ECONOMIC GROWTH, AN EVOLUTIONARY PROCESS THAT GIVES RISE TO AN ATTRACTOR Alain Villemeur1 Abstract Economic growth is seen here as the outcome of an entrepreneur-driven process of evolution in the context of an economy of competitive markets. In the course of this process the entrepreneurs implement capital and labour factors, one part of them committed to substitution and the other to complementarity with increasing returns. The theory demonstrates that the conditions of equilibrium of the different markets give rise to an attractor made up of steady states. The growth determinants for these states are employment, investment and technical productivity, with the profit share in income always being equal to 1/3. The comparison of what is learned with the empirical reality of the main developed economies demonstrates the interest of this view of growth. The attractors of the United States economy for the period 1960–2000 are given special attention. JEL classification : O30, O40, O57. Keywords : Attractor; Evolutionary process; Growth model; Market economy; United States. 1 Université Paris Dauphine, Crea , Place du Maréchal de Lattre de Tassigny 75775 Paris Cedex 16, France [email protected]. I should like to express my gratitude to Jean-Hervé Lorenzi, professor at the Université Paris Dauphine, for his invaluable advice and assistance with this research. 1 1. Introduction This article aims to explain the mechanism of economic growth with the help of a growth model that is assumed to represent it in the context of a free market economy. Growth is considered to be the result of economic development driven by innovation as initially described by Schumpeter (1934, 1939, 1942), then developed by Nelson and Winter (1982) and Gaffard (1997). Economic development has the characteristics of an evolutionary process with the entrepreneur as the central figure. This evolutionary process is by nature in perpetual disequilibrium. Could it be characterised by an attractor, as Gaffard (1997, p. 62) and Nelson (2005, p. 66) have assumed? To recap, a dynamic system attractor is a set towards which a system irreversibly evolves after a long enough time. Nelson (2005) justified this question in the following manner: “In their analysis of certain economic phenomena, for example technical advance, many economists recognize that frequent or continuing shocks, generated internally as externally, may make it hazardous to assume that the system ever will get to an equilibrium; thus the fixed or moving equilibrium in the theory must be understood as an “attractor” rather than a characteristic of where the system is”. This article also aims to explore this question. We retain the definition of an evolutionary process given by Schumpeter (1939, p. 86): “The changes in the economic process brought about by innovation, together with all their effects and the response to them by the economic system, we shall designate by the term Economic Evolution.” This process is driven by entrepreneurs who have perceived opportunities for wealth through innovations that have been implemented. Following the first evolutionary growth model (Nelson and Winter, 1982), we accept that a growth model must also be based on a model of economic evolution2. It strikes us that an evolutionary economic growth model of this kind raises two major problems. The first relates to the substitutable or complementary nature of capital and labour inputs applied by entrepreneurs, given the innovations brought into the new combinations. The second relates to the nature of the evolutionary process and the equilibrium encountered. These two problematics will be the topic of a discussion that will lead to retaining the founding hypotheses of a new model (Section 2). We will then draw up a new evolutionary model of economic growth (Section 3) in order to form, for a highly simplified free market economy, a representation of the essential mechanisms of evolution at the heart of growth. The conditions of market equilibrium (goods market, financial market, labour market) are detailed in Section 4. We show that these conditions generate the existence of an attractor formed by a set of steady states characterising long-term growth (Section 5). Some striking lessons come to light regarding distribution of income. The major theoretical lessons are then compared with the stylised facts (Kaldor, 1961, Barro and Sala-i-Martin, 1995) and the historical data on profit share in income are analyzed. The attractors are identified for the economy of the United States (1960-2000) in particular (Section 6). After creating a synthesis of the lessons learned from this contribution to the theory of growth, lines of progress are identified in order to overcome the limits of current modelling (Section 7). 2 A complete overview is in Nelson (2005) and in Verspagen (2006). 