EUROPEAN JOURNAL OF PURE AND APPLIED MATHEMATICS Vol. 7, No. 2, 2014, 156-165 ISSN 1307-5543 – www.ejpam.com Competition Among Manufacturers in Technological Innovation in the Market with Delayed Information Jair S. Dos Santos1,∗, Katia A. G. Azevedo1, Paola Torresan 2 1 Dep. de Computação e Matemática- FFCLRP-Universidade de São Paulo, Brazil 2 Faculdade de Economia e Administração de Ribeirão Preto Universidade de São Paulo, Brazil Abstract. The search for information in memory of technological competition process in the market is analysed using techniques widespread in biomathematics. Here we examined the effects that delayed information causes in choosing strategies process on the part of manufacturers to supplant a technology by introducing an alternative one. A differential-equation system with delays is presented to describe the dynamics of an endogenous model with memory. Situations of competition in which the market has several manufacturers using the same technologies and are in competition are analyzed. Conditions for stability and existence of periodic oscilations by means of Hopf bifurcation is investigated. 2010 Mathematics Subject Classifications: 34K15, 34C23, 92A15, 90A60, 90A16, 91B55 Key Words and Phrases: delay-differential equation, endogenous competition, stability, Hopf-bifurcations, innovation technology, delayed information. 1. Introduction The sharing of the benefits generated by technological knowledge involves a variety of flows which govern actions of manufacturers and consumers of technological innovation. A number of business-cycle models postulating that both, internal factors in market as demand instability, resource avaliability or external ones, as instability of government policies, setbacks in environmental regulation, influence strongly the global dynamics of all producing activities (see [1, 2, 9–11, 13]). Recently, technology planners and market agents have been interested in models that focus their attention on two causal variables, the one that captures the mea- sure of the potential quantity of consumers in an economic system susceptible to adopt the new technology and another that captures the measure of quantity of a product with new technology (see [1, 2, 7, 10]). In [10], the authors have adopted the transposition of techniques used in biomathematics to analyze the changes in migration of consumers facing new technologies (see also [3]). In ∗Corresponding author. Email addresses: jair@ffclrp.usp.br (J. dos Santos), pa.torresan@gmail.com (P. Torresan) http://www.ejpam.com 156 c© 2014 EJPAM All rights reserved. J. dos Santos, K. Azevedo and P. Torresan / Eur. J. Pure Appl. Math, 7 (2014), 156-165 157 their model they considered external and internal socioeconomic factors but, they have assume that each firm takes only into account instantaneous information about both, its own techno- logic output and the one of their competitors. They show clear-cut results that describe the dynamics of the competition among manufacturers producing innovation technology. Huang, in [7], investigates what is the impact that accurate information has on the features on the dynamics of an oligopolistic market and shows how an improvement in information accuracy (or elimination of delayed information) may destabilize an existing stable equilibrium. Our purpose is contributing to understand the basic mechanisms involved in the complexity of competition and co-existence of new technologies in innovation technology process. We will see later which techniques from dynamical systems, the memory of the production process can contain to explanation the fluctuations in indicators of innovation process technology and ensure