The Dynamics of Exploitation and Class in Accumulation Economies∗ Jonathan F. Cogliano,† Roberto Veneziani,‡ and Naoki Yoshihara§ December 10, 2014 Abstract This paper analyses the equilibrium dynamics of exploitation and classingeneralaccumulationeconomieswithpopulationgrowth,tech- nical change, and bargaining by adopting a novel computational ap- proach. First, the determinants of the emergence and persistence of exploitation and class are investigated, and the role of labour-saving technical change and, even more importantly, power is highlighted. Second, it is shown that the concept of exploitation provides the foun- dations for a logically coherent and empirically relevant analysis of inequalities and class relations in advanced capitalist economies. An index that identifies the exploitation level, or intensity of each indi- vidual can be defined and its empirical distribution studied using the standard tools developed in the theory of inequality measurement. JEL: B51 (Socialist, Marxian, Sraffian); D63 (Equity, Justice, Inequality, and Other Normative Criteria and Measurement); C63 (Computational Techniques, Simulation Modeling). Keywords: Exploitation, class, accumulation, simulation. ∗We would like to thank Marco Mariotti for his thoughtful comments and suggestions. The usual disclaimer applies. †(Corresponding author) Department of Economics, Dickinson College, Althouse 112, P.O. Box 1773 Carlisle, PA 17013, U.S. ([email protected]) ‡SchoolofEconomicsandFinance,QueenMaryUniversityofLondon,MileEndRoad, London E1 4NS, UK. ([email protected]) §The Institute of Economic Research, Hitotsubashi University, Naka 2-1,Kunitachi, Tokyo 186-0004, Japan. ([email protected]) 1 1 Introduction One of the key tenets of Marxian economics is the idea that exploitation and class are defining features of capitalist economies. This raises two issues. First, the existence of a logically coherent and empirically meaningful defi- nition of exploitation and class. Second, the economic mechanisms that lead to the emergence and persistence of exploitation and class. The first issue has received a lot of attention in the literature, leading to vast debates on the notions of exploitation and class. The received view is thatnologicallyconsistentandempiricallymeaningfuldefinitionexistswhich captures the key positive and normative insights of the Marxian theory of ex- ploitation and class in general economies. This view has been questioned in a seriesofrecentpapersbyYoshiharaandVeneziani(e.g. [26,27,19]). Inthese contributions, the concept of exploitation is analysed using a novel, general axiomatic approach which allows one to rigorously capture the normative and positive foundations of exploitation theory. Contrary to the received wisdom, it is shown that there exists a nonempty class of logically consis- tent definitions - conceptually related to the so-called “New Interpretation” (Dum´enil [5]; Foley [8]; Dum´enil, Foley, and L´evy [6]; see also Mohun [11]) - that satisfy a set of desirable properties in general convex economies with heterogeneous agents - including the existence of a robust relation between profits and exploitation, as well as between class and exploitation. These contributions, however, focus on one-period economies with no savings and accumulation, and have relatively little to say about the second issue in the opening paragraph, namely the dynamics of class, exploitation, and profits. This is not a minor issue. In his seminal theory, Roemer [14] has proved that exploitation and classes emerge as the equilibrium outcome of differential ownership of the means of production in competitive economies with optimising agents when capital is scarce, leading him to conclude that the normative relevance of exploitation reduces to an exclusive emphasis on asset inequalities. YetRoemer’sresultshavebeenderivedinone-periodmodelswhereascap- italism, according to Marx, is an inherently dynamic system geared towards capitalaccumulationandsoonemayarguethatclassandexploitationshould be analysed in a dynamic framework. In dynamic accumulation economies, it is not difficult to show that capital may become abundant, leading profits and exploitation