2 2. Founding hypotheses The evolutionary process is boosted by innovation, of which Schumpeter distinguishes five forms: the introduction of a new good, the introduction of a new method of production, the opening of a new market, the conquest of a new source of raw materials or half-manufactured goods, the carrying out of a new organisation of an industry, e.g., creation of a monopoly position (Schumpeter, 1934). What impacts do these innovations have on the economic system, particularly goods output and potential equilibrium? Schumpeter (1934, p. 121) offers an initial response by distinguishing, in his own words, “produce more” and “produce differently”. “Produce more” is symbolic of product innovations, while “produce differently” is symbolic of innovations in processes. We must go further, however, by specifying the links between capital and labour factors that may underlie “produce more” and “produce differently”. What role do substitution, complementarity or increasing returns have? This is the first question we examine. These innovations are behind the evolutionary process conducted by entrepreneurs, who are driven to make successive decisions regarding investment, production and labour. In these conditions, what kind of evolution and equilibrium are encountered? This is the second question we shall examine. 2.1. Substitution, complementarity, increasing returns? The first significant growth model in the theory of growth, the neo-classical model established by Solow (1956), is characterised by a production function (with constant returns to scale) whose two input factors, capital and labour, are substitutable. Economic growth then lies in exogenous sources, natural population growth or technical progress; in the absence of continual improvements in technology, the growth of the production per capita dies out. The Solow model, and those derived from it (Mankiw and al., 1992) proved interesting for explaining the conditional convergence of economies (Barro and Sala-i-Martin, 1995). Yet, technological progress notwithstanding, these models only explain a small part of effective growth and provide no satisfactory explanation for economic growth (Helpman, 2004). Another vision of the connection between capital and labour-that of increasing returns- has been set forth by numerous economists, in particular Young (1928) and Kaldor (1961, 1972).3 Input factors are seen as complementary, with increase in output simultaneously requiring an increase in capital and labour, a complementarity that is based on the principle of increasing returns to scale. Though empirical assessments of the growth phenomenon teach us about the primarily complementary nature of input factors (Gaffard, 1997),4 increasing returns are not sufficient to provide a subtle explanation for evolutions in the fundamentals of economies, as Salter notably demonstrates (1960). The work of Romer (1986) and Lucas (1988) brought about a third vision: endogenous growth models. These are generally based on the existence of increasing (or constant) returns as well as a broader concept of capital that encompasses human capital. All these models are a considerable contribution; we can in particular cite the model proposed by Aghion and Howitt (1988), which shows that innovation and capital accumulation can be two essential factors of 3 Young and Kaldor consider that Adam Smith was behind the first observation of increasing returns, due to his considerations on the effects of the division of labour (An Inquiry into the Nature and Causes of the Wealth of Nations). 4 See p. 172. 3 long-run growth. As Barro (1997) then notes, endogenous growth models are important for gaining a better understanding of the continuation of per capital GDP growth in developed economies. There have, however, been three major criticisms of the theory of endogenous growth. The first, made by Jones (1995), concerns the apparent discrepancy with reality in OECD countries where, since 1950, the number of scientists and engineers working in R&D has tripled in relation to the active population, without creating a corresponding increase in productivity growth.5 The second considers that human capital is not an input like others; this is in keeping with Nelson-Phelps’ approach (1966) whereby education leads to increasing capacity for innovation and adaptation to new technologies. Finally, these models do not succeed in explaining conditional convergence of economies (Aghion and Howitt, 1998).6 Even though this brief account shows that research on the true mechanisms and determinants of growth is still needed, several elements seem well established, in both theoretical and empirical terms. Empirical reality thus encourages us to acknowledge the simultaneous existence of complementary and substitutable inputs in the growth process. On a theoretical level, some economists, notably Jones and Manuelli (1990) or Howitt (2000), have developed the idea of reconciling increasing (or constant) returns with the convergence of economies linked to the Solow model. They have thus, as Barro and Sala-i-Martin point out, demonstrated the relevance of a production function considered to be the combination of an AK function and a Cobb-Douglas function, without, however, being able to give a justification for it. This dual form was also the subject