George Alogoskoufis, Dynamic Macroeconomics, 2016 Chapter 7 Externalities, Human Capital and Endogenous Growth In this chapter we examine growth models in which the efficiency of labor is no longer entirely exogenous, but may be influenced either through positive externalities from the accumulation of physical capital, or through the accumulation of human capital, or even through the generation of new ideas and innovations. These models, under certain conditions, can lead to endogenous growth, unlike the models that we have analyzed so far. 1 Initially we present a growth model which is based on the assumption of positive externalities from aggregate capital accumulation on labor efficiency. The main idea that drives this model is learning by doing, an idea introduced to growth models by Arrow (1962). This assumption can, under certain conditions, lead to endogenous growth, as in Romer (1986). In the learning by doing model, labor efficiency is a function of both exogenous technical progress, as well as aggregate capital per worker. Thus, the efficiency of labor, which is the same for all firms, depends on capital per worker in the rest of the economy. Because of learning by doing, as suggested by Arrow (1962), the accumulation of aggregate capital increases labor productivity both directly and indirectly, through knowledge spillovers, that have a direct effect on the efficiency of labor. It is assumed here that "knowledge" is like a public good, and that the accumulation of knowledge depends on the accumulation of aggregate capital. An important consequence of this approach is that diminishing returns from capital accumulation set in more slowly, and that, under certain conditions, there may even be constant or increasing returns from capital accumulation. In these latter circumstances growth becomes endogenous and is determined by the rate of accumulation of aggregate physical capital. To analyze the impact of the savings behavior of households in the process of endogenous growth, we combine this approach with the assumptions of the Solow model, for a fixed savings and investment rate, as well as with the savings assumptions of the Ramsey representative household model and the overlapping generations model of Blanchard and Weil. We then present an alternative class of growth models, which emphasizes the dependence of labor efficiency on education and training, and the accumulation of human capital. We start with a generalized Solow model, in which the accumulation of human capital is the result of spending on education and training. The model is due to Mankiw, Romer and Weil (1992), and generalizes the Solow model in two directions. First, labor efficiency depends in part on exogenous Throughout the chapter we shall mainly refer to effects on the efficiency of labor, but all the results would go through 1 if instead we referred to effects on total factor productivity. As we use a Cobb Douglas production function throughout, labor augmenting technical progress (Harrod neutral), and total factor productivity augmenting technical progress (Hicks neutral) are equivalent.
technical progress and in part on the accumulated human capital of workers. Second, the accumulation of human capital depends on spending on education and training, which is a fixed, exogenous share of total output. Thus, in this model, there is investment in both physical and human capital. Under certain conditions, this generalized Solow model can also lead to exogenous growth. We also present an alternative model in which investment in human capital depends not on the share of income spent on education and training, but on the amount of time devoted to education and training. These models (Lucas 1988, Jones 2002) assume that workers spend part of their time in education and training, and that this investment of their time increases their human capital and the efficiency of labor. Under certain conditions, these models can also lead to endogenous growth. Finally, we present a class of models in which technical progress is the result of investment in research and development. These models also emphasize the externalities associated with the production of new ideas and innovations that increase labor efficiency. Although models of research and development have existed since the late 1960s, their full economic implications for the functioning of markets were developed in the early 1990s, based on the work of Romer (1990). Under certain conditions, these models can also lead to endogenous growth. It is worth noting that endogenous growth models do not necessarily imply convergence in per capita output, as the corresponding exogenous growth models. Initial conditions matter for the evolution of per capita output and income. However, the available empirical evidence from postwar international data (see Mankiw, Romer and Weil 1992 and Barro 1997) suggests that the issue of convergence of per capita incomes cannot be dismissed easily. The convergence observed can be explained by broader exogenous growth models, in which there is learning by doing, or accumulation of human capital and ideas and innovations, but not to an extent that completely neutralizes the diminishing returns from the accumulation of physical and human capital. 7.1 Externalities from Capital Accumulation and Endogenous Growth We start with a class of models in which physical capital accumulation leads to improved labor efficiency, through a process of learning by doing. It is assumed in these models that the accumulation of aggregate capital implies a direct externality that leads to higher efficiency of labor, as a higher aggregate capital stock implies a higher stock of accumulated experience of all workers in the production process. This is what Arrow (1962) termed, learning by doing. In this class of models, capital has both direct effects on labor productivity, but also indirect effects, through the accumulated economy wide efficiency of workers. The indirect impact of capital accumulation on labor efficiency is an example of a positive externality. 7.1.1 Definitions The definitions of the main variables and parameters in this class of models are similar to the ones we have used in our previous models. Y K L h Output (Physical) Capital Stock Labor knowledge, or labor efficiency or human capital per worker!2