to disappear (Devine and Dymski [4]). Perhaps more sur- prisingly, Veneziani [17, 18] and Veneziani and Yoshihara [20] have proved that if savings are allowed in a dynamic capitalist economy, then asset in- equalities are necessary for exploitation to emerge, but alone they are not 2 sufficient for it to persist even if agents do not accumulate in equilibrium. These results cast doubts on the claim that asset inequalities are necessary and sufficient for the emergence and the persistence of exploitative relations, and raise the issue of the determinants of exploitation and class. This paper adopts the conceptual approach to exploitation proposed by Yoshihara and Veneziani ([26, 27, 19]) in order to study the dynamics of asset inequalities, exploitation and classes. We significantly generalise Roe- mer’s [13, 14] accumulation economies with maximising agents in order to incorporate nonstationary prices, population growth, time-varying consump- tion norms, technical change, and distributive conflict. We analyse - both formally and computationally - the dynamic equilibrium trajectories of the economies and their class and exploitation structures, and generalise some fundamentalinsightsoriginallyprovedbyRoemer[14],includingtheso-called Class-Exploitation Correspondence Principle (henceforth, CECP). To be specific, we consider three main models exploring different mecha- nisms determining the emergence and persistence of exploitation and class. We start off by analysing a basic economy with constant consumption, pop- ulation, and technology: this benchmark scenario confirms the insights of the previous literature by showing that accumulation leads exploitation to disappear because the economy eventually becomes labour constrained. We then extend the model to consider capitalists’ decisions to innovate and norm-based consumption dynamics. Empirically, the long-run evolution of capitalist economies has indeed been characterised by an increase in (av- erage) consumption opportunities and by a tendential expansion of technical knowledge, leading to a progressive increase in labour productivity (Flaschel et al. [7]). Theoretically, our analysis confirms that labour-saving technical progress may play a crucial role in making exploitation persistent by guar- anteeing the persistent abundance of labour (Skillman [15]). In other words, the capitalists’ control of investment and innovation decisions can make ex- ploitation persistent by maintaining labour unemployment over time. Although many actual capitalist economies have indeed gone through long spells of labour unemployment, one may argue with Roemer [14] that a general theory of exploitation and classes should not crucially depend - either positively or normatively - on structural imbalances in factor markets. Exploitation and class are characteristics of capitalist relations of production and full employment does not make capitalist economies non-exploitative. Thereforeweanalyseanextensionofthebasicmodelwithpopulationgrowth, technical change and accumulation in which full employment occurs in every period and distribution is determined using a general Nash bargaining pro- cedure with the bargaining power of each agent endogenously determined as 3 a function of their ownership of the means of production and class solidarity. The results are quite striking: technical change and population growth are notsufficienttomakeexploitationpersistentunlesscapitalistsaresufficiently powerful and class solidarity among propertyless workers is sufficiently weak even if the economy never becomes labour constrained. Capitalist power is an essential determinant of the persistence of exploitation and class. In all economies, we analyse the evolution of the structure of exploita- tive relations. By deriving a robust correspondence between class and ex- ploitation status, the CECP yields relevant normative insights on capitalist economies. Yet, the CECP draws a rather partial, coarse picture of the structure of exploitative relations: two economies with similar numbers of agents belonging to each class and each exploitation category may still be very different. Based on [26, 27, 19], we propose a novel