Financial Intermediation Versus Stock Markets in a Dynamic Intertemporal Model
Abstract
EconStor is a publication server for scholarly economic literature, provided as a non-commercial public service by the ZBW.
Full text
Rovelli, Riccardo; Fulghieri, Paolo; Bhattacharya, Sudipto Working Paper Financial Intermediation Versus Stock Markets in a Dynamic Intertemporal Model Quaderni - Working Paper DSE, No. 300 Provided in Cooperation with: University of Bologna, Department of Economics Suggested Citation: Rovelli, Riccardo; Fulghieri, Paolo; Bhattacharya, Sudipto (1997) : Financial Intermediation Versus Stock Markets in a Dynamic Intertemporal Model, Quaderni - Working Paper DSE, No. 300, Alma Mater Studiorum - Università di Bologna, Dipartimento di Scienze Economiche (DSE), Bologna, https://doi.org/10.6092/unibo/amsacta/5008 This Version is available at: https://hdl.handle.net/10419/159143 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-nc/3.0/
FINANCIAL INTERMEDIATION VERSUS STOCK MARKETS IN A DYNAMIC INTERTEMPORAL MODEL by Sudipto Bhattacharya1, Paolo Fulghieri 2 and Riccardo Rovelli 3 August 1997 Forthcoming: Journal of Institutional and Theoretical Economics 1 London School of Economics and Political Science, UK, and CEPR. 2 INSEAD, France, and CEPR. 3 University of Bologna, Italy.
ii ABSTRACT We consider the transitions among intragenerational and alternative intergenerational financing and liquidity risk-sharing mechanisms, in an Overlapping Generations model with endogenous levels of long-lived investments. The existence and characterization of a Self-Sustaining Mechanism, stable across generations, are established. The long-run equilibrium outcome, in a Proposal Game across generations, is shown to depend on the risk-aversion and propensity for early liquidity needs of the agents. JEL classification: D23, G10, G20. Acknowledgements: We are grateful to CEMFI (Madrid, Spain), CEPR (London, UK), ICER (Torino, Italy), IGIER (Milano, Italy) and the Studienzentrum Gerzensee (Switzerland) for research and financial support. Discussions with Patrick Bolton, John Boyd, Ed Green, Martin Hellwig, Jorge Padilla, Ailsa Roell, Bruce Smith and Elu von Thadden have been helpful. We are grateful for the comments of the discussants, Thorsten Hens and Christian Keuschnigg, at the International Seminar on the New Institutional Economics (Universität des Saarlandes, June 1997). Alessandro Secchi made several useful suggestions. Giacomo Elena and Wim Deblauwe provided excellent research assistance. All errors remain our responsibility. A previous version of this paper was circulated as “Turnpike Banking : Only the Meek Shall Inherit the Earth”.
iii I. INTRODUCTION AND SUMMARY Models of intertemporal liquidity risks, or "preference shocks" - developed in Bryant [1980] and further examined in the works of Diamond and Dybvig [1983], Bhattacharya and Gale [1987], and Jacklin [1987] among others - have recently been extended to the realm of ongoing dynamic overlapping generations (OLG) economies; see Bencivenga and Smith [1991], Qi [1994], Fulghieri and Rovelli [1993] and Dutta and Kapur [1994]. These models may be used to evaluate the essential tradeoffs which arise in two distinct areas of current debate. The first is on the evaluation of alternative (or possibly complementary) arrangements for the monetary system, for example the relative merits of a market-value-based vs. an intermediated banking system. The second interpretation, clearly requiring a considerably longer definition of the time unit, pertains to the debate on the viability and desirability of differing social security (pension) systems, and in particular the choice between Pay-As-You-Go (PAYG) versus Funded or Capital-reserve systems (e.g. Boldrin and Rustichini, 1995). Our paper considers the latter set of tradeoffs in an OLG model outside of long-run steady states. In our paper, agents choose among intertemporal liquidity-sharing or insurance mechanisms, in an OLG economy in which the capital stock is possibly longer-lived than agents’ lives. We focus on the comparisons between “Banking” or PAYG systems, of both an intra- and inter-generational nature, versus Stock Market based inter-generational financial systems. The essential tradeoffs across these institutional mechanisms arise from the need for financial contracting to provide agents with liquidity insurance, related to the uncertainty regarding their preferences for their intertemporal allocation of consumption1. Such preference shocks, coupled with real investment opportunities that are long-lived and technologically illiquid, create a demand for interim stock markets, or financial intermediaries and contracts such as deposits withdrawable (fully or partially) on demand. In dynamic OLG economies, there is in addition the possibility of designing contracts (or mechanisms) for sharing such liquidity risk across generations, taking advantage of the "life cycle" structure to the demand for savings of each generation. With the long horizons envisioned, one may think of events such as uncertain workers’ disability as a source of early withdrawal demand. In recent papers, by Qi [1994], Fulghieri and Rovelli [1993], and Dutta and Kapur [1994], alternative institutional arrangements for such intertemporal consumption-smoothing and liquidity risk-sharing mechanisms, within and across generations, have been examined and compared.