! George Alogoskoufis, Dynamic Macroeconomics, 2016 Chapter 7 t n s ρ α A δ r w time (an exogenous continuous variable) the exogenous rate of growth of the labor force the savings rate the pure rate of time preference the direct elasticity of output with respect to capital total factor productivity the depreciation rate of the capital stock real interest rate real wage 7.1.2 The Production Function Production of goods and services takes place through a large number of competitive firms. The production function of firm i is given by, $ Y i (t) = F(K i (t),h(t)l i (t)), i = 1, 2,... (7.1) The production function is characterized by constant returns to scale and satisfies the usual conditions of a neoclassical production function. We can thus write the production function as, y i (t) = f (k i (t),h(t)) (7.2) where, y = Y /L k = K /L h i i i i i i Output per worker Capital Stock per worker Efficiency of labor In what follows we shall assume a Cobb-Douglas production function of the form,! Y i (t) = AK i (t) α (h(t)l i (t)) 1 α, 0<α<1 (7.3) A is total factor productivity, and α the direct elasticity of output with respect to the capital stock. Those two parameters, as well as the efficiency of labor h, are assumed the same for all firms. 7.1.3 Externalities from the Accumulation of Capital Unlike the growth models we have analyzed so far, where labor augmenting technical progress was assumed exogenous, we will assume that the aggregate efficiency of labor is a function of both time and the aggregate capital stock per worker in economy. This dependence is justified on the basis of the learning by doing idea of Arrow (1962). Economies with a high capital stock per worker are characterized by higher labor efficiency as well, because of the accumulated experience of workers in the production process. As noted by Arrow (1962), I advance the hypothesis here that technical change in general can be ascribed to experience, that it is the very activity of production which gives rise to problems for which favorable responses are selected over time. I therefore take cumulative gross!3
investment (cumulative production of capital goods) as an index of experience. Each new machine produced and put into use is capable of changing the environment in which production takes place, so that learning is taking place with continually new stimuli. This at least makes plausible the possibility of continued learning in the sense, here, of a steady rate of growth in productivity. Following Arrow we thus assume that the efficiency of labor is determined by,! h(t) = K(t),! (7.4) L(t) ( e gt ) 1 β 0 β where K is the aggregate capital stock (an index of experience) and L is aggregate employment of labor. From (7.4) a higher aggregate capital stock per worker implies a higher efficiency of labor for all workers in the economy, irrespective of the capital stock employed by the firm that employs a particular worker. The parameter β measures the elasticity of labor efficiency with respect to the aggregate capital stock per worker, and g is the exogenous rate of technical progress. Substituting (7.4) in (7.1), and after we aggregate the production of all firms, we can derive the aggregate production function as,! Y (t) = A(K(t)) α +β (1 α ) (e gt L(t)) 1 (α +β (1 α )) (7.5) From (7.5) it follows that output per worker is given by,! y(t) = A(k(t)) α +β (1 α ) e gt (7.6) where y=y/l and k=k/l. β ( ) 1 α β (1 α ) The shape of this production function is depicted in Figure 7.1, for different values of the parameter β. For β=0, there is no effect of the aggregate accumulation of capital on labor efficiency, and we are back to an aggregate Cobb Douglas production function without externalities. The efficiency of labor depends only on the exogenous rate of technical progress g. For 0<β<1 the accumulation of capital implies a positive externality on the efficiency of labor, but the aggregate marginal product of capital tends to fall as the economy accumulates more capital per worker. Capital accumulation leads to diminishing returns, although the productivity of capital declines at a slower rate than if there were no externalities. For β=1, the aggregate marginal product of capital is constant, equal to total factor productivity A, and is not affected by the accumulation of capital. There are no diminishing returns to capital accumulation as the aggregate marginal product of capital is constant. Finally, for β>1, the marginal product of capital increases with capital accumulation, but this assumption violates the condition of constant returns to scale for the aggregate economy. In the case β<1 the properties of the exogenous growth models we have analyzed so far do not change by much. There are decreasing returns to capital accumulation, and the economy converges to a balanced growth path that can be analyzed in the way that we demonstrated for exogenous growth models in previous chapters. However, if β=1 or β>1, then (7.6) leads to an endogenous growth model.!4
! George Alogoskoufis, Dynamic Macroeconomics, 2016 Chapter 7 In what follows we shall concentrate on the special case β=1, which implies endogenous growth, without violating the assumption of constant returns to scale. In the special case β=1, (7.5) and (7.6) take the form, Y (t) = AK(t) (7.7) y(t) = Ak(t) (7.8) (7.8) implies that aggregate output per worker is a linear function of capital per worker. The marginal and average productivity of capital is constant. Because of the linearity of the aggregate production function, the rate of growth of output per worker, or per capita income g is equal to the rate of growth of capital per worker. The accumulation of capital per worker does not lead diminishing returns. It thus follows that, y (t)! (7.9) y(t) = k (t) k(t) = g where g is now endogenous, and is determined by the rate of accumulation of capital. For the same reason, from (7.7), the rate of growth of aggregate output g+n, is determined by the share of net investment to total output. As a result, we shall have that,! g + n = Y (t) / Y (t) = K (t) / K(t) = (I(t) / K(t)) δ = A(I(t) / Y (t)) δ (7.10) where I denotes gross investment in physical capital and δ the depreciation rate. 2 7.1.4 Determination of the Real Interest Rate and the Real Wage Assuming that firms operate under perfect competition, they maximize profits by taking the prices of inputs as given. Profit maximization implies that the real interest rate will be equal to the marginal product of capital for individual firms, and the real wage to the marginal product of labor for individual firms. From the point of view of each individual firm, the efficiency of labor is exogenously given. Thus, from (7.3), the net marginal product of capital for each individual firm is given by, α 1 1 α α Ak (t) h(t) -δ i Substituting h(t) from (7.4), with the assumption that β=1, we get the marginal productivity condition, r(t) = α Ak i (t) α 1 k(t) 1 α δ (7.11) 2 This endogenous growth model is known as the ΑΚ Model, from the form of the aggregate production function (7.7).!5