reliable assessment of possible impacts upon the socioeconomic factors that affect both demand and supply through innovation (see [2, 14]). Form the view point of many researchers, the production process of the market depends endogenously on its past history and in several researches, on mathematical modelling taking into account this problem (see [2, 5–7, 9, 11, 13]). The authors in [5] construct two linear continuous-time dynamic oligopoly models with partial adjustment towards the best response and analyse the effects on local stability caused by lagged information. The authors, in [6], establish a class of economic models with two delays, one of them in production process and another in consumption. They discuss the stability of the equilibrium point for economic system and the existence of Hopf bifurcation. If the economy is in recession the government uses fiscal mechanisms or monetary policy and increases its expenditure to stimulate consumption and investment (pro-cyclical policy, see [12]). Takeuchi and Yamamura, in [11], investigate how the fiscal policy with a time delay affects stability in an economy. They assume that there exists a time delay between policy making to adjust the economy and its implementation which is divided into two factors: recognizing and decision making, which are lagged actions. Having to control the uncertainty effect about adoption of new technology, individual man- ufacturer attempt to delay decision making until they receive more accurate information to minimize uncertainty and to ensure improvement in performance. So, in their decisions tech- nology planners and market agents are always concerned about the existence of time delay between the moment when it is necessary to act and the moment of recognizing the necessity of action. We assume that they act consciously and that there are lags between the time that information are obtained and the time when decisions related to it are implemented. We propose a similar model of endogenous competition as presented by the authors, in [10]. Let x(t) be the potential quantity of consumers within an economic system at instant t. Let us denote by y(t) the quantity of a product with new technology put on the market at instant t. We assume that the evolution rate of the quantity of consumers in an economic system susceptible to adopt the new technology increases with the function g and decreases with the sum p0 and p1. The function p0 represents the functional response rate of manufac- turers with time delay incorporated. The function p1 represents the instantaneous functional response rate of manufacturers. The function s+ q(y) expresses the rate of specific extinction of the manufacturers due to intraspecific competition among them. The growth rate of the J. dos Santos, K. Azevedo and P. Torresan / Eur. J. Pure Appl. Math, 7 (2014), 156-165 158 product with new technology depends on past history with p0 and on the current moment at p1. ẋ(t) =x(t)g(x(t))− ym(t)[p0(x(t −σ) + p1(x(t))] ẏ(t) =y(t){−s− q(y) + γ[p0(x(t − r)) + p1(x(t))]y m−1(t)}. (1) where g, p0, p1, q ∈ C1([0,∞), R), σ > 0, r > 0 and m ≥ 1 is the constant of mutual interfer- ence. Assumption 1. Let be p(u) = p0(u) + p1(u). Assume that p(0) = 0, q(0) = 0, and p′(u) > 0, q′(u)> 0, for u≥ 0. Moreover, there is u0 > 0 such that g(u0) = 0, g(0)u> 0, limu→0(ug ′(u) + g(u))> 0, g ′(u)< 0 for u ∈ [0;∞). The same way [10], it follows from Assumption 1 that there are x0 > 0 and y0 > 0 so that P0 = (x0, y0) is the unique positive equilibrium point of (1), where y0 = � x0 g(x0) p(x0) � 1 m and p is defined in Assumption 1. Assumption 2. Let (x0, y0) be the positive equilibrium point of the system (1). Le us assume