Applied Mathematical Sciences, Vol. 8, 2014, no. 180, 8965-8976 HIKARI Ltd, www.m-hikari.com http://dx.doi.org/10.12988/ams.2014.410817 A Generalized Solow Nonautonomous Model with Delay and Bounded Population Growth Corina G. Averbuj and Juan José M. Martínez Center for Research in Economic Theory and Applied Mathematics (CIETyMA) National University of San Martin, Buenos Aires, Argentina Copyright c 2014 Corina G. Averbuj and Juan José M. Martínez. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Abstract In this paper we consider a nonautonomous functional differential equation obtained when the classical Solow model extends by adding a delay in the new capital installed and the active population is not decreasing and asymptotically bounded. The existence of an unique solution is proved and sufficient conditions are given for both stability and boundedness. The evolution of this equation over time and its convergence, or not, to steady states is exemplified through numerical simulations. Mathematics Subject Classification: 37G40, 91B62 Keywords: Solow model; delay; nonautonomous dynamical systems 1 Introduction Economic growth is a dynamic process, focusing on how and why output, capital, consumption and labour population change over time.the temporary nature of it makes it a long-term phenomenon influenced both by the increase in productive resources and by the efficiency in their use. The analysis of the temporal dynamics begins, usually, with the study of the Solow-Swan growth model, ([17] and [18]). This is an abstract and simple representation of a complex economy that captures many real facts. Also, this
8966 Corina G. Averbuj and Juan José M. Martínez model have been in continuous development (see, for instance, [14], [15] and [19]). This work focuses on two main issues affecting long-term growth. The first one, called time-to-build, incorporates a constant delay between investment and new capital installed (see [3], [8] and [10]). The second one incorporates a bounded growth rate in the workforce (see, [1], [6] and [9]). Both points are markedly different in emerging and in developed economies. In an emerging economy delays are higher due to structural weaknesses and limitations, among other factors, and the higher growth of population. On the other hand, in developed economies population growth occurs at increasingly lower rates and there are minor delays between investment and installed capital. Therefore, these features are of vital importance in the comparison of such economies. Our main contribution is to extend the results of Solow (1956), Swan (1956) and Ferrara-Gerrini (2009), proving that under adequate conditions, there is a unique asyntotically stable and boundness solution. The paper is organised as follows. Section 2 introduces the classical Solow model and proposes a model that extends it, proving the existence and uniqueness solution of proposed model. In section 3 we prove that under adequate conditions, the solution is asymptotically stable and bounded. In section 4 we show numerical simulations. 2 Extension of Solow model 2.1 The model The Solow model describes the evolution over time of some kind of gross output (e.g. GDP) Y (t)based upon the input factors, labour L(t), capital K(t), and technology A(t) 1. This model treats market structure as given and technology as exogenous (rather like a black box). Let us suppose that the economy is closed and with a fully employed labour supply. The expression we propose to link the input variables is 2 Y (t) = F (K(t t ), A(t t )L(t t )) (1) 1 Economists normally use the short-hand expression technology to capture factors other than physical and human capital that affect economic growth and performance. 2 The production function F (K(t), L(t), A(t)) is too general to achieve balanced growth, and only some very special types of production functions are consistent with balanced growth. The functional form, F (K(t), L(t)A(t)), namely labor-augmenting or Harrodneutral, implies that an increase in technology A(t) increases output as if this economy would have more labor. Balanced growth in the long run is only possible if all technological progress is Harrod-neutral (see, for instance, [2]).
A generalized Solow non-autonomous model 8967 where F denotes the aggregate production function and t > 0, is a fixed time delay between the investment and the new capital installed. The classical Solow model describes the evolution of the physical capital over time through the differential equation dk = I(t) εk(t), (2) dt where 0 < ε < 1 is the stock capital depreciation rate and I(t) is the saving. If a fraction s of the output is saved, we have I(t) = sy (t) 3. Then we have the dynamic capital equation given by: dk dt = sf (K(t t ), A(t t )L(t t )) εk(t). (3) We consider a change of variables given by k(t) = K(t), which denotes A(t)L(t) the capital per capita and technology. 1 We can write f(k(t)) = F (K(t), A(t)L(t)) = F (k(t), 1) and after A(t)L(t) some computations, we can see the per capita capital dynamics by dk dt = sh(t)f (k (t t )) (ε + γ T + γ L (t)) k (t), (4) where 1. h(t) = e γ T.t L (t t ), L(t) 2. the labour force rate is monotonically decreasing such that γ L (t) = dl (t) dt L(t) 0, (5) t 3. γ T is the constant growth technology rate and df 4. is bounded for k k dk 0, i.e., there exists a constant η > 0 such that df dk η. We note that the per capita production function f is not necessarily concave, so that (4) extends the Solow model. In the following subsection we show the existence and uniqueness of solution defined in [ t, ) 2.2 Existence and uniqueness of solution Let the equation (4) and initial conditions k(t) = φ 0 (t), t [ t, 0], φ 0 C([ t, 0], R) (6) 3 In equilibrium I(t) = Y (t) C(t), where C(t) is consumption.