of Lorenzi and Bourlès’ first works (1995) and led to Villemeur researching and elaborating a new growth model (2002, 2004) in order to help to explain situations of economic divergence. These various elements of reflection have led us to posit that growth is the result of an evolutionary process during which firms implement new productive combinations spurred by innovation, with capital and labour inputs being in part substitutable and in part complementary with increasing returns to scale.7 Capital8 is thus substituted for labour over vintage capital (“produce differently”), while capital can give rise to increasing returns through the implantation of extra machines (“produce more”) ; substitution and complementarity are the two facets of “creative destruction”. 2.2. Which evolution, which equilibrium? As Nelson and Winter (1982) highlighted, the evolutionary process is a process of continuing disequilibrium, taking the decisions made by entrepreneurs into consideration.9 Despite these imbalances, economic trajectories have certain regularities, as Kaldor (1960) first formalised. This invites reflection on the nature of the evolutionary process. Firstly, we consider that the evolutionary process is led by entrepreneurs making decisions, in production, investment and labour, for example, and as such can therefore seize upon or spark off innovations. Hanusch (1988, p. 2) sums up that vision: “the principal economic agent is 5 The model developed by Aghion and Howitt (1998) overcame this criticism. 6 The authors come to the conlusion that these models (the Solow-Swan and AK models), do not constitute the ultimate explanation of the growth process. 7 Nelson and Winter’s growth model (1982) does not take the dual aspect of production factors into account. 8 In accordance with Nelson-Phelps’ approach, it is accepted that human capital is not a production factor; the capital variable K will therefore not encompass human capital. 9 “The fact that there are winners and losers and that the process is one of continuing disequilibrium” (p.276). 4 the entrepreneur, who establishes new combinations by introducing new products as well as new production methods”. This vision of the entrepreneur’s role follows Schumpeter’s and cannot be replaced by the vision of an R&D sector that brings innovations (Ebner 2000). We retain the concept proposed by Gaffard (1997, p.88) of a sequential evolutionary process in production, investment and employment defined as “a succession of balances established over elementary (short) periods, when these balances do not constitute an intertemporal balance over a longer period of time.” Entrepreneurs then make decisions in relation to expectations formulated for the ensuing periods. The sequential method’s specific approach will be at the heart of the new model put into play. This method will be used to model the sequence of decisions made by the entrepreneur to create new combinations; it also enables use of analytical tools that make it transparent.10 We place ourselves in the framework of a free market economy whose competitive functioning is assumed, in the long run, to bring about conditions of equilibrium in the goods market, the financial market and the labour market. For example, the competitive goods market is thus assumed to oblige entrepreneurs to keep to minimum production costs (Schumpeter, 1942, p. 97). We assume that firms hence aim to reduce production costs (per investment unit); profit is then, by essence, “the result of carrying out new combinations” (Schumpeter, 1934, p. 136). The free market economy should also favour the regularity of growth (Amendola and Gaffard, 1998).11 Moreover, given the innovation dynamic, it is important to consider the existence of a co-evolution (Malerba, 2006) between technique, firms, innovation activities and the different markets. The attractor is defined as the set of states towards which fundamentals evolve in the absence of disequilibrium. It is determined by simultaneous realization of equilibrium conditions on the different markets: in other words, it is the result of the presumed perfect functioning of the markets. 3. The evolutionary model of economic growth 3.1. The sequential and dual evolutionary process in a free market economy Generally speaking, we consider that entrepreneurs ensure production of a technique A, using different vintages of machines.12 By installing additional production machines, they make complementary use of capital and labour to create extra output.13 By replacing, for an unchanged volume of output, old machines by new, more efficient ones that therefore require fewer staff, they are substituting capital for labour.14 10 The Nelson and Winter model (1982), based on Markovian processes and a simulation, has the drawback of not being analytical. 11 “The viability of the growth process, then and not stability is the relevant concept. This depends mainly on the functioning of prevailing institutions (markets, governments, and so on), that is, on the ability of these institutions to provide the means and the environment that make it possible for the economy to remain viable by continuously adapting to shocks and imbalances as they arise” (p. 6). 