that a) x0 g(x0) = ym 0 p(x0) = y0(s+ q(y0)) = [(m− 1)γ]−1 y2 0 q′(y0), b) p0(x0) = p1(x0), τm = m m−1 y0q′(y0)< 1. c) 0< −p′0(x0)< p′1(x0) and x0 g ′(x0) + g(x0)− ym 0 p′1(x0)< 0 . System (1) becomes the system (26) in [10] if, 0 < m ≤ 1 and σ = r = 0. Stability of competition among new technologies available close to P0 was analysed in [10]. Since the model (1) depends on two delays, that leads to great complexity in the analysis of competition among new technologies available on the market (see [4]). We observe that in this system the delays can not be eliminated by any change of variable. Its dynamics are studied in terms of the local stability of P0 and of the Hopf bifurcation that is proven to exist as one of the delay crosses some critical value. To achieve our goals, we shall analyse how the roots of (3)) are distributed with respect to the imaginary axis. This is a classical problem that, in addition to being important by itself, plays an important role in the study of asymptotic behavior in the theory of delay differential equations (see [4]). If we set κ11 = x0 g ′(x0) + g(x0) − ym 0 p′1(x0), κ12 = −mym−1 0 p(x0), κ21 = γp′1(x0)ym 0 , κ22 = (m − 1)γym−1 0 p(x0) − y0q′(y0), b11 = −ym 0 p′0(x0), b21 = γym 0 p′0(x0), we obtain the system � ẋ(t) = κ11 x(t) + κ12 y(t) + b11 x(t −σ) ẏ(t) = κ21 x(t) + κ22 y(t) + b21 x(t − r) (2) that is the linearized system of (1), close to P0. From Assumption 2a it follows that κ22 = 0. If a = ym+1 0 q′(y0)(−p′0(x0)), b = ym 0 (−p′0(x0)), c = γy(m+1) 0 q′(y0)p′1(x0) and d = x0 g ′(x0) + g(x0)− ym 0 p′1(x0), then the characteristic equation of the system (2) is given by H(λ) = λ2 − dλ− bλe−λσ − ae−λr + c = 0. (3) J. dos Santos, K. Azevedo and P. Torresan / Eur. J. Pure Appl. Math, 7 (2014), 156-165 159 Let λ = x + i y be a solution of equation (3). Separating real and imaginary parts in (3) we obtain the following equations system for x and y � x2 − (y2 − c)− d x − be−σx[x cos yσ+ y sin yσ]− ae−r x cos y r = 0 2x y − d y + be−σx[x sin yσ− y cos yσ] + ae−r x sin y r = 0. (4) The solutions of the system (4) with null real part are solutions of the system � b y sin yσ+ a cos y r = −y2 + c −b y cos yσ+ a sin y r = d y. (5) Suppose y 6= 0 is a solution of (5), then we must have sin(σ− r)y = (−y2 + c)2 + d2 y2 − b2 y2 − a2 2ab y := %(y). (6) If we define u(y) = s + q(y), we indicate elasticity of u with respect to y at y0 by εq. Analogously, we have εg , εp0 , εp1 and εp are the elasticities of g, p0, p1 and p at x0, respectively (see Assumption 1). Remark 1. Authors in [10] consider the system (1) with σ = r = 0, 0 < m ≤ 1 and the corresponding to configuration of the equilibrium P0 is E3 (see [10, pp. 364]). They indicate the diagonal elements of the variational matrix associated to E3 by H = g(x0)[εg + 1− εp] and R = γ(m− 1)(s+ q(y0))[1− εq] ≤ 0. Using H + R and L = HR−mγ(y0)2m−1p(x0)p′(x0) they describe the stability of E3. We follow the alternative offered Bléair and Mackey, [2]which make it possible an endoge- nous explanation for erratic behaviour of competition with memory among new technologies available, when the innovation process operates around equilibrium point. The dynamics is described in terms of elasticities (see also [8]). We also show that the dynamics of the model (1) depends, essentially, on delays and on elasticities of the rate of evolution of the consumers, on the functional response rate of manufacturers and on the rate of specific extinction of man- ufacturers owing to the intraspecific competition among then manufacturers. From Assumption 2 it follows that b > 0, c > a > 0, d < 0, εg < 0, εp1 > 2(1+ εg), 0< −εp0 < εp1 , εp = εp0 +εp1 2 , εq = m− 1, (7) It also follows from