8968 Corina G. Averbuj and Juan José M. Martínez It is easy to verify that under the assumptions given in the previous section, there exists a unique solution to (4)-(6) in a neighbourhood of (0, φ 0 (0)). See [11]. Remark 2.1 Using the step method, we can see by induction that the solution is unique and defined in [ t, ). For details see [7]. 3 Stability of solution Knowing that the problem has an unique solution, given the role of capital in the initial period [ t, 0], in this section we will study the stability of these long-term solutions. We asked ourselves: if we initially have similar capital endowments, what will happen in the long term with the stock of capital corresponding to the respective solutions? Our main result in this section is to prove that under suitable conditions of the rate of technology and investment rate exogenous growth the solution of our problem is asymptotically stable in t = 0. Definition 3.1 If k(t) : ( t, ) R is a solution of dk (t) = g [t, k(t), k(t dt t )], then k is asymptotically stable in t = 0, if the equilibrium solution of the equation (7), z = 0 is asymptotically stable in t = 0: where dz dt (t) = G [t, z(t), z(t t )], G(t, 0, 0) = 0 (7) G [t, z(t), z(t t )] = g [t, k(t) + z(t), k(t t ) + z(t t )] g [t, k(t), k(t t )] (8) Before showing the solution of equation (6) is asymptotically stable in t = 0, we consider the following proposition Proposition 3.2 Suppose there are R(t) and β(t) continuous functions such that for all t 0, y 1, y 2, u R (y 1 y 2 ) (G(t, y 1, u) G(t, y 2, u)) R(t)(y 1 y 2 ) 2 (9) G(t, y, u) G(t, y, v) β(t) u v (10)
A generalized Solow non-autonomous model 8969 for all t 0, y R, u, v [w, ), w > 0 and the operator G [t, z, z(t t )] = g [t, z + y, z(t t ) + y(t t )] g [t, y, y(t t )] (11) Let R 0 and k constants such that R(t) R 0 < 0 and β(t) π < 1 t 0. R(t) Then the equilibrium solution z = 0 for (6) is asymptotically stable in t = 0 (See [12] and [5]). We now state the main result of this section: Theorem 3.3 There exists R(t) and β(t) continuous functions that satisfy (9) and (10) such that the solution k(t) of (7)-(8) is asymptotically stable in t = 0. Let the operator g(t, k(t), k(t t )) given by s.h(t).f[k (t t )] [ε + γ T + γ L (t)].k (t) (12) Then the operator G(t, z, z(t t )) is given by sh(t)f(z(t t )+y(t t )) (ε+γ T +γ L (t))(z+y) sh(t)f(y(t t ))+(ε+γ T +γ L (t))y Taking R(t) = (ε + γ T + γ L (t)) a simple computation shows that (y 1 y 2 )(G(t, y 1, u) G(t, y 2, u)) = = (y 1 y 2 ) [ (ε + γ T + γ L (t))y 1 + (ε + γ T + γ L (t))y 2 ] = (ε + γ T + γ L (t))(y 1 y 2 ) 2 (13) As R(t) (ε + γ T ) < 0, let R 0 = (ε + γ T ). To find β(t), we see that G(t, y, u) G(t, y, v) = sh(t)f(u) sh(t)f(v) + sh(t)f(v z) sh(t)f(u z) 2sh(t)η u v (14) where we have used that f η. Therefore we can take β(t) = 2sh(t)η. k Finally, to find π < 1 we write β(t) R(t) = t 2sηe γt L(t t ) ε + γ T + γ L (t) L(t) 2sηe γ T t γ T < 1 (15) Let π = 2sηe γ T t and taking suitable γ T and s to π < 1 the following proof γ T using the previous proposition. Moreover, we have the following proposition Proposition 3.4 Under suitable γ T and s, the solution of (7)-(8) is bounded.