index of the level, or intensity of exploitation for individual agents, whose distribution provides a finer and more comprehensive picture of exploitative relations. The anal- ysis of its distribution yields relevant normative insights, and it raises some interesting issues that are conceptually analogous to those discussed in the literature on the measurement of income inequality. Another contribution of the paper is methodological. Given the com- plexity of the economies considered, the paper adopts a novel computational approach to Marxian exploitation theory. Pioneering work applying compu- tational methods to Marxian theory includes Wright [23, 24, 25], Cogliano [2], and Cogliano and Jiang [3]. But the latter contributions focus on Marx- ian price and value theory and the circuit of capital rather than exploitation and class. More related to our work is an unpublished paper by Takamasu [16], which adopts a computational approach to study class formation in ac- cumulation economies. Yet this paper does not analyse exploitation and it onlyconsidersaverybasicscenariowithconstanttechnology, populationand consumption. Moreover, there is no explicit analysis of agents’ maximising decisions or of the equilibrium conditions. By moving beyond the straightjacket of analytical solutions, a computa- tional approach allows us to study the equilibrium determination of exploita- tion status and the Class-Exploitation Correspondence Principle, and trace the co-evolution of exploitation and class over time in complex economies withendogenoustechnicalchange, populationgrowth, norm–basedconsump- tion dynamics, and generalised N-agent bargaining. The results obtained are robust with respect to changes in the specification of technology, population, preferencesand, especially, endowments, butalsotoalternativespecifications of some of the behavioural assumptions. The rest of the paper is structured as follows. The general framework 4 is described in section 2. Section 3 analyses the benchmark economy with stationary technology, population and consumption. Section 4 derives some key theoretical results concerning class and exploitation in the basic model. Section 5 presents the index capturing the level of exploitation of each agent. Section 6 analyses the dynamics of the basic model, and its exploitation and class structures, using computational techniques. Section 7 extends the analysis to economies with endogenous consumption and technical change, whereas section 8 focuses on the role of bargaining and power in economies with full employment, technical change and population growth. Section 9 discusses the robustness of the results. Section 10 concludes. 2 The framework Consider a dynamic extension of Roemer’s [14] accumulating economy with a labour market and only one good produced and consumed.1 In every period t = 1,2,..., let N denote the set of agents with cardinality N and generic t t element ν. At the beginning of each production period t, there is a finite set, P , of Leontief production techniques (A ,L ), where 0 < A < 1 and L > 0, t t t t t and all agents have access to the techniques in P . t Ineveryperiodt, agentshaveidenticalpreferencesbutpossesspotentially different endowments of labour, lν , and capital, ων , inherited from previ- t−1 t−1 ous periods. The distribution of endowments at the beginning of t is given by Π = (cid:0)lν (cid:1) ∈ RN and Ω = (cid:0)ων (cid:1) ∈ RN . In every t, each agentt−ν1∈ N t−is1thν∈erNetfore c+o+mpletelyt−id1entifietd−1byν∈aNdtuplet+(cid:0)lν ,ων (cid:1) ∈ R2 . t t−1 t−1 + (cid:0) (cid:1) An agent ν ∈ N endowed with lν ,ων can engage in three types of pro- t t−1 t−1 duction activity: she can sell her labour power zν; she can hire others to t operate a technique (A ,L ) ∈ P at the level yν; or she can work on her own t t t t to operate (A ,L ) ∈ P at the level xν. t t t t Following Roemer [13, 14], we assume that production takes time and current choices are constrained by past events. To be precise, wages are paid expostandw ∈ R denotesthenominalwagerateattheendoft, butevery t + agent must be able to lay out in advance the operating costs for the activities she chooses to operate using her wealth Wν . Letting p ∈ R denote the t−1 t + price of the produced commodity at the end of t and beginning of t+1, the market value of agent ν’s endowment - her wealth - is Wν = p ων . The t−1 t−1 t−1 1Given our focus on the dynamics of exploitation and class, the one-good assumption yields no loss of generality. The model can be extended to include n commodities, albeit at the cost of a significant increase in technicalities and computational intensity. 