustration. This can be understood by taking logarithms on both sides of the labor market equilibrium condition (5). The slope of the equilibrium wage-hours locus is given by (1 α)(1+χ) 1, while the slope of the labor supply curve is σ. For the DHL parameterization with α =0.29, χ =0.5 and σ =0, the Benhabib-Farmer condition for local indeterminacy, (1 α)(1+χ) 1 > σ, is satisfied. 8 This is related to the importance of the discount factor and the capital depreciation rate (both of which affect the intertemporal tradeoff between consumption goods at different dates) in optimal growth models that display complicated dynamics (see, for example, Mitra, 1998, and Baierl, Nishimura and Yano, 1998). 6

8 3 Variable Capital Utilization Given empirical findings of Burnside (1996) and Basu and Fernald (1997), it is now well known that the required degree of increasing returns for a one-sector RBC model to exhibit equilibrium indeterminacy is too high to be empirically plausible. This implies that the large markup (χ =0.5) used by DHL is not consistent with these empirical estimates of aggregate returnsto-scale in the U.S. economy. 9 In light of this criticism, Wen (1998) finds that a laissez-faire one-sector RBC model with varying capital utilization can produce multiple equilibria under an empirically realistic level of increasing returns. With a similar motivation, this section shows that adding variable capital utilization to the DHL economy substantially reduces the level of externalities needed to generate procyclical responses of output, consumption, investment, employment, labor productivity and real wage to changes in government purchases. 3.1 The Economy We incorporate variable capital utilization àlagreenwood, Hercowitz and Huffman (1988) into the above model. As a result, the rate of capital depreciation is no longer a constant. Consider a representative household that maximizes its expected lifetime utilities (6), subject to c t + k t+1 (1 δ t )k t + g t = y t = z t (u t k t ) α h 1 α t, 0 < α < 1, (15) δ t = 1 φ uφ t, φ > 1, (16) z t =(ū t kt ) αχ h(1 α)χ t, χ 0, (17) where u t denotes the rate of capital utilization that is endogenously determined, and g t follows the stochastic process as in (12). In addition, since φ > 1, (16) postulates that the rate of capital depreciation δ t (0, 1) is an increasing function of capital utilization, and the productive externality z t is specified as a function of economy-wide average levels of utilized capital and labor inputs. In a symmetric equilibrium where k t = k t, h t = h t and u t =ū t,the aggregate production function is 9 This appears to be the main reason why DHL do not interpret choosing their model s parameter values as a calibration exercise (p. 246). 7

9 h i y t = (u t k t ) α h 1 α (1+χ) t, (18) and the first-order conditions are given by the transversality condition (11) and 10 ηc t h σ t =(1 α) y t h t, (19) 1 1 = βe t [ (1 δ t+1 + α y t+1 )], (20) c t c t+1 k t+1 α y t = u φ 1 t k t, (21) u t where (19) governs the labor supply decision and (20) is the consumption Euler equation. Moreover, (21) equates the marginal gain (more output) and marginal loss (higher capital depreciation) of a change in u t. Rearranging (21) shows that the equilibrium rate of capital utilization is an increasing function of the marginal product of capital, α y t k t. Finally, combining (18) and (21) yields the following reduced-form social technology as a function of capital and labor inputs: where α(1+χ)(φ 1) φ α(1+χ) α(1+χ)(φ 1) φ α(1+χ) φ α(1+χ) k y t = α α(1+χ) t φ(1 α)(1+χ) φ α(1+χ) ht, (22) < 1 to guarantee the existence of an interior steady state since this condition implies diminishing marginal product of capital. Notice that in comparison with (3), (22) displays a larger labor elasticity of output for all χ 0, and a higher level of aggregate returns-to-scale with respect to k t and h t. As a result, variable capital utilization amplifies the impact of changes in government spending since it enriches the model s endogenous propagation mechanism by providing an additional channel to change output. Following the procedure described in section 2, we first derive the steady-state interest rate, hours worked and capital stock as follows: r = 1 β (1 δ), (23) 10 Notice that in a decentralized equilibrium of the model, the capital rental rate, r t,andtherealwage,w t, are given by (4) and (5), respectively. 8