8970 Corina G. Averbuj and Juan José M. Martínez To see this, we will use the following theorem: Theorem 3.5 (Theo.3.1 ([20])) Assume that g(t, k, y) = sh(t)f(k(t t )) c(t)k(t) is continuously bounded with respect t, c(t) > 0, and f(k) is Lipschitz m sh(t) continuous with constant m > 0. If lim sup < 1, then the solution t c(t) of (7)-(8) is bounded. Proof of Proposition 3.4 Just take c(t) = ε + γ T + γ L (t), h(t) = se γ T t L(t t ) and the proof follows. L(t) 4 Numerical Simulations In this section we shall reproduce the behavior of the solution of equation (4) 4 for two particular production functions and logistic population growth 5. First, let us consider the most common example of production function used in macroeconomics, the Cobb-Douglas production function 6 F (K(t), A(t)L(t)) = a.k(t) α (A(t)L(t)) 1 α 0 < α < 1, a > 0 (16) and therefore f(k) = ak α. In the original Solow model, i.e. when there is no technological progress or time-to-build, and labor force growths exponentially, the economy always converges to the so called steady state where the capital per capita is constant. This steady state will be completely determined by the saving rate s, the depreciation rate ε, and the rate of population growth γ L (see, for instance, [2] and [4]). Replacing the function given by (4) in equation (16) yields 4 The simulations were done with the MATLAB subroutine dde23. 5 ( ) Namely L(t) = L L and γ 1+e d.r.t L (t) = L (t) = r L, such that d = ln 1+e r.t d L0 1. Where L 0 is the initial population, L( is the population for infinitely large times and r is a growth parameter with γ L (0) = r. 1 L0. L ) 6 It can easily be verified that this production function satisfies continuity, differentiability, positive and diminishing marginal products (F K > 0, F L > 0, F K < 0, F L < 0), and constant returns to scale (F (λk, λal) = λf (K, AL)). In addition to these standard assumptions on the production function, the following boundary conditions, the Inada conditions, are often imposed in the analysis of economic growth and macroeconomic equilibrium: lim F K (K, AL) = and lim F K (K, AL) = 0 for all L > 0 and all A, lim F L (K, AL) = K 0 K L 0 and lim (K, AL) = 0 for all K > 0 and all A. Moreover, F (0, LA) = 0 for all L and A. L F L The assumptions made about the production function ensure the existence of interior equilibria.
A generalized Solow non-autonomous model dk = sah(t)k α (t t ) (ε + γt + γl (t)) k (t), dt Note that one cannot guarantee the the existence of a steady state. The sufficient condition for stability of equation (15) is 7 8971 (17) 2sαae γt t < 1, γt.k01 α (18) where k can be considered as the initial value. As the Fig. 1 shows, although this condition is not met, the solution remains asymptotically stable8. Fig.1 α = 0.5, s = γ T, γ T.t = 0.5, a = 2 L0 = 1, L = 100, r = 0.1 On the other hand, the function ak α, α > 0, a > 0, b > 0 (19) 1 + bk α can be thought of as a generalization of the production function of CobbDouglas (b = 0, 0 < α < 1)9. Thus, from equation (4) and the production function (19) one gets f (k) = where f (k) satisfies bounded derivative for k k. ase γt t 8 The observed values are consistent with k 1 α = γt 9 This function does not necessarily presents diminishing returns and relaxes Inada assumptions, thus generating a richer dynamics. 7
8972 Corina G. Averbuj and Juan José M. Martínez dk dt = sah(t) k α (t t ) 1 + bk α (t t ) (ε + γ T + γ L (t)) k (t), (20) For α 1, df is decreasing and the maximum of the derivative is k dk 0 = k(0). A non-strict bound could be again η = αak 1 α a In contrast, for α > 1, α+1 (α+1) α 4α ( α 1 b 0. df dk is a maximum in kα = α 1 (α+1)b ) α 1 α. This bound is independent of the initial value: and now η = a α+1 (α + 1) α 2α ( α 1 b ) α 1 α s γ T e γ T t < 1 (21) Fig.2 s = γ T, γ T.t = 0.5, a = 4, b = 1 L 0 = 1, L = 100, r = 0.1
A generalized Solow non-autonomous model 8973 Fig.3 α = 2, s = γ T, γ T.t = 0.5, a = 4, b = 1 L 0 = 1, L = 100, r = 0.1 In Fig. 2 the time evolution of the effective capital for α = 0.5 and α = 2 are compared. For α = 2, one clearly sees that the solution is unstable (cf. Fig. 3), which is consistent with the bound (21): 3.15 1. From an economic point of view, the instabilities may be responsible for a crisis in long-term growth. Fig. 2 and Fig. 3 show two distinct possibilities: sustained growth or collapse without growth. It is shown in the figures that these facts are associated with the existence, or not, of asymptotically converging states. If there exists an stationary states k, of the condition (5) and equation (20), it follows that se γ T t (k ) α a 1 + b (k ) α (ε + γ T ) k = 0 (22) To illustrate this, take α = 2. From relation (22) a steady states exists if a 2 s e γ T t b γ T 1 t 1 ( a ln γ T 2 b s γ T ) (23) For the set of parameters of Fig. 4 and Fig. 5 there are asymptotic equilibrium states if t 10 ln 2 6.9315. These values are consistent with those observed in these figures.