iv Under some assumptions, in particular the restriction of the analysis to steady states only, both Qi [1994] and Fulghieri and Rovelli [1993] show that there exist intergenerational financial intermediation contracts that can attain the Golden Rule (Phelps, 1961) levels of investment and consumption-smoothing. The paper of Fulghieri and Rovelli, and the later work of Bhattacharya and Padilla [1996], also focus on the comparison of intergenerational Banks versus Stock Markets for investments that are long-lived in nature. In contrast to the earlier models, which focus either on a static economy (Bhattacharya and Gale, 1987) or on intra-generational mechanisms only (Bencivenga and Smith, 1991), Fulghieri and Rovelli find that stock markets result (in steady state) in underinvestment in the long-lived technology, relative to the Golden Rule optimum, without any investment in short-term liquid technologies or early liquidation2. This occurs essentially owing to the incompleteness of contracts for resource transfers across generations, which in a stock market economy can take place only through real investments in long-lived technologies and bilateral trading at their interim market valuations. In all of the above mentioned papers, the analysis of the functioning of intermediaries and stock markets is confined to steady states, in which the inheritance of a stationary level of investment by prior generations is taken as given. However, given the endowment structure in these models, this is a restrictive assumption, especially vis à vis the first generation which only has its initial endowment to start a resource-transfer cum liquidity-insurance scheme. While Qi [1994] briefly examines this issue (see his Propositions 5 and 6), his analysis is focused on PAYG systems only, rather than on the choices among alternative risk sharing mechanisms. More specifically, the following questions remain unanswered: given initial conditions, is it possible to predict which allocational mechanisms would be chosen by the generations present, and would the same choice be repeated by the subsequent generations, i.e., be intergenerationally stable or Self-Sustaining? For example, consider the three institutional mechanisms which we shall analyze below: (i) Intragenerational Bank (B), (ii) Intergenerational Banking (IB) and (iii) Intergenerational Stock Market (SM). Suppose these may be ranked in a particular way by the steady state welfare criterion. Then, is it necessarily true that, when the issues related to (a) transition to the steady state, and (b) non-cooperative intergenerational choices are taken into account, the most efficient steady state allocational mechanism would indeed be chosen and be intergenerationally self-sustaining? We seek to address these questions in this paper. In defining our notions of inter-generational choices over mechanisms for financing of investment and consumption smoothing, we shall not be employing notions such as the Core, whose
v applicability to the OLG setting is dubious at best. As Esteban [1986] and Esteban and Millan [1990] have shown, even when the Golden Rule allocation does improve the welfare of the first generation (the Samuelson, 1958, case of Gale, 1973), it is not in the intergenerational Core (never for a single commodity setting)3. Fortunately, as Esteban and Sakovics [1993] have emphasized, negative results like those mentioned above do not imply that intergenerational transfer mechanisms become infeasible. Society may cope with such a conundrum, created by the "not in the Core" result, by restricting the language or strategy space of intergenerational proposals for choosing its Institutional Mechanisms for intergenerational reallocations. In the context of the Esteban and Sakovics model, of a pure-exchange OLG environment with agents' endowments tilted to the early period(s) of life, they argue that in a time-stationary model the appropriate notion of an Institutional proposal is that of a time-stationary transfer scheme, from each current and future generation when young to its preceding generation when old. Institutions survive if and only if they constitute Subgame Perfect Equilibria (SPE) (Selten, 1975) in such an intergenerational proposal game. A similar restriction is employed by Boldrin and Rustichini [1995], in their modelling of intergenerational social security schemes versus intra-generational accumulation of short-lived investments. In our model below, we define the concept of a Self-Sustaining investment and consumption allocation Mechanism, using restrictions