10 k = ½ h = (1 α) r η [(1 θ) r αδ] αh (1 α)(1+χ) r α(1+χ) φ φ ¾ 1 1+σ, (24) 1 1 α(1+χ), (25) where δ and θ are the (exogenously given) capital depreciation rate and government spending to output ratio at the steady state. Given (23)-(25), the steady-state expressions of all the remaining variables can then be easily derived. Next, we log-linearize the equilibrium conditions around the steady state to obtain a dynamical system with the same form as (14). To maintain comparability with DHL and other existing studies, we restrict our attention to the cases with a unique equilibrium. For the quantitative analysis, we use the DHL (annual) parameterization in which α =0.29, σ =0, θ =0.2, η =3.78, β =0.96, together with the steady-state capital depreciation rate, δ = 0.1. The selected values of β and δ imply that φ = Regarding the degree of productive externalities, we consider two cases, (i) constant returns-to-scale or zero markup, χ =0, and (ii) mild increasing returns or small markup, χ =0.1, which can be characterized as empirically plausible vis-à-vis recent empirical estimates reported by Burnside (1996) and Basu and Fernald (1997). 11 In both formulations, the steady state is a saddle point, hence the model exhibits a unique rational expectations equilibrium. Furthermore, the model continues to exhibit saddle path stability in either case at the quarterly frequency when β =0.99 and δ = A Permanent Change in Government Spending Table 1 presents the effects on the steady-state labor supply, capital stock (investment), output and consumption (labor productivity and real wage) of a one-percent permanent increase in the share of government purchases, i.e., raising θ to By contrast, the small-markup parameterization in DHL uses χ =0.2, which is at the upper end of the empirically plausible range. 12 The minimum level of externalities needed to generate multiple equilibria for the annual parameterization is As is discussed above, this value is higher than that required for the quarterly specification, See Wen (1998) for an analysis of the region of indeterminacy in this model without government. 9

11 Table 1: Effects of A Permanent Government Spending Shock 13 h k, i y c, y/h,w χ =0 1.71% 1.71% 1.71% no change χ = % 1.96% 1.96% 0.25% As is well known in the literature, a permanent increase in the share of government spending raises the steady-state labor supply because of a negative wealth effect. In addition, this response is independent of the externality parameter χ because it does not enter the steadystate expression of hours worked (see equation 24). In the standard RBC model with constant returns-to-scale (χ =0), a higher labor supply shifts up the marginal product schedule for capital, thus increasing investment and capital accumulation along the transition path. In the long run, there will be higher levels of capital and labor inputs, but the capital to labor ratio remains unchanged (see Baxter and King, 1993). Therefore, capital, investment and output all rise by the same percentage as labor hours at the steady state. On the other hand, a constant capital to labor ratio in the long run implies that the relative factor price ratio, w/r, is fixed. Since r is invariant to movements in θ (see equation 23), the real wage, consumption and labor productivity do not change at the steady state. As in the above benchmark case, a higher θ raises the long-run labor supply, capital stock and aggregate output in the specification when χ =0.1. However, the quantitative results here are quite different. In particular, combined with variable capital utilization, this level of productive externalities is sufficiently strong to generate an endogenous increase in labor productivity since under this parameterization, the elasticity of output with respect to labor φ(1 α)(1+χ) input, φ α(1+χ) in (22), is greater than one. It follows that both consumption and the real wage also rise in the long run. By contrast, χ needs to be around 0.5 for a permanent increase in θ to crowd in consumption and raise the real wage in the DHL economy without varying capital utilization. Moreover, for all admissible levels of productive externalities, a higher steady-state labor supply leads to a more than proportional rise in the capital stock (see equation 25) and investment. Finally, to maintain the constant marginal product of capital (= r) at the steady state, output will be increased by the same percentage as the capital stock in the long run. Notice that for any given χ 0, the impact on the steady-state output, 13 Since i = δk at the steady state, investment and capital stock exhibit the same percentage responses. Moreover, in our parameterization with indivisible labor (σ =0), consumption is proportional to labor productivity and the real wage (see equations 19 and 5). Hence, c, y/h and w are listed in the same column. 10

12 employment, capital stock, and investment to a permanent government spending shock are qualitatively identical in DHL and the current setting. 3.3 A Temporary Change in Government Spending This section examines the macroeconomic effects of a one-time positive innovation to government purchases, described by (12), equal to 1 percent of initial (steady-state) output. As in DHL, we study two possibilities: (i) low persistence, γ =0.6, and (ii) high persistence, γ =0.95, of government spending shocks for each of the two returns-to-scale considered. Figure 1 plots the impulse response functions in the low-persistence (γ =0.6)case. 14 When there is no externality (χ =0), a temporary increase in government spending, financed by lump-sum taxation, lowers the amount of net resources available to the economy. Therefore, the representative household will decrease its consumption and investment expenditures at the impact period. This negative wealth effect shifts the labor supply curve outward, raising employment and output, and reducing the real wage. Furthermore, a higher labor supply leads to more intensive utilization of capital, which both contribute to an immediate rise in the interest rate(see equation 21). In sum, varying capital utilization does not affect the familiar qualitative results of a standard RBC model in response to a transient government spending shock with low persistence. In the low-externality formulation when χ =0.1, the output response is about twice as large as that in the above benchmark case because increasing returns strengthens the expansionary effect of the spending shock. Moreover, the outward shift of the labor supply curve is now more than offset by an endogenous increase in labor productivity coming from externalities and variable capital utilization. 15 As a result, consumption and the real wage both rise on impact, peak after three periods and then fall back gradually to the steady state. On the other hand, the effect of a higher labor supply dominates the increase in the interest rate, resulting in a sharp positive initial response of investment. This stimulates capital accumulation, which in turn causes the interest rate to display a cyclical convergent pattern towards the steady state. Notice that falling consumption is accompanied by the interest rate below its steady-state value. 14 Given our calibration of indivisible labor, consumption, labor productivity and the real wage exhibit the same percentage responses. 15 In other words, mild increasing returns-to-scale together with varying capital utilization raise the labor elasticity of output to be above one (see equation 22). Hence, in equilibrium, the marginal product of labor is an increasing function of employment. 11