5 wealth that is not used for production activities can be invested to purchase goods to sell at the end of the period, δν. t Our main behavioural assumption postulates that agents wish to accu- mulate as much as possible, subject to consuming b ∈ R per unit of t ++ labour performed, Λν ≡ Lxν + zν. Within every period t, we consider b t t t t as a constant parameter, but we do allow for the possibility that b changes t endogenously over time. This modelling choice is motivated by our focus on the dynamics of ex- ploitation and class in capitalist economies characterised by a drive to accu- mulate, rather than on consumer choices. Theoretically, it is also consistent with the classical-Marxian tradition where consumption is largely the prod- uct of social norms, rather than utility-maximising behaviour, and it allows us to analyse the issue of the persistence of class and exploitation abstract- ing from heterogeneous individual consumption behaviour. Unlike in many accumulation models in the Marxian tradition, however, the introduction of an (endogenously determined) subsistence bundle raises some interesting theoretical and technical issues, as it imposes a relevant and oft-neglected constraint on the set of equilibria. 3 The basic model In this section, we set up and analyse the basic model, which is characterised by constant population, technology, preferences, consumption norms, and labour endowments over time. The focus on the basic model is motivated by analytical clarity and because it provides a theoretical benchmark and start- ing point for our analysis. However, the framework, concepts, and definitions presented in this section, and in the next one, can be easily extended and the results derived continue to hold in more general economies (as confirmed also by the simulations). (cid:0) (cid:1) Let N = N, P = P = {(A,L)}, b = b, and lν = (lν) for t t t t−1 ν∈N ν∈N all t, and suppose that the economy is sufficiently productive to produce a surplus: 1 − vb > 0, where v = L(1 − A)−1 denotes the embodied labour value.2 In every t, given (p ,w ), every agent ν ∈ N chooses xν, yν, zν, t t t t t and δν to maximise her wealth subject to purchasing b per unit of labour t performed (1) and to the constraints set by her capital (2) and labour (3) 2The condition 1−vb > 0 is equivalent to (1−bL) > A: it implies that if Ax units of capital are invested in the production process, (1−bL)x>Ax units of output (net of necessary consumption) are produced. 6 endowments. Formally, every ν solves the following programme MPν: t max p ων t t (xν;yν;zν;δν) t t t t subject to p xν +[p −w L]yν +w zν +p δν = p bΛν +p ων (1) t t t t t t t t t t t t t p Axν +p Ayν +p δν = p ων , (2) t−1 t t−1 t t−1 t t−1 t−1 Lxν +zν (cid:53) lν, (3) t t xν,yν,zν,δν,ων (cid:61) 0. (4) t t t t t Let Aν (p ,w ) be the set of actions ξν ≡ (xν; yν; zν;δν) that solve MPν t t t t t t t t (cid:0) (cid:1) at prices (p ,w ). Let Vν p ων ;(p ,w ) ≡ maxp ων be the value of t t t t−1 t−1 t t t t MPν realised by the actions in Aν (p ,w ). Let (p,w) ≡ {(p ,w )} and t t t t t t=1,... let (xν; yν; zν;δν) ≡ ξν = {ξν} . A basic accumulation economy is t t=1,... defined by the set of agents, N, technology, (A,L), consumption bundle, b, labour endowments, Π, and initial capital endowments, Ω ; and is denoted as 0 E(N;(A,L);b;Π,Ω ), or, as a shorthand notation, E . Let x ≡ (cid:80) xν, 0 0 t ν∈N t andlikewisefory , z , δ , ω , Λ , andl. BasedonRoemer[14], theequilibrium t t t t t notion can be defined. Definition 1: A reproducible solution (RS) for E(N;(A,L);b;Π,Ω ) is a 0 vector (p,w) and associated actions (ξν) , such that at all t: ν∈N (a) ξν ∈ Aν (p ,w ), for all ν ∈ N (individual optimality); t t t (b) A(x +y )+δ (cid:53) ω (capital market); t