12 The idea of the vintage capital model was put forward by Johansen (1959) and was then the subject of numerous studies. 13 These machines can be new or old designs. 14 This hypothesis complies with the learning-by-doing model developed by Arrow (1962). Thanks to learning- by-doing, the machines that are built are increasingly efficient; this is materialised in a reduction in the number of workers needed to operate them. 5 In “produce more”, capital and labour are complementary, as returns are increasing. Production-employment elasticity will be assumed to be variable, given the extreme diversity of productive combinations that can be retained, calling on different levels of per capita production. In “produce differently”, capital is substituted for labour, with jobs being cut in relation to volumes of investment;15 here too various combinations can be retained. According to the terminology introduced by Phelps (1963), the technique is assumed to be a putty-putty model, given the perfect substitutability of the inputs, as in the Solow model. Complementarity is employed for new production equipment and the proportion of factors is not fixed, due to the existence of numerous possible technical combinations. For previously installed equipment, the different machine vintages make it possible to substitute capital for labour, again variably depending on the combinations used. The evolutionary process wherein an entrepreneur implements a new productive combination is sequential, with entrepreneurs being led to make successive investment, production and employment decisions while taking consumer demands into account and formulating expectations for the ensuing periods.16 During this process, remunerations are paid to employees and entrepreneurs (or shareholders). The orientation of productive combinations therefore depends on the price of factors, which places income distribution at the heart of firms’ investment decisions. Throughout this process, we take it that consumers reinforce these choices; in other words, production-demand balance is achieved throughout the sequence of decisions. For the sake of simplification, we accept that for each period the prices of goods are fixed.17 Given the competition in the goods market framework, entrepreneurs will aim for competitiveness through their investments throughout the sequence. One condition for equilibrium is that entrepreneurs adopt productive combinations that minimise the unit production cost (per investment unit) related to the investments made, while taking into account the return on capital expected18 by the financial market (Robinson, 1954). In the financial market framework, the condition for equilibrium imposes an expected return on capital equal to return on capital. As Robinson theorised (1954, pp. 130-131), if this condition is not met, the volume of invested capital is not adapted. We also accept that in the long-term, the return on capital is a constant through time, while entrepreneurs aim to keep the profit share in income constant; this amounts to the definition of neutrality for technical progress as per Harrod.19 The condition for equilibrium in the labour market framework is that entrepreneurs aim to implement productive combinations for which wage gains are independent of the employment 15 These latter considerations are coherent with the stylised fact (Pianta, 2006) which states that product innovation contributes to creating jobs, while process innovation contributes to cutting jobs. 16 A simple model of the investment and output sequences (Nelson and Winter, 1982) that corresponds to Schumpeter’s analysis is Y = A K , the output Y of a firm i at time t being equal to capital stock times the it it it it technical productivity it is employing (p.284). Note that this function is similar to the AK function proposed by Rebelo (1991). 17 This follows the concept of “temporary equilibrium” defined by Hicks (1939). 18 Return on capital is defined as the ratio between the amount of profit and the volume of capital employed. It is sometimes termed “rate of return on capital”. The return expected by the entrepreneur is assumed to be a function of the financial market’s requirements. 19 These final conditions are indeed equivalent to the conditions defining a neutrality for technical progress in the sense in which Harrod sees it (constancy of the profit share in income and of the capital productivity), as long as the prices of production and capital are constant, which they are in this model. 6 growth rate.20 With this last condition translating the existence of a wage increase norm that imposes itself on all firms, independently of employment growth. If this condition is not met, the employees of firms that create jobs will be favoured or put at a disadvantage. The evolutionary process of growth in a free market economy is mainly characterised by these major rules. Nevertheless, for reasons of clarity, other rules specifying it are introduced as the model is constructed. 