motivated by analogous considerations. The spirit of these is that proposals that make later generations strictly worse off than the proposing one are not allowed. Such a restriction may arise spontaneously for market-based mechanisms, or be a result of societal Norms embodied in a legislated Pay As You Go (PAYG) mechanism such as IB. In particular, we restrict the set of proposals by each generation to be one of the three Institutions that were indicated above: Transition to Steady-State of Intergenerational Banking (TIB), Intergenerational Stock Market (SM), and Intragenerational Banking (B) (The first of these takes into account the lack of inherited prior capital investments for the first generation). Some of these proposals may require that an older and a younger generation agree on the "sharing" of the inherited capital stock plus the endowment of the young. Hence, each stage game in our model is one of Proposals and Counter-Proposals by the two overlapping generations, unlike in Esteban and Sakovics [1993] where only young propose transfers to the currently old. Our paper is organized as follows. In Section II, we set out the basic model and delineate the Golden Rule and other steady state allocations under the three alternative institutional arrangements of B, IB, and SM. The dynamics of transition to these steady states, and the ranking
vi of welfare levels attainable under each alternative institutional mechanism ( TIB, SM, B), taking transition into account, are the foci of Section III. Intergenerational Proposal Games over these (transition) Institutions and their subgame perfect equilibria, which we term intergenerationally Self- Sustaining Mechanisms, are described and characterized in Section IV. In Section V we conclude with a discussion of our results in the context of recent literature and with suggestions for further research. The major results that emerge from the analysis in the paper are the following. First, we point out that there exists a feasible stationary consumption path (TIB) such that capital accumulation along that path will lead, within a finite horizon, to the Golden Rule steady state optimal allocation. Furthermore the consumption levels attained along this path ensure that the welfare of each generation involved in the transition strictly exceeds that attainable from any intragenerational mechanism4. Second, we show that there exists a faster one-shot transition to the steady state intergenerational stock market (SM) allocation (Fulghieri and Rovelli, 1993)5. Third, we prove that a transition to SM that benefits the proposing current young, which exists unless they are extremely risk averse, must involve strictly rationing the participation of the concurrent preceding generation in the initial market (only). Fourth, we prove that under reasonable parameter values, there exists an intergenerational conflict in any stage game. Specifically, at any time, the currently young may stand to gain by proposing a (rationed and immediate) transition to the stationary intergenerational stock market allocation (SM), whereas the currently middle-aged are made strictly worse off, relative to the "slowest turnpike" allocation (TIB), from accepting such a proposal. Finally, in the last step we delineate how this conflict may be resolved, in the sequence of equilibria of the intergenerational proposal games across overlapping generations. The equilibrium outcome of the intergenerational proposal games depends crucially on the relative expected utility payoff(s) of the middle-aged generation(s) from their rationed transition to the stock market (SM), versus their fallback option of doing intragenerational liquidity risk-sharing (B), based on their inherited endowment net of prior payout obligations. It turns out that the currently middle-aged generation(s) are better off from their fallback option (B), relative to the constrained (for them only) transition to SM proposed by their “children”, if and only if (i) the proportion of the generation requiring early liquidity (withdrawal) is small, and (ii) agents in the generation are ex ante highly risk-averse (or averse to unsmoothed consumption profiles). If either of these conditions on agents’ ex ante characteristics is not satisfied then, and only then, the stock market (SM) institutional arrangement (which is always suboptimal relative to IB in the steady