!! George Alogoskoufis, Dynamic Macroeconomics, 2016 Chapter 7 where r(t) is the real interest rate, and the expression on the right hand side is the net marginal product of capital from the vantage point of each individual firm. (7.11) implies that all firms will select the same capital per worker, as they face the same real interest rate and the same labor efficiency per worker. Therefore, k (t) = k(t) i. As a result, the real interest rate will be equal to, i r(t) = α A δ (7.12) The real interest rate is constant and equal to the private marginal product of capital, as calculated by each individual firm. Because of its small size, each competitive firm does not take into account the effects of its own choice of capital per worker on the aggregate capital stock per worker in the economy, treating labor efficiency as exogenously given. Labor efficiency depends on the aggregate capital stock per worker and not on the firm specific capital per worker. This externality is the reason why each firm underestimates the social marginal product of capital, and why the real interest rate, as determined in competitive markets, is lower than the social (aggregate) marginal product of capital. We shall return to this market failure below, when we discuss the efficiency of the competitive equilibrium. In the competitive equilibrium, the real wage will be equal to the marginal product of labor as calculated by each individual firm. It is thus determined by,! w(t) = (1 α )Ak i (t) α k(t) 1 α = (1 α )Ak(t) = (1 α )y(t) (7.13) where w(t) is the real wage per worker, and we have made use of the property that in the competitive equilibrium k (t) = k(t) i. i The real wage is a constant share of output per worker. If output per worker is growing at a rate g, then the real wage per worker will also be growing at a rate g. Note that this endogenous growth model predicts a constant real interest rate and growing real wages, in agreement with the Kaldor stylized facts analyzed in Chapter 2. However, unlike the exogenous growth models we have analyzed so far, this model does not predict convergence. As we shall see, the growth rate is constant and endogenously determined. Therefore, an economy that starts with low capital per worker, and one that starts with high capital per worker, will not experience convergence to the same per capita output and income, even if they have the same technological and population parameters and the same preferences of consumers. They will have the same endogenous growth rate, but they will not converge to the same per capita output and income (see Section 7.1.10 below). 7.1.5 The Savings Rate and the Endogenous Growth Rate Let us first assume that, as in the Solow model, consumer behavior is described by a constant savings rate. Per capita consumption is thus given by, c(t) = (1 s)y(t) (7.14)!6
! George Alogoskoufis, Dynamic Macroeconomics, 2016 Chapter 7 where s is the savings rate. In equilibrium, total output will be equal to consumption plus gross investment. In per capita terms,! y(t) = c(t) + k (t) + (n + δ )k(t) (7.15) Combining (7.8), (7.14) and (7.15), the change in the capital stock per worker will be given by,! k (t) = sy(t) (n + δ )k(t) = (sa n δ )k(t) (7.16) Dividing both sides of (7.16) by k(t) we get that, g = k (t) = sa (n + δ ) k(t) (7.17) (7.17) determines the endogenous growth rate g. The higher the savings rate s, and total factor productivity A, the higher the growth rate of per capita income. On the other hand, population growth (employment) and the depreciation rate have a negative impact on the endogenous growth rate. In this model the savings rate plays a similar role to its role in the exogenous growth Solow model. In the exogenous growth Solow model, the savings rate has a positive impact on steady state capital per effective unit of labor k*, and the growth rate during the convergence process towards k*, but does not affect the long-term growth rate, which is equal to the exogenous rate of technical progress g. In this endogenous growth Solow model, with positive externalities from capital accumulation, the saving rate s determines the gross investment rate, and, through the gross investment rate the endogenous growth rate, as the accumulation of capital does not imply diminishing returns for the marginal product of capital, and growth continuous for ever. Therefore, in this endogenous growth Solow-Arrow-Romer model, "poor" and "rich" economies in terms of initial capital, will not converge to the same per capita income, even if they have the same savings rate, the same total factor productivity, the same rate of population growth and the same depreciation rate. They will simply have the same steady state growth rate, without converging to the same per capita income. 7.1.6 Externalities and Endogenous Growth in the Ramsey Model The assumption of an exogenous savings rate is not theoretically satisfactory. The savings rate is determined by underlying factors associated with the preferences and constraints of households. How is the endogenous growth determined in a representative household model, where households maximize the intertemporal utility from consumption? We assume a representative household, which maximizes the intertemporal utility function,!7