3.2. Production activity In the considered economy, goods are production goods as much as consumer goods. We place ourselves at the period [t, t+dt] of a sequence of decisions for an entrepreneur engaged in a process of sequential evolution. This entrepreneur invests the volume K& to plan for additional output, K being the volume of capital. So x is the share of the increment of capital engaged in “produce more”; x will be named “complementarity rate” in the series. The increment of capital engaged in “produce more” is xK& (x∈R 0≤ x≤1). The output growth rate We accept that additional production, linked to a new productive combination using the technique A, takes the form AxK& .21 The parameter A characterises the productivity of the investment employed in complementarity: it will be termed “technical productivity”22. This parameter is supposed constant over a long period and common to all entrepreneurs. Since, by definition, substitution occurs in continuous production, additional production is written: Y& = AxK& A>0 0≤ x≤1 (1) We assume that production meets demand and that the balance between investment and saving is achieved. Hence: Y& K& I = Ax = Axi = Axs with K& = I =i = s Y =C+I (2) Y Y Y where I is investment (net), C is consumption, i is investment rate (net) and s is saving rate. These two latter rates are assumed equal and will from now on be called “rate of accumulation”. The output growth rate is proportional to technical productivity, the complementarity rate and the rate of accumulation. Thus we obtain a simple result: economic growth is self-maintained and all the stronger when the rate of accumulation or the proportion of investment employed in complementarity are high. The employment growth rate Let us now address the relation between employment L and investment. In the framework of complementarity, let us designate as L the creations of jobs associated to the investment c xK& . Accounting for innovations, we accept that the marginal rate of labour productivity 20 In his study of 27 industrial sectors of the American economy from 1923 to 1950, Salter (1960) noted the lack of correlation between labour productivity gains and employment growth. Other economists have also demonstrated this fact in the United States in a similar form: there is no correlation between labour productivity and employment; let us mention the works of Hansen and Wright (1992). 21 It comes, by differentiation in relation to time, from the form proposed by Nelson and Winter (1982) to represent Schumpeter’s analysis. 22 It should be noted that A is a marginal technical productivity 7 (Y&/L ) is higher than the labour productivity, the coefficient of proportionality being the c elasticity e that we will suppose is variable. Returns to scale are increasing and translated by c a coefficient higher than the unit, i.e. Y& Y L A A =e e >1 ⇒ c = xs =ε xs with ε = 0<ε < A (3) L c L c L e c c e c c c c where ε is defined as the “job creations coefficient”. c In the substitution framework, the rate of job losses (relation between job losses L and l employment L) is assumed to be proportional to the rate of investment employed in ( ) substitution 1−x i, with the coefficient of proportionality ε being named “job loss l coefficient”23 i.e ; L l =ε(1−x)s ε ≥ 0 L l l (4) Hence the employment growth rate that is the result of the balance between the job creations rate and the job loss rate: L& ( ) =ε xs−ε 1−x s ε >0 ε ≥ 0 A>ε (5) L c l c l c We assume that there is a connection between choices made with regard to complementarity and substitution; when the entrepreneur favours the choice of combinations that tend to create jobs (a higher job creations coefficient), the combinations chosen for substitution destroy fewer jobs (lower job loss coefficient). Consequentially, we set down the following relation: ε +ε =εmx 0<ε ≤εmx 0≤ε <εmx (6) c l c c c l c where εmx is the maximal coefficient of job creations.24 The constraints of labour organisation c impose limits to creating or cutting of jobs. This last parameter reflects the organisational limit to creating or cutting jobs, taking into account the organisations which the entrepreneur has implemented in the context of use of technique A.25 In the end, employment growth is finally written: L& ( ) =εmxxs− εmx −ε s 0<ε ≤εmx A>εmx (7) L c c c c c c Employment growth rate thus depends on the rate of accumulation, the complementarity rate, the job creations coefficient, and a parameter reflecting the organisational limit to creating or cutting jobs. 4. The conditions of equilibrium in the markets We will now successively determine the states that are the result of equilibrium in the different markets. 23 In a way this reflects the fact that the entrepreneur cuts jobs in proportion to the “deficit” of demand for his products, a deficit represented by the term A(1−x)s. The job creations coefficient and the job loss coefficient are the elasticities of labor to investment (related to complementarity and substitution). 24 It is also the maximal coefficient of job losses. 