vii state) becomes the unique subgame - and “trembling-hand” perfect (Selten, 1975) equilibrium of the intergenerational proposal games6. II. THE MODEL AND STATIONARY ALLOCATIONS II.A Optimal Allocations and Intermediated Outcomes We consider an extended version of the overlapping generation (OLG) model of Samuelson [1958], that was first advanced by Bryant [1980]. Time is indexed by t = 0, 1, 2, ..., ∞. Each generation consists of a continuum of agents of unit measure. An agent born at time point t ≥ 0 is alive at time point (t+1), and possibly also at time point (t+2). Each generation is endowed with one unit of the single commodity, spread equally across its members, at birth; this may be thought of as labor income or prior savings. To simplify matters, we assume that an agent born at t only wishes to consume from his savings at time-point (t+1) if he is an early dier, or at time-point (t+2) if she is a late dier. Such simple ex post corner preferences, introduced by Diamond and Dybvig [1983], imply that the allocations produced by intragenerational stock markets and banking systems coincide, provided market participation is suitably restricted; see Jacklin [1987], and Bhattacharya and Gale [1987]. Agents become early or late diers with probabilities ε and (1-ε) respectively. These events are independent across the continuum of agents in each generation, so that a proportion (1-ε) of agents born at time t is alive at time (t+2), when they die. Thus, the ex ante expected utility of a representative agent of generation t is given by: t, , V = U( ) + (1-)U( ) ε ε C C t t1 2 (1) where { } C C t t, , , 1 2 are the (certain) allocations of consumption to early and late diers of generation t, respectively. The utility function U(.) is assumed to be strictly increasing and concave, with U’(0) = ∞. There are two investment technologies in the model. The first is long-lived with constant returns to scale: investment of It at time t returns RIt, R > 1 , at time (t+2). The second technology is that of storage, without any loss or depreciation, between time points t to (t+1). The long-lived investment It can also be liquidated at time (t+1) to yield Q Lt+1 , where 0 ≤ Q ≤ 1, and 0 ≤ Lt+1 ≤ It ; we assume Q = 1 , so that early liquidation and storage technologies are equivalent7.
viii The Golden Rule (Phelps, 1961) stationary optimal intergenerational investment and consumption plan thus solves the problem: {} [ ] * , , , , , V ( ) ( ) ( ) , , ≡ + −Max C C ILC C t t t t t t1 2 1 2 1ε ε U U (2) s.t.: ε ε C C t t t t t t R IL L It − − − − + − ≤ + − + − ∀ 1 1 2 2 2 1 1 1 , , ( ) ( ) , , (2.a) 0 L I t t ≤ ≤ ≤ +11(2.b) As shown, for example, in Fulghieri and Rovelli (1993, Proposition 1), the solution to this problem is characterised by full investment of the initial endowment of each new generation, no liquidation of invested capital, and a level of consumption for early and late diers which is constant and equal to R. Furthermore, Qi [1994], Fulghieri and Rovelli [1993], and Dutta and Kapur [1994] give the conditions under which the Golden Rule allocation may be implemented by an intergenerational banking system. These results are summarised in the following proposition. Proposition 1. (a) (Fulghieri and Rovelli 1993).The Golden Rule (Stationary) Optimal Intergenerational Allocation Rule, which maximizes (2) subject to (2.a) and (2.b), is given by: t *t *,1 *, * I = 1,L = 0,C=C=R,t. t t 2∀ (b) (Qi 1994, Fulghieri and Rovelli 1993, and Dutta and Kapoor 1994). If agents/depositors can be subject to a No Redepositing Condition within (and across) banks, then a single welfare-maximizing bank (or equivalently, many representative competing banks) would, in equilibrium, offer each generation the deposit contract: It in deposits leads to withdrawal rights RIt at time (t+1) or time (t+2), at the choice of the depositor. Furthermore, all agents born at t would deposit all their initial endowment in banks, which would invest it (It =1) in long-lived investments only, without early liquidation. Thus intergenerational banking (IB) would implement the Golden Rule optimum. Hence in particular: C C R t i IB t i, , * = = , { } i ∈∀ t12, , (c) If redepositing after early withdrawal by agents is not monitorable (indistinguishable from an initial deposit), especially if done across banks, but depositing at one bank by another can be prohibited, then ex ante Bertrand competition among banks would lead to the outcome:
xv solutions. The first, due to Fulghieri and Rovelli [1993], and Dutta and Kapur [1994], satisfies: pR τκκ +=∀≥,0. But, as we shall see in the next section, it also requires rationing of the very first generation to sell in such a market at τ. The second solution is a twoperiodic one, with pfor τκ κ + = ≥ 1 1, and odd, and pR τκ κ + = ∀ ≥ ,2 and even, where τ is the first generation to trade in such a market. To simplify notation (only) in the following ‘steady-state’ result, we assume τ = 0, although that is inconsistent with the 0-th generation having no prior investments to buy in the secondary market. Proposition 3. The set of stationary or periodic intergenerational steady state stock market equilibria, satisfying equations (7a,b,c,d,e,f,g,h) and (8), are isolated, being one of the two following, assuming the anticipation of stock market trading commencing at some t=0: (a) { } = R, t=1,2,..., ,pt∀ ∞ (6.e) implying: {I} = I, t=,2,..., , tSM ∀ ∞1 which, from equation (9) satisfies: 1-I = R[-(1-I)], SM SM ε that is: SM I = 1-1+R, εR(10) or: (b) { } = 1 pt for t = 1, 3, 5, ..., odd ; and (6.f) { } = R pt for t = 2, 4, 6, ..., even (6.g) with: It = 1 for t ≥ 0 even, and (11.a) It = 1 - εfor t 1 ≥ odd, (11.b) which clearly satisfies equation (9). The consumption allocations at the two alternative stock market allocations are given, using equations (8.a,b), by: (a) C R C R t t t, , ;, 1 2 0= = ∀≥,(12) and: (b) C C R t t t, , ;, 1 2 1 0 = = ∀ ≥ and even ;
xvi C R C R t t t, , ;, 1 2 0 = = ∀ ≥ and odd. (13) Proof. For Part (a), see Fulghieri and Rovelli [1993] or Dutta and Kapur [1994]. For Part (b), it suffices to note that eqs. (6.f, g) , (11.a, b) and (13) satisfy (9). Remarks. (i) If stock market equilibrium commences at time t = 0, with I t - 1 = 0, then given (7.a) the only intergenerational stock market equilibrium consistent with the initial condition is the periodic stock market equilibrium (b). (ii) Also note that, since iterating equation (6.d) implies that p p t t t = ∀ ≥ +21,, no other periodic stock market equilibrium with periodicity k > 2 exists. We now move on to consider the dynamics of transition to the steady-state intergenerational allocation and risk-sharing mechanisms described above. I. INTRODUCTION AND SUMMARY Models of intertemporal liquidity risks, or "preference shocks" - developed in Bryant [1980] and further examined in the works of Diamond and Dybvig [1983], Bhattacharya and Gale [1987], and Jacklin [1987] among others - have recently been extended to the realm of ongoing dynamic overlapping generations (OLG) economies; see Bencivenga and Smith [1991], Qi [1994], Fulghieri and Rovelli [1993] and Dutta and Kapur [1994]. These models may be used to evaluate the essential tradeoffs which arise in two distinct areas of current debate. The first is on the evaluation of alternative (or possibly complementary) arrangements for the monetary system, for example the relative merits of a market-value-based vs. an intermediated banking system. The second interpretation, clearly requiring a considerably longer definition of the time unit, pertains to the debate on the viability and desirability of differing social security (pension) systems, and in particular the choice between Pay-As-You-Go (PAYG) versus Funded or Capital-reserve systems
xvii (e.g. Boldrin and Rustichini, 1995). Our paper considers the latter set of tradeoffs in an OLG model outside of long-run steady states. In our paper, agents choose among intertemporal liquidity-sharing or insurance mechanisms, in an OLG economy in which the capital stock is possibly longer-lived than agents’ lives. We focus on the comparisons between “Banking” or PAYG systems, of both an intra- and inter-generational nature, versus Stock Market based inter-generational financial systems. The essential tradeoffs across these institutional mechanisms arise from the need for financial contracting to provide agents with liquidity insurance, related to the uncertainty regarding their preferences for their intertemporal allocation of consumption1. Such preference shocks, coupled with real investment opportunities that are long-lived and technologically illiquid, create a demand for interim stock markets, or financial intermediaries and contracts such as deposits withdrawable (fully or partially) on demand. In dynamic OLG economies, there is in addition the possibility of designing contracts (or mechanisms) for sharing such liquidity risk across generations, taking advantage of the "life cycle" structure to the demand for savings of each generation. With the long horizons envisioned, one may think of events such as uncertain workers’ disability as a source of early withdrawal demand. In recent papers, by Qi [1994], Fulghieri and Rovelli [1993], and Dutta and Kapur [1994], alternative institutional arrangements for such intertemporal consumption-smoothing and liquidity risk-sharing