Assumption 2 that, 2a = −εp0 τm g(x0), 2b = −εp0 g(x0), 2c = εp1 τm g(x0), 2d = g(x0)[2+ 2εg − εp1 ],moreover a = τm b and εp1 a = −εp0 c. (8) So, we can check, directly, that equation (3) is equivalent to equation λ2 − g(x0) 2 {(2+ 2εg − εp1 )λ+ (−εp0 )[λe−λσ +τme−λr]− εp1 τm}= 0. (9) J. dos Santos, K. Azevedo and P. Torresan / Eur. J. Pure Appl. Math, 7 (2014), 156-165 160 2. Stability For a moment assume that in (8) either 2+ 2εg − εp1 = 0 and εp1 < −εp0 or −εp0 τm g(x0) > 0 and εp1 τm g(x0) < 0, so it is not true that all roots of the equation (9) have negative real part, once on the real axis H(0) = τm[εp1 − (−εp0 )] < 0 and H(λ)→∞ as λ → ∞, so, there will be unbounded solutions of (1) and we do not uniform ultimate boundedness. As can been seen, from certain combination of the elasticities with time delay, erratic changes can arise around their fundamental values, which causes instability in the complex system of competition among the new technologies available. Remark 2. From 0< 17(−εp0 )< 2(3+ p 22 p 2− p 2)εp1 it follows that ρ0 =: �32(ε2 p1 − ε2 p0 )c 8ε2 p1 − 17ε2 p0 � 1 2 < �8(ε2 p1 + ε2 p0 )c ε2 p1 (3− p 2) � 1 2 =: ρ1. (10) For a sake of simplicity, let be ρ2 =: 8 p 2ε 3 2 p1 ( p 2+6)ε 3 2 p1+2 p εp1 + q 2(−εp0 ) Theorem 1. In addition to Assumption 1 and 2 we assume Æ g(x0)[2+ 2εg − εp1 ] [4( y∗ d ) 2 + 1]−1, ℜ � λ̇(σ∗, r∗, y∗)|x=0 � 6= 0 and (σ∗, r∗, y∗) is a Hopf bifurcation point and close to P0, the system (1) oscilates (see (8)). Proof. With a simple computation we verify that 2ab[%(y) + y%′(y)] = 4y2 + 2[d2 − (b2 − 2c)]y (see (6)). It follows from p 2(−εp0 ) + εp1 − 2εg − 2 = 0 that d = − p 2b (see the first item of the Proposition 1). Let be m% = ε2 p0 g(x0)−2τmεp1 τ1ε2 p0 g(x0) and n% = 2τm 4 q 3ε5 p1 (ε2 p1 −ε2 p0 )3 3ε2 p0 and 2(y∗)3 = n%τm(ε0 g(x0))2 4 . It follows from (13) that 0 ≤ m% ≤ 1 and 0 ≤ n% ≤ 1. We can choose 0 ≤ r∗ ≤ σ∗ satisfying 0 ≤ (σ∗ − r∗)y∗ ≤ σ∗ so that the straight line z = m% y + n% is tangent to the graphic of the function % in (6) at the point (y∗,%(y∗)). Since % is a convex function on interval (0,∞), then %(y) ≥ m% y + n% for all y ≥ 0. Analogously, we choose 0 ≤ r̄ ≤ σ∗ and ȳ satisfying 0≤ (σ∗− r̄) ȳ ≤ σ∗ so that the straight line z = msin y + nsin is tangent to the graphic of the function sine defined in (6) at the point ( ȳ , sin(σ∗ − r̄) ȳ). It is easy see that msin = msin(r̄, ȳ) = (σ∗ − r̄) cos(σ∗ − r̄) ȳ and nsin = nsin(r̄, ȳ) = sin(σ∗ − r̄) ȳ − (σ∗ − r̄) ȳ cos(σ∗ − r̄) ȳ . Since r∗ and y∗ were chosen so that %′(y∗) = m% and %(y∗) = n%, it follows from (5) and Implicit Function Theorem that the unique solution of the system � msin(r̄, ȳ) = m% nsin(r̄, ȳ) = n% (14) will be (r∗, y∗). REFERENCES 162 By using implicit derivative in (3) we can show that ℜ � λ̇(σ, r, y)|x=0 � = b y2[2y sinσy − d cosσy] + r ba cos(σ− r)y. (15) Because sin(σ∗ y∗) > [4( y∗ d ) 2 + 1]−1, we have ℜ � λ̇(σ∗, r∗, y∗)|x=0 � > 0 (d is defined in (8)). Remark 3. If the elasticities of the functions involved in model (1) satisfy the conditions of Theo- rem 2 we are able to localize a non-null purely imaginary root of equation (9) and to show that as the delay σ crosses the crtical value σ∗ there are two simple roots of equation (9) crossing transversely the imaginary axis from left to right, while all others have negative real part. We can also verify that (σ∗, r∗) determines a sequence {(σ∗p, r∗q )}(p,q)∈N2 of critical values that lie in a smooth manifold K defined by system (14). Using Hopf-bifurcation theory we show that the manifold K can be chosen in such way that each one of this critical values, near the equilibrium point P0, is associated to a nonconstant periodic solution of system (1). 