t t t−1 (c) Ly = z (labour market); t t (d) (x +y )+δ (cid:61) bΛ +ω (goods market). t t t t t At a RS, in every period (a) all agents optimise; (b) aggregate capital is sufficient for production plans; (c) the labour market clears; (d) aggregate supply is sufficient for consumption and accumulation plans. The economy E(N;(A,L);b;Π,Ω ) can thus be interpreted either as a sequence of gener- 0 ations living for one period or as an infinitely-lived economy analysed in a sequence of temporary equilibria. For any (p,w), the profit rate at t is π = pt−pt−1A−wtL. Given the struc- t pt−1A ture of the economy, we shall focus on equilibria with strictly positive prices.3 3It immediately follows from MPν that if there is some t(cid:48) such that p = 0, then at t t(cid:48) any RS it must be p =0 for all t>t(cid:48). t 7 Hence the profit rate is well defined at all t, and we can take the produced commodity as the num´eraire, setting p = 1, all t. Let the normalised price t vector be denoted as (1,w), where 1 = (1,1,...) and, at any t, w is the real (cid:98) (cid:98)t wage rate. It is immediate to prove that at any RS, if ω > 0, then w (cid:61) b t−1 (cid:98)t and π (cid:61) 0, all t. t Given the previous observations, by constraints (1)-(2), it follows that at any RS, for all ν ∈ N and all t, the following equation must hold ων = [1−A−w L](xν +yν)+(w −b)(Lxν +zν)+ων . (5) t (cid:98)t t t (cid:98)t t t t−1 Equation (5) has a number of implications. Lemma 1 proves that if the profit rate is strictly positive, then all wealth is used productively and if the wage rate is above subsistence, then the labour constraint (3) binds, for all agents at the solution to MPν.4 t Lemma 1: Let(1,w)beaRSforE . Atanyt: ifπ > 0, thenA(xν +yν) = (cid:98) 0 t t t ων , all ν ∈ N; and if w > b, then Lxν +zν = lν, all ν ∈ N. t−1 (cid:98)t t t Proof: By equation (5), if π > 0, but A(xν +yν) < ων , some ν ∈ N, t t t t−1 then ν can increase yν and capital accumulation, contradicting optimality. t Similarly, if w > b and Lxν +zν < lν, ν ∈ N, then ν can increase zν and (cid:98)t t t t capital accumulation, contradicting optimality.(cid:4) Next, it is possible to derive an explicit expression for the value of MPν t and for the growth rate of capital, gν, for all agents. t (cid:0) (cid:1) Lemma2: Let(1,w)beaRSforE . ThenVν ων ;(1,w ) = (1+π )ων + (cid:98) 0 t t−1 (cid:98)t t t−1 (w −b)lν, and gν = π +(w −b) lν , for all ν ∈ N. (cid:98)t t t (cid:98)t ωtν−1 Proof: Straightforward from equation (5).(cid:4) Lemma 2 has some interesting implications concerning the dynamics of accumulation. Let πmax ≡ 1−A−bL. Firstly, at all t, the aggregate growth A rate of the economy is g = π +(w −b) l . Hence, if l = LA−1ω , then t t (cid:98)t ωt−1 t−1 g = πmax, and if w = b, then gν = g = πmax, for all ν ∈ N such that t (cid:98)t t t ων > 0. Secondly, if w > b, then for any ν,µ ∈ N, gν > gµ if and only t−1 (cid:98)t t t if lν > lµ . Finally, if π = 0 then gν = (1−vb) lν , for all ν ∈ N such ωtν−1 ωtµ−1 t t v ωtν−1 that ων > 0, and g = (1−vb) l . Therefore, if there exists t(cid:48) (cid:61) 1 such that t−1 t v ωt−1 4To be precise, a RS should be denoted as (cid:0)(1,w),(xν; yν; zν;δν) (cid:1). In what (cid:98) ν∈N follows, we simply write (1,w) for the sake of notational simplicity. (cid:98) 8 π = 0 for all t (cid:61) t(cid:48), then the growth rate of the basic economy decreases t over time and tends asymptotically to zero. Lemma 3 derives a useful property of the set of solutions of MPν. t Lemma 3: Let (1,w) be a given price vector such that π (cid:61) 0 and w (cid:61) (cid:98) t (cid:98)t b, some t. If ξν solves MPν, then ξ(cid:48)ν ∈ R4 also solves MPν whenever t t t + t x(cid:48)ν +y(cid:48)ν = xν +yν and z(cid:48)ν −Ly(cid:48)ν = zν −Lyν. t t t t t t t t Proof: It is easy to check that ξ(cid:48)ν satisfies constraints (1)-(2). Moreover, t labourperformedisthesameinξν andξ(cid:48)ν, sinceL(x(cid:48)ν +y(cid:48)ν)+(z(cid:48)ν −Ly(cid:48)ν) = t t t t t t L(xν +yν)+(zν −Lyν). Then the result follows from equation (5).