! George Alogoskoufis, Dynamic Macroeconomics, 2016 Chapter 7 c(t) 1 θ 1! U = e (ρ n)t dt (7.18) t=0 1 θ under the household budget constraint,! k(t) = r(t)k(t) + w(t) c(t) nk(t) (7.19) Variables in (7.18) and (7.19) are defined per capita. Income per capita is capital income per capita, plus labor income per capita. Each household member supplies one unit of labor. From the first order conditions for a maximum, we get the Euler equation for consumption (see Chapter 2). c (t)! (7.20) c(t) = 1 ( θ r(t) ρ) Given that the real interest rate is constant and equal to αa-δ, from (7.12), the rate of growth of per capita consumption is also constant, and given by, c (t) c(t) = 1 ( θ α A δ ρ) = g (7.21) (7.21) determines the endogenous growth rate g, as on a balanced growth path all per capita variables grow at the same rate. If that was not so, either consumption or investment would eventually grow to be equal to output. From (7.21), a higher pure rate of time preference of households, relative to the private marginal product of capital to firms (the real interest rate), results in a lower endogenous growth rate of per capita income and consumption. This is because a higher pure rate of time preference of households implies lower savings and a lower rate of accumulation of capital. On the other hand, a higher total factor productivity A, or a higher private contribution of capital to output α, lead to a higher endogenous growth rate, as both result in a higher equilibrium real interest rate, and higher savings and capital accumulation rates. For the opposite reason, the depreciation rate δ has a negative impact on the endogenous growth rate. To find the savings rate, we will use (7.8) and (7.15). These imply that,! k (t) = Ak(t) c(t) (n + δ )k(t) (7.22) (7.22) is the aggregate equilibrium condition of equality of total output to consumption plus investment. Dividing through by k(t), k (t) $ (7.23) k(t) = A n δ c(t) k(t) = g!8
! George Alogoskoufis, Dynamic Macroeconomics, 2016 Chapter 7 From (7.23) it follows that, c(t) k(t) = A n δ g (7.24) From (7.24), the share of consumption to total output is equal to, c(t) $ (7.25) y(t) = c(t) Ak(t) = 1 n + g + δ A From (7.25) and (7.21), the aggregate savings rate is equal to, $ s = 1 c(t) (7.26) y(t) = n + g + δ = 1 ( α A (1 θ)δ ρ +θn ) A θa The saving rate in this endogenous growth Ramsey model is constant, because the real interest rate is constant. It depends positively on the real interest rate, αα-δ, the rate of population growth, n, and the elasticity of intertemporal substitution in consumption 1/θ, and negatively on the pure rate of time preference of households. The effect of the depreciation rate of capital on the endogenous savings rate depends on the elasticity of intertemporal substitution in consumption. If the latter is equal to one, which implies θ=1, then the depreciation rate of capital has no effect on the savings rate. If the elasticity of intertemporal substitution in consumption is greater than one (θ<1), then the depreciation rate of capital has a negative effect on the savings rate. The opposite happens if the elasticity of intertemporal substitution is lower than one (θ<1). 7.1.7 The Inefficiency of Competitive Equilibrium with Externalities We have already demonstrated in Chapter 3, that in the exogenous growth Ramsey model without externalities, the competitive equilibrium maximizes social welfare. Does this result carry over to the Ramsey model with externalities from capital accumulation, as the endogenous growth model we have analyzed? In order to examine the optimality of the competitive equilibrium, let us assume there is a social planner who maximizes (7.18) under the macroeconomic constraint (7.22) instead of the private constraint (7.19). From the first order conditions for a maximum, the Euler equation for the social planner would take the form, c (t)! (7.19 ) c(t) = 1 ( θ A δ ρ) = g* > 1 ( θ α A δ ρ) = g From (7.19 ) we can deduce that the endogenous growth rate in the competitive economy is lower than the socially efficient growth rate. This is because the competitive real interest rate (αα-δ) underestimates the social net marginal product of capital (Α-δ), which, because of the positive externality from capital accumulation, is higher than the private net marginal product of capital.!9
! George Alogoskoufis, Dynamic Macroeconomics, 2016 Chapter 7 Since the positive externality from capital accumulation is not reflected in the real interest rate, households have a smaller incentive to save and accumulate capital, and, as a result, the investment rate and the growth rate of the economy are lower than what would be socially optimal. Thus, the competitive equilibrium is not optimal, and could be improved upon by a social planner who would chose savings and capital accumulation taking the externality into account. Various other policies, such as the subsidization of capital could improve efficiency in this case. 7.1.8 Externalities and Endogenous Growth in the Blanchard-Weil Model We next turn to the determinants of endogenous growth in an overlapping generations model with externalities from capital accumulation. We shall utilize the Blanchard Weil model of perpetual youth, where population growth comes from the entry of new households. The number of members of each household is constant, and the rate of entry of new households is equal to n. Every household has an infinite time horizon. The household born at instant j maximizes the intertemporal utility function, U j = t= j e ρt u( c( j,t) )dt (7.27) where, c(j,t) u ρ consumption of household born at instant j at instant t instantaneous utility function pure rate of time preference Following Blanchard (1985) and Weil (1989) we shall assume that the instantaneous utility function is logarithmic. ( ) = lnc( j,t)! u c( j,t) (7.28) All households supply one unit of labor and receive a real wage w, and, to the extent that they have accumulated savings, they also receive capital income. Household j thus maximizes (7.27) under the constraint, k ( j,t) = r(t)k( j,t) + w(t) c( j,t) (7.29) where k(j,j) = 0. From the first order conditions, and after aggregation, we get the following equation for the evolution of aggregate consumption, 3! C (t) = (r(t) + n ρ)c(t) nρk(t) (7.30) 3 See Chapter 3 for a detailed derivation of aggregate consumption in the Blanchard Weil model.!10