4. Conclusions With endogenous framework, the stabilization of competition on a market that operates with several manufacturers using same technologies, is accurately analyzed. It is clear from Theorems 1 and 2 that production delays in commodities markets are potentially destabilizing factors. Theorem 1 shows that planners can search for information in the memory of an eco- nomic system without doing harm to parameter systems governing technological innovation process. Theorem 2 shows that if parameters are close to the boundary of the stability region the system undergoes Hopf bifurcation. With this phenomenon the technological innovation process becomes unstable and fluctuating. This lead us to believe that our analysis of the model (1) is able to offer technology planners a reasonable explanation for cyclical behaviour in competitive markets, and it suggests how market agents must act to avoided fluctuations. References [1] G. Barbirol and D. Ritelli. Dymamical systems in analysing competitiveness and co- existence among technologies. International Journal of Systems Science, 28(4):347–356, 1997. [2] J. Bléair and M. C. Mackey. Consumer memory and price fluctuations in commodity markets: an integrodifferential model. Journal of Dynamic of Differential Equations, 1(3):299–325, 1989. [3] H. L. Freedman. Stability analysis of a prey-predator system with mutual interference and density-dependent death rates. Bulletin of Mathematical Biology, 41(1):67–78, 1979. [4] J. K. Hale and S. M. V. Lunel. Introduction to Functional Differential Equations. Springer Verlag - New York - 1993, 1993. REFERENCES 163 [5] T. Howroyd and A. Russel. Cournot oligopoly models with time delays. Journal of Math- ematical Economics, 1(13):97–103, 1981. [6] C. Huang, C. Peng C., Chen, and F. Wen. Dynamics analysis of a class of delayed economic model. Abstract and Applied Analysis, 2013(Article ID 962738):12 pages, 2013. [7] W. Huang. Information lag and dynamic stability. Journal of Mathematical Economics, 44(1):513–529, 2013. [8] C. M. Mackey. Commodity price fluctuation: Price dependent delays and nonlinearities explanatory factors. 48(1):497–509, 1989. [9] A. Matsumoto and F. Szidarovszky. Nonlinear delay monopoly with bounded rationality. Chaos Solitons and Farctais, 45(4):507–519, 2012. [10] D. Ritelli, G. Barbirol, and P. Fabbri. Predation among technologies on market: A mod- elistic analysis. Journal of Mathematical Economics, 27(1):347–374, 1997. [11] Y. Takeuchi and T. Yamamura. Analysis of the kaldor model with time delay: monetary policy and government budget constraint. Nolinear Analysis Real World Applications, 5(1):277–308, 2004. [12] E. Wolfstetter. Fiscal policy and the classical growth cycle. Journal of Economics, 42(4):375–393, 1982. [13] T. Yamamura. Analysis of the Kaldor Model with Time Delay: Monetary Policy and Gov- ernment Budget Constraint. PhD thesis, Graduate School of Science and Engineering Shizuoka University, 2001. [14] K. Zhu and J. P. Weyant. Strategic decisions of new technology adoption under asym- metric information: a game-theoretic model. Decision Sciences, 34(4):643–675, 2003. Appendix In order to simplify notation we set x̄ = b+ p b2 + 2a 2 , ȳ = x̄ 2 + b 2 − d 2 + √ √ ( x̄ 2 + b 2 − d 2 )2 + a+ c. (A1) Unfortunately, the analysis is not easy, since it involves hard computations. Proposition 1. Assume (7) are satisfied and (i) d 0 so that if x̄ < ε0, all roots of the equation (3) have negative real part. Proof. We observe that a, b and c are positive, d is negative and c > a. We suppose that y = 0 in (4). The system reduces to x2 − (be−σx + d)x − ae−r x + c = 0 If ϑ(x) = x2 − (be−σx + d)x − ae−r x + c then ϑ(0) > 0. If x > 0, the first inequality