8974 Corina G. Averbuj and Juan José M. Martínez Fig.4 α = 2, s = γ T +ε, γ T.k 0 = 1.0, a = 4, b = 1 L 0 = 1, L = 100, r = 0.1 Furthermore, Fig. 5 shows that although the solution is not asymptotically stationary itself, it remains bounded. References Fig.5 α = 2, s = γ T, γ T.k 0 = 1.0, a = 4, b = 1 L 0 = 1, L = 100, r = 0.1 [1] E. Accinelli, and J.G. Brida, Population growth and the Solow-Swan model, International Journal of Ecological Economics & Statistics 8 S07 (2007) 54-63.
A generalized Solow non-autonomous model 8975 [2] D. Acemoglu, Introduction to Modern Economic Growth, Princeton University Press, Princeton, 2009. [3] P. Asea and P. Zak, Time to build and cycles, Journal of Economic Dynamic and Control 23 (1999) 1155-1175. http://dx.doi.org/10.1016/s0165-1889(98)00052-9 [4] R.J. Barro and X. Sala-i-Martin, Economic growth, MITT Press, Cambridge, 2003. [5] A. Bellen, N. Guglielmi and M. Zennaro, Numerical stability of nonlinear delay differential equations of neutral type, Journal of Computational and Applied Mathematics 125 (2000) 251-263. http://dx.doi.org/10.1016/s0377-0427(00)00471-4 [6] J.G. Brida and E.J. Limas Maldonado, Closed Form Solutions to a Generalization of the Solow Growth Model, Applied Mathematical Sciences, 1 (40) (2007) 1991-2000. http://www.m-hikari.com/ams/ams-password- 2007/ams-password37-40-2007/bridaAMS37-40-2007.pdf [7] S. A. Campbell, Introduction of delay differential equations, University of Waterloo, Canada, 2007. [8] F. Collard, O. Licandro, and L. Puch, The short-run dynamics of optimal growth models with delays, Economics Working Papers ECO2004-04, European University Institute, 2004. http://cadmus.eui.eu/bitstream/handle/1814/1888/eco2004-6.pdf [9] M. Ferrara and L. Guerrini, The Ramsey model with logistic population growth and Benthamite felicity function revisited, WSEAS Transactions on Mathematics, 8 (2009) 41-50. http://www.wseas.us/elibrary/transactions/mathematics/2009/29-200.pdf. [10] L. Guerrini, The Solow-Swan Model with Kaldor-Pasinetti Saving and Time Delay, Applied Mathematical Sciences, 6 (72) (2012) 3569-3573. http://www.m-hikari.com/ams/ams-2012/ams-69-72- 2012/guerriniAMS69-72-2012-2.pdf [11] J.K. Hale and S.M. Verduyn Lunel, Introduction to functional differential equations, Springer, New York, 1993. http://dx.doi.org/10.1007/978-1- 4612-4342-7 [12] H. Hu Peng, Analytical and numerical stability of nonlinear neutral delay integro-differential equations, Journal of the Franklin Institute 348 (2011) 1082 1100. http://www.wseas.us/elibrary/transactions/mathematics/2009/29-200.pdf
8976 Corina G. Averbuj and Juan José M. Martínez [13] V.B. Kolmanovskii and V.R. Nosov, Stability of functional differential equations, Mathematics in Science and Engineering, vol. 90, Academic Press, London, 1986. [14] N. Kaldor, A model of economic growth, Economic Journal 67 (1957) 591-624. http://dx.doi.org/10.2307/2227704 [15] L. Pasinetti, Rate of profit and income distribution in relation to the rate of economic growth, Review of Economic Studies 29 (1962) 267-279. http://dx.doi.org/10.2307/2296303 [16] H. Smith, An introduction to delay differential equations with sciences applications to the life, Texts in Applied Mathematics, vol. 57, Springer, New York, 2010. [17] R.M. Solow, A contribution to the theory of economic growth, Quarterly Journal of Economics, 70 (1956) 65-94. http://dx.doi.org/10.2307/1884513 [18] T.W. Swan, Economic growth and capital accumulation, Economic Record 32 (1956) 334-361. http://dx.doi.org/10.1111/j.1475-4932.1956.tb00434.x [19] H. Usawa, On a Two-Sector Model of Economic Growth, The Review of Economic Studies 29 (1) (1961) 40-47. http://dx.doi.org/10.2307/2296180 [20] H. Yuan, Boundedness and global convergence of non-autonomous neural networks with variable delays, Nonlinear Analysis: RealWorld Applications 10 (2009) 2195 2206. http://dx.doi.org/10.1016/j.nonrwa.2008.04.004 Received: October 15, 2014; Published: December 18, 2014