!! George Alogoskoufis, Dynamic Macroeconomics, 2016 Chapter 7 From (7.30), after division by L(t), the evolution of per capita consumption is described by,! c (t) = [r(t) ρ]c(t) nρk(t) (7.31) Dividing (7.31) by per capita income y(t), and after we replace the real interest rate from (7.12) and the capital output ratio from (7.8), we get the following equation for the evolution of the ratio of private consumption to total output, c _! (t) = (r(t) ρ g)c _ (t) nρ k _ (t) = (α A δ ρ g)c _ (t) nρa 1 (7.32) where, for any variable Χ, we define its ratio to total output Υ as, x _ = X Y From the per capita capital accumulation equation (7.22) we get, g = 1 c _ (t) A n δ (7.33) (7.33) suggests that the aggregate growth rate g+n is equal to the net savings (=investment) rate, times total factor productivity, minus the depreciation rate. Because of constant returns to capital accumulation total factor productivity is equal to the (constant) average and marginal product of capital. From (7.32) and (7.33) we can determine the two endogenous variables, that is the growth rate and the savings rate. Given that both are non predetermined variables, the equilibrium is achieved instantaneously. For a constant ratio of private consumption to output, (7.32) implies,! c _ nρa 1 (t) = (7.34) αa δ ρ g The higher the growth rate, the greater the ratio of private consumption to total output for a given difference between the real interest rate and the pure rate of time preference of households. The non-linear relationship between the consumption rate and the growth rate is represented by the positively sloped curve in Figure 7.2. Obviously, as g tends to αa-δ-ρ, the ratio of private consumption to total income tends to infinity. The capital accumulation equation (7.33) is represented by the straight, negatively sloped, line. The higher the ratio of private consumption to total income, the smaller the savings rate, and the smaller the investment rate and the endogenous growth rate.!11
The equilibrium is determined at the intersection of the two loci (point E). Since both private consumption and the rate of economic growth are non predetermined variables, the economy jumps directly to the equilibrium E. It is worth noting that the equilibrium ratio of private consumption to output is greater than in the corresponding representative household model, and that the endogenous growth rate is lower. The equilibrium in the corresponding representative household model (n=0) is determined at point ER. It is also relatively simple to show that an increase in the pure rate of time preference of households leads to an increase in the ratio of private consumption to output and a reduction in the endogenous growth rate. 7.1.9 Fiscal Policy and Endogenous Growth The effects of fiscal policy, when we introduce the government, are similar to the effects in the exogenous growth Blanchard Weil model. The exogenous growth version of the model determines the capital stock and consumption per efficiency unit of labor, while the endogenous growth version determines the growth rate and the ratio of consumption to total output. Assuming that the government stabilizes the public debt to output ratio and that it selects a fixed ratio of primary government expenditure to total output, the endogenous growth Blanchard Weil model takes the following form. c _ (t) = (r(t) ρ g)c _ (t) nρ(k _ (t) + b _ ) = (α A δ ρ g)c _ (t) nρ(a 1 + d _ ) (7.35) c _ (t) = 1 c _ g (n + g + δ )A 1 (7.36) (7.35) is the aggregate private consumption function, when households hold assets in both capital and government bonds, and is derived from (7.32). (7.36) is derived from(7.33), with the addition of primary government expenditure, and describes the capital accumulation rate, which is equal to the endogenous growth rate, as a result of total domestic savings. In (7.35) and (7.36) we assume that,! c _ g = c (t) g,! d _ = d(t) y(t) y(t) i.e that the government selects a constant government debt and primary expenditure to output ratio. As a result, the model determines the ratio of private consumption to total output c/y, and the growth rate g. The equilibrium is depicted in Figure 7.3. The dashed positively sloped curve depicts (7.35) for a constant ratio of private consumption to total output, and is given by,!12
c _ (t) = c(t) y(t) = nρ(a 1 _ + d ) (α A δ ρ g) (7.37) The higher the growth rate, the greater the ratio of private consumption to total output, for a given difference between the real interest rate and the pure rate of time preference of households. The non-linear relationship is represented by the dashed upward sloping curve in Figure 7.3. Obviously, as g tends to αa-δ-ρ, the ratio of private consumption to total output tends to infinity. For a positive ratio of public debt to output, this curve lies above the corresponding curve for zero government debt (solid curve). Current households consider government debt as part of their wealth, and thus consume more relative to income, because they know that part of the future taxes to finance government debt will be paid by future generations. The dashed negatively sloped straight line represents the negative relationship between growth (capital accumulation) and the private consumption ratio, since the higher the private consumption to output ratio, the lower ratio of savings and investment to total output. With a positive ratio of primary government expenditure to total output, this line is below the corresponding line for zero primary government expenditure. It is evident that both the ratio of government debt to output and the ratio of primary government expenditure to output have a negative impact on the endogenous growth rate g. The reason is that both government debt and primary government expenditure cause a reduction in total savings and capital accumulation, and thus reduce the growth rate. Government debt causes an increase in private consumption by current generations, reducing the rate of accumulation of capital and the growth rate (point Eb in Figure 7.3). Primary government expenditure causes a smaller decline in private consumption, reducing total savings, the rate of capital accumulation and the growth rate (point Ecg in Figure 7.3). The combination of the two is even greater reduction in the endogenous rate of economic growth. In the corresponding representative household model (with n=0), the rate of economic growth is not affected by the level of government debt and primary government expenditure. The endogenous growth rate is determined by the difference of the real interest rate of the pure rate of time preference (point gr in Figure 7.3). Primary government expenditure crowds out an equal amount of private consumption, and thus does not affect aggregate savings, while government debt does not affect private consumption as it is considered equivalent to the present value of future taxes for the representative household. 7.1.10 Exogenous and Endogenous Growth Models and Convergence We have already argued that, unlike the exogenous growth models, in endogenous growth models of the AK form, such as those examined in this section, there is no convergence process. In exogenous growth models, any two economies characterized by the same parameters describing the technology of production, household preferences and economic policy, will converge to the!13