in (i) gives ϑ(x) > x2 − (b+ d)x − a+ c > 0. From this it follows that p(λ) = 0 has no positive solution with null imaginary part (see (3)). Let consider the sets Si for i ∈ {0, 1,2, · · · , 6} given by S0 ={(x , y) ∈ R : 0< x ≤ x̄ and 0< ȳ ≤ y}; S1 ={(x , y) ∈ R2 : 0< 4σy ≤ π, and 2x > b}; S2 ={(x , y) ∈ R2 : π≤ 4σy ≤ 2π and 0< y ≤ x}; S3 ={(x , y) ∈ R2 : π≤ 2σy ≤ 2π and x > 0}; S4 ={(x , y) ∈ R2 : 0< y ≤ x and y ≥ x̄}; S5 ={(x , y) ∈ R2 : 0< x ≤ y and x ≥ x̄}; S6 ={(x , y) ∈ R2 : π≤ 4σy ≤ 2π and 0< x ≤ y}. (A2) For each (x , y) ∈ R2 +, the first equation in (4) becomes equivalent to Γ1(x , y) = 0, (A3) where Γ1(x , y) = x − y − d x − c x + y − be−xσζ1(x , y)− e−x r x + y a cos y r and ζ1(x , y) = (x + y)−1[x cos yσ+ y sin yσ]. We note that |ζ1(x , y)| ≤ 1 and −ae−r x x + y ≤ −ae−r x cos r y x + y ≤ ae−r x x + y ≤ ae−r x y . (A4) If ε∗(x , y) = −y2+(b+ x−d)y+a+ c, then yΓ1(x , y)< ε∗(x , y) since −d x < −d(x+ y). If (x , y) ∈ S0, we can verify that ε∗(x , y)< 0 and so, equation A3 has no solution belonging to S0. In fact, the function y(x) = (b+x−d)+ p (b+x−d)2+4(a+c) 2 is increasing in x and gives a positive solution for the second degree equation in y given by ε∗(x , y) = 0. The second equation in (4) is equivalent to Γ2(x , y) = 2x y − d y + be−σx[x sinσy − y cosσy] + ae−r x sin r y = 0. (A5) REFERENCES 165 Since Γ2(x , y)> y(2x− b) e 2x > b, there is no solution of equation A5 for (x , y) belongs to S1. It is easy to see that x sinσy − y cosσy ≥ 0 for (x , y) ∈ S2. Moreover, for all (x , y) ∈ S2, we have Γ2(x , y) ≥ 2x y − d y + ae−r x sin r y > 0. Thus, there is no solution of (A3) that belongs to S2. Let consider (x , y) ∈ S3, then Γ2(x , y)≥ 2x y−d y which is positive because d < 0. Hence, there is no solution of (A3) that belongs to S3. For each (x , y) ∈ S4, we define ζ2(x , y) = (x y)−1[x sin yσ − y cos yσ]. we can check directly that yζ2(x , y) ≥ −2 and y sin y r ≥ −x . Then, Γ2(x , y) > (x−1 y−3)[2y2 − 2b y − a], which is positive if y ≥ x̄ (see A1), whence (A5) has no solution at S4. Let (x , y) ∈ S5. Because xζ2(x , y)≥ −2 and (x y)−1 sin(y r)≥ −x−2, then Γ2(x , y)> (x−3 y−1)[2x2−2bx−a] that is positive if x ≥ x̄ (see (A1)) and so, has no solution into S5. We assume (x , y) ∈ S6, since x y ≥ x2 and −b y p 2 2 < −b ye−σx cosσy , σΓ2(x , y)> 2x2σ− π4 (d+ b p 2) is positive if d < −b p 2 (see (1)). Hence, there is no solution of (A3) that belongs to S6. Finally, the system (4) has no purely imaginary solution. If the roots of the system (4) with null real part exist, they must be roots of the system (5) We suppose y 6= 0 is a solution of (5), then we must have sin(σy − r y) = %(y) (see (6))). It is clear that %(y) = −%(−y), thus we will consider only y > 0. A direct calculation shows that 2ba y2%̇(y) = 3y4 + (d2 − b2 − 2c)y2 − (c2 − a2) and that y0 = � p [d2 − (2c + b2)]2 + 12(c2 − a2)− [d2 − (2c + b2)] 6 � 1 2 , (A6) is a global minimum point of % on the interval (0,∞). Since ab y2%̇(y) = ab y%(y) + y4 − (c2 − a2), then y4 0 = (−d2 y2 0 + b2 y2 0 +2c y2 0 −a2+ c2)/3 and 3ab y0%(y0) = 2(c2−a2)+(d2−2c− b2)y2 0 , the inequality of (3) and the first inequality in (1) implies that %(y0) > 1. Therefore, (3) has no purely real root on the right half complex plane. By general arguments on the compacity of the interval [0, π4 ] and the continuity of %, we arrive at the existence of a ε0 > 0 so that System (4) has no solution belonging to Sε0 = {(x , y) ∈ R2 : 0 ≤ x < ε0, 0 ≤ 4σy ≤ π}. If R2 + = {(x , y) ∈ R2, x > 0, y > 0}, the conditions (i), (ii), (iii) and x̄ < ε0 imply that R2 + ⊂ S ∪ Sε0 , where S = ⋃6 i=0Si (see (A2)). So, it can proven that System (4) has no solution that lies in Cl(R2 +) and the proof of the Theorem is complete.