(cid:4) t t t t Lemma 3 implies that if (xν; yν; zν;δν) solves MPν, then there is an- t t t t t other vector (0; y(cid:48)ν; z(cid:48)ν;δν) which solves MPν. In the simulations, this t t t t allows us to select one of the many potential solutions of MPν by setting t xν = 0 for all ν ∈ N. t Theorem 1 characterises the equilibria of the economy. (cid:0) (cid:1) Theorem 1: Let (1,w),(ξν) be a RS for E . At any t: (cid:98) ν∈N 0 (i) If π > 0 and w > b, then l = LA−1ω . Furthermore, for any (1,w(cid:48)) t (cid:98)t t−1 (cid:98)t such that π(cid:48) (cid:61) 0 and w(cid:48) (cid:61) b, (ξν) also satisfies conditions (a)-(d) of t (cid:98)t t ν∈N Definition 1; (ii) If l > LA−1ω > 0 then w = b; t−1 (cid:98)t (iii) If l < LA−1ω then π = 0. t−1 t Proof: Part (i). By Lemma 1, A(xν +yν) = ων and Lxν +zν = lν, for all t t t−1 t t ν ∈ N. Therefore, A(x +y ) = ω and, by Definition 1(c), L(x +y ) = t t t−1 t t Lx +z = l. Since (x +y ) = A−1ω , we have L(x +y ) = LA−1ω = l. t t t t t−1 t t t−1 To prove the second part of the statement, take any (1,w(cid:48)) such that π(cid:48) (cid:61) 0 (cid:98)t t and w(cid:48) (cid:61) b. Then, it is immediate to show that ξν solves MPν at (1,w(cid:48)) for (cid:98)t t t (cid:98)t all ν and (ξν) satisfies conditions (b)-(d) of Definition 1 by assumption. t ν∈N Part (ii). At any RS, it must be w (cid:61) b. Suppose, contrary to the (cid:98)t statement, that w > b. Then, for all ν ∈ N, by (2), A(xν +yν) (cid:53) ων (cid:98)t t t t−1 and by Lemma 1, Lxν +zν = lν. But, since l > LA−1ω , Ly < z holds, t t t−1 t t contradicting Definition 1(c). Therefore w = b. (cid:98)t Part (iii). At any RS, it must be π (cid:61) 0. Suppose, contrary to the t statement, that π > 0. Then, for all ν ∈ N, by (3), Lxν + zν (cid:53) lν and t t t by Lemma 1, A(xν +yν) = ων . But, since l < LA−1ω , Ly > z holds, t t t−1 t−1 t t contradicting Definition 1(c). Therefore π = 0.(cid:4) t 9 Theorem 1 defines the theoretical framework for the analysis of the dy- namics of the economy and provides the foundations for the parameterisa- tions of the price vector (1,w). Although it only identifies necessary condi- (cid:98) tionsfortheexistenceofaRS,itdoesshedsomelightonhowtoconstructthe dynamic general equilibria of the economy. Consider part (ii) of the proof. Suppose l > LA−1ω , some t. If w = b, then π > 0 and labour performed t−1 (cid:98)t t does not produce any net income for accumulation, and for all ν ∈ N, any (0; yν; zν;0) with Ayν = ων solves MPν. Therefore since Ay = ω and t t t t−1 t t t−1 l > LA−1ω , we can choose a suitable profile (zν) such that Ly = z t−1 t ν∈N t t and all conditions of Definition 1 are satisfied at t. Consider part (iii) of the proof. Suppose l < LA−1ω , some t. If π = 0, t−1 t then w > b and capital holders are indifferent between using their wealth (cid:98)t productively and just carrying it for sale at the end of the period, and for all ν ∈ N, any (0; yν; zν;δν) with zν = lν solves MPν. Therefore since t t t t t z = l and l < LA−1ω , we can choose a suitable profile (yν) such that t t−1 t ν∈N Ly = z and all conditions of Definition 1 are satisfied at t. t t 4 Exploitation and Class in the Accumula- tion Economy The concept of exploitation in the accumulation economy can now be intro- duced. Inwhatfollows, exploitationstatusisdefinedineveryperiodt: thisis a natural assumption if the model describes a series of one-period economies, otherwise it reflects a focus on within period exploitation.5 Unlike in subsistence economies, focusing on the bundle consumed by an agent may be misleading as both poor and rich agents consume b per unit of labourexpended,buttheirpotential consumptionisverydifferent. Definition 2 identifies exploitation status in terms of the bundles of goods that an agent can purchase with her income. More precisely, for all ν ∈ N and all (p ,w ), t t (cid:0) (cid:1) let cν satisfy p cν = Vν Wν ;(p ,w ) +p bΛν −p ων . Then t t t t t−1 t t t t t t−1 Definition 2 [Roemer [14]]: Agent ν is exploited at t if and only if Λν > vcν; t t she is an exploiter if and only if Λν < vcν; and she is neither exploited nor t t an exploiter if and only if Λν = vcν. t t Theorem 2 characterises the exploitation status of every agent, based on Wν their wealth per unit of labour performed t−1: Λν t 5For a discussion of within period and whole life exploitation, see Veneziani [17, 18]. 10