same balanced growth path, even if they start from different initial conditions. In endogenous growth models they will have the same endogenous growth rate, but they will not converge to the same per capita income. Their initial differences will remain for ever. The analysis of this is presented in Figure 7.4, where it is assumed that two, otherwise identical, economies are characterized by different initial conditions. Economy 1 has initially less capital and output per worker in relation to economy 2. In terms of the production technology, the preferences of consumers and fiscal policy the two economies are identical. If the two economies are characterized by the conditions that apply to an exogenous growth model, then they both converge to the balanced growth path y*(t). Initially, economy 1 will have a higher growth rate of output per head than economy 2. Gradually the two economies will converge to the balanced growth path y*(t), and will be characterized by the same per capita output and the same growth rate of per capita output. If the two economies are characterized by the conditions that apply to an endogenous growth model, then their initial differences will remain. Economy 1 and 2 economy will be characterized by the same rate of growth of per capita output, but their initial percentage differences in per capita output, per capita consumption the real wages will persist. Economy 1 will follow the balanced growth path y1(t), and economy 2 the balanced growth path y2(t). Their per capita output, consumption and real wages will be growing at the same rate, but they will not converge. The paths of per capita output of two such economies (in logarithms), under exogenous and under endogenous growth, are illustrated in Figure 7.4. It is worth noting that the available empirical evidence from post war international experience indicates that convergence cannot be dismissed easily. 4 7.2 Investment in Human Capital and Economic Growth In the learning-by-doing endogenous growth model that we presented in the previous section, endogenous growth is essentially a by-product of the accumulation of physical capital, because of the assumption that labor efficiency is a function of the aggregate physical capital stock per worker. An alternative class of growth models emphasizes the education and training of workers and the accumulation of human capital that it implies. In this section we will present a number of alternative models in which the efficiency of labor depends on investment in human capital, through education and training. 7.2.1 The Generalized Solow Model and Spending on Education and Training We start with a model in which the accumulation of human capital is the result of spending on education and training. The model is due to Mankiw, Romer and Weil (1992), and is a generalization of the Solow model, to allow for the accumulation of human as well as physical 4 See for example Mankiw, Romer and Weil (1992) and Barro (1997).!14
capital. The efficiency of labor depends in part on exogenous technical progress and in part on human capital, which accumulates due to investment in education and training. Firms are competitive, and their production technology is Cobb Douglas, of the form,! Y (t) = AK(t) α a(t)l(t), 0<α<1 (7.38) where a(t) denotes the efficiency of labor at instant t. We shall assume that the efficiency of labor is a function of human capital per worker, and exogenous technical progress. This function takes the form, ( ) γ e gt ( ) 1 α ( ) 1 γ 0 γ 1! a(t) = h(t),! (7.39) where h is human capital per worker, resulting from accumulated education and training, and γ a parameter which measures the relative contribution of human capital in the efficiency of labor. A fraction 1-γ of the efficiency of labor depends on exogenous technical progress, which grows at a rate g. Substituting (7.39) in (7.38), we get the following aggregate production function, ( ) (1 γ )(1 α )! Y (t) = AK(t) α H (t) γ (1 α ) e gt L(t) (7.40) where H=hL is aggregate human capital. We will assume, generalizing the Solow model, that total household income is either consumed or invested in physical capital or invested in human capital. The percentage of output and income invested in physical capital equals sk, and the percentage invested in human capital equals sh. These percentages are considered exogenous, in the same way that the savings rate is assumed exogenous in the Solow model. Furthermore we assume that δ, the depreciation rate, is the same for both physical and human capital. With these assumptions, the accumulation equations for physical and human capital are given by,! K (t) = s K Y (t) δ K(t) (7.41)! H (t) = s H Y (t) δ H (t) (7.42) If γ=0, then we get the familiar neoclassical production function, as spending on education and training does not contribute to labor efficiency of workers, which is a simple function of time (exogenous technical progress). If 0<γ<1, we have a generalized model of exogenous growth, in which the per capita capital stock and output on the balanced growth path are positive functions of both the proportion of total income spent on physical capital accumulation sk, and the proportion spent on education and training sh.!15
The equilibrium condition in the goods and services market in this case is given by, ( )! Y (t) = C(t) + K (t) + H (t) + δ K(t) + H (t) (7.43) which, in conjunction with (7.41) and (7.42) implies a consumption function of the form,! C(t) = (1 s K s H )Y (t) (7.44) In the case where 0<γ<1, we can divide (7.40), (7.41), and (7.42) by the number of workers L times the exogenous labor efficiency e gt, and the model takes the form,! y (t) = A k (t) α h (t) γ (1 α ) (7.45) k! (t) = s K y (t) (n + g + δ )k (t) (7.46) h! (t) = s H y (t) (n + g + δ )h (t) (7.47) where,! y ~ (t) = Y (t),! k ~ (t) = K(t),! h ~ (t) = H (t). e gt L(t) e gt L(t) e gt L(t) 7.2.2 The Balanced Growth Path in the Generalized Solow Model On the balanced growth path, all per capita variables grow at the rate of exogenous technical progress g. It therefore follows that, α γ (1 a)! s K y ~ * = s K A k ~ * h ~ * = (n + g + δ )k ~ * (7.48) α γ (1 a)! s H y ~ * = s H A k ~ * h ~ * = (n + g + δ )h ~ * (7.49) From (7.48), (7.49), and the production function (7.45), it follows that,! k ~ * = s K h ~ * (7.50) s H From (7.50) it follows that on the balanced growth path, the ratio of physical to human capital is constant, and equal to the ratio of the investment rates in physical and human capital respectively. This follows because equilibrium investment on the balanced growth path is a proportion n+g+δ of both physical and human capital, per exogenous efficiency unit of labor. Substituting (7.50) in the production function and solving (7.48) and (7.49),!16