25 This is coherent with the stylised fact whereby division of labour is limited by the use of technique and not by the size of the market (Malerba, 2006). 8 4.1. Taking the goods market into account We are still considering an entrepreneur at a time t engaged in a sequential and dual evolutionary process in a free market economy. For an investment K& , the entrepreneur must arbitrate between numerous productive combinations. Given the competition, entrepreneurs will aim to minimise the cost of production (per investment unit) that results from the investment; he takes this decision by taking into account return on capital expectations z a and assuming that wage w is constant. The cost of production C induced by the investment decision includes the cost linked to m employment growth and the cost of invested capital linked to the entrepreneur’s profitability expectations. In the end, the cost of production (per investment unit) is written: . C wL& + z K& [ ] m = a = (1−c)εmxx+(1−c)ε −(1−c)εmx + z (8) K& K& c c c a ( ) c being the profit share in income such as wL = 1−c Y . In a context of limited rationality, the entrepreneur is supposed to minimise the cost per investment unit under two particular constraints: - the share of profit c in income is constant in order to maintain the share of profit when choosing new combinations; - the cost of job creations for output, per investment unit, is inversely proportional to the expected return on capital. This constraint reflects the fact that the greater the expected return on capital, the fewer jobs will be created, given the risks taken by the entrepreneurs who create jobs to produce goods. Hence the two constraints: wL wLεxs c c =c c = c =(1−c)xε = 2 z ≠ 0 x ≠ 0 1 K& K& c z a a In fact, the entrepreneur’s optimisation programme is the following: at each time t, the entrepreneur’s decisions are deduced from: ⎧C ⎫ Min⎨ m ⎬, subject to:26 c =c xε z =c' ⎩ K& ⎭ 1 c a 2 The minimisation programme is equivalent to: Min{(1−c )εmxx+(1−c )ε + z } subject to: c' = xεz 1 c 1 c a 2 c a To solve this easily, we obtain a function with 2 variables (by substitution of the constraint in the function to be minimised) for which we write the first-order conditions to find the extrema, i.e. c' f(x,ε ,z ) =(1−c )εmxx+(1−c )ε + 2 c a 1 c 1 c xε c ∂f c' ∂f c' = (1−c )εmx − 2 =0 =(1−c )− 2 = 0 ∂x 1 c x2ε ∂ε 1 xε2 c c c These conditions define a minimum,27 for it is easy to show that the Hessian matrix of order 2 is defined positive.28 In the end, the conditions of minimisation are written:29 26 In other words, the constraints are exogenous variables at x,ε ,z . c a 27 We have of course assumed that the variables are strictly positive. 9 z z x = a ε =εmxx = a (9) (1−c)εmx c c 1−c c At optimum competitiveness, the complementarity rate and the coefficient of job creation depend on the expected return on capital. This reflects the fact that the choice of productive combinations is oriented by the expected return on capital. The condition for a solution to exist in the realm of definition requires expected return on capital to be below a level that depends on the technique employed and the share of profits in income. (x,ε )∈]0,1]x]0,εmx] ⇔ za ∈]0,1]⇔ z ≤(1−c)εmx (10) c c (1−c)εmx a c c In the framework of a sequential and dual evolutionary process, the entrepreneur’s competitive choices thus determine, in relation to the expected return on capital for the ensuing periods, the share x of capital engaged in complementarity and the coefficient of job creation ε . Given the relations (7) and (9) : c L& ( ) 1 L& 1 =εmxxs− εmx −ε s ε =εmxx ⇒ x = + (11) L c c c c c 2εmxs L 2 c The output growth rate (2) becomes: Y& A L& A = + s (12) Y 2εmx L 2 c 4.2. Taking the financial market into account Up to now, return on capital expectations has been formulated throughout the sequential evolutionary process. In conditions of financial market equilibrium, expected return on capital meet return on capital to reach a value at equilibrium z , which is a time constant, and e entrepreneurs keep the profit share in income constant: Y z = z = z =constant c =constant with z = c (13) a e K This is translated by the following conditions: z Y Y& K& z z z = z = z ⇒ x = e =constant⇒ = = Axs = s e ⇒ x = e (14) a e (1−c)εmx K Y K c Ac c These conditions lead us to determine profit share in income and the complementarity rate: εmx z c = c x = e (15) εmx + A Ac c 4.3. Taking the labour market into account From now on we take it that the entrepreneur takes the existence of the labour market into account. The condition for equilibrium in this market is defined by a wage increase norm that is imposed on all firms, regardless of the employment growth rate. Given that entrepreneurs maintain a constant profit share in income (financial market), these two conditions are written: w& Y& L& w& L& c =constant ⇔ = − independent of w Y L w L 28 The two principal minor diagonals of the Hessian matrix are strictly positive. 29 These conditions are easily obtained by writing equals between the three terms of the sum to be minimised. 10