13 Figure 2 presents the dynamic responses of aggregate variables in the high-persistence (γ =0.95) case. Compared to the low-persistence configuration, the negative wealth effect of a positive transient shock to government spending is stronger, as is shown by Aiyagari, Christiano and Eichenbaum (1992). Consequently, employment rises more than before, thereby generating a higher and more persistent output response for both returns-to-scale that we consider. When the model exhibits constant returns-to-scale, a more responsive labor supply leads to a rise in investment, a stronger increase in the interest rate, and a larger fall of consumption and the real wage. On the other hand, the impulse responses in the low-externality specification are qualitatively identical to those when γ =0.6, i.e., consumption, investment, labor productivity, the real wage and the interest rate all rise immediately following the shock. In fact, even when the shock is purely transitory(γ =0), a temporary increase in government purchases can raise the real wage, and crowd in consumption and investment at this level of productive externalities (χ =0.1) due to an endogenous increase in labor productivity. 16 Overall, as in DHL, our result that a one-sector RBC model can produce procyclical movements in macroeconomic aggregates to a permanent or temporary government spending shock is driven by an endogenous, positive response of labor productivity. In the parameterization with logarithmic utility and indivisible labor, DHL find that such response requires a large markup around 1.5. On the contrary, this section shows that, ceteris paribus, a mild increasing returns of 1.1 is strong enough to yield the same outcome because variable capital utilization amplifies the effects of a spending shock by providing an additional margin to change output. 4 Conclusion DHL have shown that a one-sector RBC model with sufficiently strong increasing returns is capable of producing qualitatively realistic business cycles driven solely by changes in government purchases. Our analysis illustrates a close relationship between DHL and a growing literature that explores indeterminacy and sunspots in the RBC framework (see Benhabib and Farmer, 1999, for a survey). Specifically, the condition which governs the minimum degree of increasing returns needed for government spending shocks to generate positive comovements among output, consumption, investment and employment is identical to that necessary for 16 With an externality in the indeterminacy region, Benhabib and Wen (2000) show that a one-sector RBC model with varying capital utilization can display hump-shaped, trend reverting impulse responses to transitory demand shocks for output, investment and hours that mimic those observed in the U.S. data. 12

14 a laissez-faire one-sector RBC model to exhibit equilibrium indeterminacy. In addition, not only does variable capital utilization substantially reduce the required magnitude of aggregate returns-to-scale for the existence of stationary equilibria in a one-sector RBC economy under laissez-faire, it also enhances the endogenous propagation mechanism in the DHL model so that their results can be obtained at an empirically plausible level of increasing returns. 13

15 References [1] Aiyagari, S.R., L.J. Christiano and M. Eichenbaum (1992), The Output, Employment, and Interest Rate Effects of Government Consumption, Journal of Monetary Economics, 30, [2] Baierl, G., K. Nishimura and M. Yano (1998), The Role of Capital Depreciation in Multisector Models, Journal of Economic Behavior and Organization, 33, [3] Basu, S. and J.G. Fernald (1997), Returns to Scale in U.S. Production: Estimates and Implications, Journal of Political Economy, 105, [4] Baxter, M. and R.G. King (1993), Fiscal Policy in General Equilibrium, American Economic Review, 83, [5] Benhabib, J. and R.E.A. Farmer (1994), Indeterminacy and Increasing Returns, Journal of Economic Theory, 63, [6] Benhabib, J. and R.E.A. Farmer (1999), Indeterminacy and Sunspots in Macroeconomics, in Handbook of Macroeconomics, edited by J. Taylor and M. Woodford, , Amsterdam: North Holland. [7] Benhabib, J. and Y. Wen (2000), Indeterminacy, Aggregate Demand and the Real Business Cycle, unpublished manuscript, New York University and Cornell University. [8] Burnside, C. (1996), Production Function Regressions, Returns to Scale, and Externalities, Journal of Monetary Economics, 37, [9] Devereux, M.B., A.C. Head and B.J. Lapham (1996), Monopolistic Competition, Increasing Returns, and the Effects of Government Spending, Journal of Money, Credit, and Banking, 28, [10] Farmer, R.E.A. and J.-T. Guo (1994), Real Business Cycles and the Animal Spirits Hypothesis, Journal of Economic Theory, 63, [11] Greenwood, J., Z. Hercowitz and G.W. Huffman (1988), Investment, Capital Utilization, and the Real Business Cycle, American Economic Review, 78, [12] Mitra, T. (1998), On the Relationship Between Discounting and Complicated Behavior in Dynamic Optimization Models, Journal of Economic Behavior and Organization, 33, [13] Rotemberg, J. and M. Woodford (1992), Oligopolistic Pricing and the Effects of Aggregate Demand on Economic Activity, Journal of Political Economy, 100, [14] Wen, Y. (1998), Capital Utilization under Increasing Returns to Scale, Journal of Economic Theory, 81,

16

17