( ) 1! h ~ * = A s α 1 α K s H (1 γ )(1 α ) (7.51) n + g + δ ( ) 1! k ~ * = A s 1 γ (1 α ) γ (1 α ) K s H (1 γ )(1 α ) (7.52) n + g + δ ( ) 1! y ~ A s α γ (1 α ) K s H (1 γ )(1 α ) * = α +γ (1 α ) (7.53) ( n + g + δ ) From (7.53), per capita output on the balanced growth path is given by,! y *(t) = y ~ A s α γ (1 α ) *e gt K s H (1 γ )(1 α ) = α +γ (1 α ) e gt (7.53 ) ( n + g + δ ) where,! y(t) = Y (t). L(t) ( ) The level of per capita output on the balanced growth path depends positively on total factor productivity A, the shares of output invested in physical and human capital (sk και sh), and negatively on the population growth rate n, the rate of exogenous technical progress g, and the depreciation rate δ. The rate of growth of per capita output on the balanced growth path is equal to the rate of exogenous technical progress g. Thus, in the case that 0<γ<1 we have a generalized Solow model of exogenous growth. 7.2.3 Endogenous Growth in the Generalized Solow Model If γ=1, there is no exogenous technical progress. We then move to an endogenous growth model. Setting γ=1 in the production function (7.39) we get,! y(t) = Ak(t) α h(t) 1 α (7.54) where, y=y/l, k=k/l και h=h/l. In this model, the balanced growth path would be defined by, 1 k *(t)! (7.55) h *(t) = s K s H!17
! g = y *(t) (7.56) y *(t) = As α Ks 1 α H (n + δ ) On the balanced growth path, the ratio of physical to human capital is stabilized at the ratio of the investment rates in physical and human capital. Because both physical and human capital are growing at the same rate, we have endogenous growth. The accumulation of physical capital causes an increase in income, that in turn causes an increase in human capital through expenditure on education and training. This in turn leads to a further rise in output, which causes new investment in physical capital. The parallel accumulation of physical and human capital leads to endogenous growth. The endogenous growth rate depends positively on total factor productivity A, and a weighted average of the income ratios invested in physical and human capital sk and sh. The rate of growth of population n, and the depreciation rate δ, have a negative impact on the endogenous growth rate. In all other respects, the properties of the endogenous growth, generalized, Solow model, are similar to those of the learning by doing model of Arrow and Romer, as analyzed in the previous section. The main difference is that in this generalized Solow model, the accumulation of human capital is not a simple by-product of physical capital accumulation, but the result of explicit investment on human capital, through education and training. 7.2.4 The Jones Model of Human Capital Accumulation We next present an alternative model, in which human capital depends on the time span devoted to education and training. We will assume that workers spend part of their time ν in education and training, and that this investment of time involves a rate of return ψ. The model is analyzed in Jones (2002). The production technology is described by the production function (7.38). The efficiency of labor per employee is determined by,! a(t) = h(v,ψ )e gt (7.42) where the function h(v,ψ) determines human capital per worker, as a function of the time interval ν devoted to education and training, and the rate of return of investment in human capital ψ. A plausible form of this function is the exponential function. We shall therefore assume that,! h(v,ψ ) = e ψ v (7.43) The exponential form implies that the proportional change in the function with respect to the time devoted to education and training is equal to the rate of return ψ. From (7.43) we get, h(v,ψ )! =ψ e ψ v =ψ h(v,ψ ) (7.44) v!18
Substituting (7.43) in (7.42) and the resulting equation in the production function, we get that output per worker is determined by,! y(t) = Ak(t) α (e ψ v e gt ) 1 α (7.45) Combining (7.45) with the savings assumption of the Solow model, i.e that there is an exogenous constant savings rate, per capita output on the balanced growth path is determined by, α sa 1 α! y *(t) = A (7.46) n + g + δ e ψ v e gt Per capita output on the balanced growth path is a positive function of the amount of time spent in education and training v, and the rate of return on investment in human capital ψ. However, in other respects, this model is an exogenous growth model, similar to the Solow model. 7.2.5 The Lucas Model of Endogenous Growth An alternative model of investment in human capital is the endogenous growth model of Lucas (1988), which was based, in part, on Uzawa (1965). Lucas assumed that labor efficiency is equal to human capital per worker.! a(t) = h(t) (7.47) With regard to the production of human capital per worker he assumed that this depends on the existing stock of human capital per worker, times the proportion of non-leisure time that workers devoted to education and training. He thus assumed that,! h (t) = ζ (1 u(t))h(t) (n + δ )h(t) (7.48) According to (7.48), for the production of human capital only human capital is required. u is defined as the proportion of the non-leisure time of households that is devoted to the production of goods and services, while 1-u is the proportion of time devoted to education and training. Education and training results in the accumulation of human capital. ζ is the productivity of human capital in the production of new human capital, n is the population growth rate and δ the depreciation rate of human capital, assumed for simplicity to be equal to that of physical capital. In steady state, u will be constant at u*. Taking the integral of (7.48) and substituting in (7.47), we get, ( )t! a(t) = h(t) = e ζ (1 u*) (n+δ ) (7.49) Lucas assumed that u is chosen by a representative household in order to maximize its intertemporal utility of consumption. As a result, the choice of u depends both on the preferences of the representative household, and on the technological parameters characterizing the production of goods and services and human capital.!19
From the form of (7.49), it follows that the endogenous growth rate is equal to g=ζ(1-u*)-(n+δ), which works like the exogenous rate of technical progress in exogenous growth models. The higher is the steady state proportion of non-leisure time devoted to education and training, the higher the endogenous growth rate. In the Lucas model, 1-u* is chosen endogenously, and, in conjunction with the exogenous ζ, δ and n, determines the steady state growth rate of per capita output. 5 7.3 Growth Models based on the Generation of Ideas and Innovations We now turn to a final category of growth models, in which technical progress is a result of ideas and innovations, that lead to higher total factor productivity or labor efficiency. These models emphasize the externalities involved in generating new ideas and innovations that increase the efficiency of production. As the learning by doing model of Arrow emphasizes the externalities from capital accumulation, so the ideas and innovations models emphasize the externalities of the production of ideas and innovations. Although models in this category, sometimes called research and development (R&D) models, date from the late 1960s, the microeconomic foundations of these models and their implications for the functioning of markets were developed in the early 1990s, inspired by the work of Romer (1990). These models, under certain conditions, can lead to endogenous growth as well. 6 7.3.1 Key Features of Ideas and Innovations A crucial assumption of such models is that ideas and innovations can improve the efficiency of production, either through total factor productivity or through labor efficiency. These models recognize that, unlike most other goods and services, the use of an idea and/or innovation by a particular firm, or employee, does not prevent to use of the same idea and innovation from other firms or workers. The use of a particular idea or innovation is non rivalrous, in contrast to the use of a particular machine or a specific employee. From the time an idea or innovation has been produced, anyone with knowledge of this idea can use it, independently of how many others use it simultaneously. If a firm uses a specific machine or a particular employee, it automatically excludes any other firm from simultaneously using this same machine or this same employee. The property of non-rivalry gives ideas and innovations a character of a quasi public good. On the other hand, in contrast to purely public goods, the use of an idea can be partially excludable by law. This allows the producer of an idea to charge for the use of his idea. For example, if an idea or innovation is legally covered by a patent, then a firm or an employee who wants to use this idea or innovation will have to pay a fee to the holder of the patent, for their copyright. 5 See Annex to Chapter 7, for an extensive presentation of the Lucas (1988) model. The dynamic analysis of this model has been generalized by Caballe and Santos (1993), Faig (1995) and Mulligan and Sala-i-Martin (1995). 6 For research and development models of the 1960s see Uzawa (1965), Phelps (1966), Shell (1966) and Nordhaus (1969). For models that build on the ideas of Romer (1990), see Grossman and Helpman (1991), Aghion and Howitt (1992) and Jones (1995).!20
The use of ideas and innovations is not rivalrous, but the degree of excludability that can be achieved varies considerably between different ideas and innovations. For example, the use of an encrypted television signal is easily excludable, while the use of computer software is excludable with higher difficulty. Purely public goods and services are both non rivalrous and non excludable. Other goods and services, such as ideas that can be covered by copyright laws, may be non rivalrous, but may be excludable. Consequently, the producers of goods and services that use an idea or innovation, may be charged for the benefits arising from their use. The production of non rivalrous and non excludable goods and services implies externalities, which are not reflected in the remuneration of producers of these goods and services. Goods and services that result in positive externalities, such as ideas and innovations, will be produced in smaller quantities than would be socially desirable, and goods and services that result in negative externalities, such as pollution of the environment, will be produced in larger quantities than would be socially desirable. If the use of ideas and innovations is both non rivalrous and non excludable, then the market will produce fewer ideas and innovations than would be socially desirable. However, if the use of ideas and innovations can be made excludable by the protection of the law on copyright or a patent, then the production of ideas and innovations can rise, as producers of ideas and innovations will be paid for the value of their product. In the analysis that follows, we shall ignore many of the microeconomic characteristics of the markets for ideas and innovations, and will focus on a simple growth model in which technological progress is simply the result of the production of new ideas and innovations. 7 7.3.2 The Key Elements of an Ideas and Innovations Growth Model In order to present the key elements of a growth model based on the production of ideas and innovations we start with the production function of goods and services. We assume a Cobb Douglas production function, of the form, ( ) 1 α! Y (t) = AK(t) α H (t)l y (t) (7.50) where H is the existing aggregate stock of ideas and innovations, affecting the efficiency of labor in the goods and services sector, and Ly is the number of workers who are employed in the goods and services sector. Apart from goods and services, the economy produces new ideas and innovations. The new ideas and innovations produced per instant are proportional to the number of research workers, i.e those employed in the production of ideas and innovations. This relationship can be written as,! H (t) = h(h (t), L h (t))l h (t) (7.51) 7 Due to the increasing returns implied by ideas and innovations, the assumption of perfect competition is no longer appropriate, and one should ideally use models of imperfect competition. See Romer (1990).!21