mechanisms, within and across generations, have been examined and compared. Under some assumptions, in particular the restriction of the analysis to steady states only, both Qi [1994] and Fulghieri and Rovelli [1993] show that there exist intergenerational financial intermediation contracts that can attain the Golden Rule (Phelps, 1961) levels of investment and consumption-smoothing. The paper of Fulghieri and Rovelli, and the later work of Bhattacharya and Padilla [1996], also focus on the comparison of intergenerational Banks versus Stock Markets for investments that are long-lived in nature. In contrast to the earlier models, which focus either on a static economy (Bhattacharya and Gale, 1987) or on intra-generational mechanisms only (Bencivenga and Smith, 1991), Fulghieri and Rovelli find that stock markets result (in steady state) in underinvestment in the long-lived technology, relative to the Golden Rule optimum, without any investment in short-term liquid technologies or early liquidation2. This occurs essentially owing to the incompleteness of contracts for resource transfers across generations, which in a stock market economy can take place only through real investments in long-lived technologies and bilateral trading at their interim market valuations.
xviii In all of the above mentioned papers, the analysis of the functioning of intermediaries and stock markets is confined to steady states, in which the inheritance of a stationary level of investment by prior generations is taken as given. However, given the endowment structure in these models, this is a restrictive assumption, especially vis à vis the first generation which only has its initial endowment to start a resource-transfer cum liquidity-insurance scheme. While Qi [1994] briefly examines this issue (see his Propositions 5 and 6), his analysis is focused on PAYG systems only, rather than on the choices among alternative risk sharing mechanisms. More specifically, the following questions remain unanswered: given initial conditions, is it possible to predict which allocational mechanisms would be chosen by the generations present, and would the same choice be repeated by the subsequent generations, i.e., be intergenerationally stable or Self-Sustaining? For example, consider the three institutional mechanisms which we shall analyze below: (i) Intragenerational Bank (B), (ii) Intergenerational Banking (IB) and (iii) Intergenerational Stock Market (SM). Suppose these may be ranked in a particular way by the steady state welfare criterion. Then, is it necessarily true that, when the issues related to (a) transition to the steady state, and (b) non-cooperative intergenerational choices are taken into account, the most efficient steady state allocational mechanism would indeed be chosen and be intergenerationally self-sustaining? We seek to address these questions in this paper. In defining our notions of inter-generational choices over mechanisms for financing of investment and consumption smoothing, we shall not be employing notions such as the Core, whose applicability to the OLG setting is dubious at best. As Esteban [1986] and Esteban and Millan [1990] have shown, even when the Golden Rule allocation does improve the welfare of the first generation (the Samuelson, 1958, case of Gale, 1973), it is not in the intergenerational Core (never for a single commodity setting)3. Fortunately, as Esteban and Sakovics [1993] have emphasized, negative results like those mentioned above do not imply that intergenerational transfer mechanisms become infeasible. Society may cope with such a conundrum, created by the "not in the Core" result, by restricting the language or strategy space of intergenerational proposals for choosing its Institutional Mechanisms for intergenerational reallocations. In the context of the Esteban and Sakovics model, of a pure-exchange OLG environment with agents' endowments tilted to the early period(s) of life, they argue that in a time-stationary model the appropriate notion of an Institutional proposal is that of a time-stationary transfer scheme, from each current and future generation when young to its preceding generation when old. Institutions survive if and only if they constitute Subgame Perfect Equilibria (SPE) (Selten, 1975) in such an intergenerational proposal game. A
xix similar restriction is employed by Boldrin and Rustichini [1995], in their modelling of intergenerational social security schemes versus intra-generational accumulation of short-lived investments. In our model below, we define the concept of a Self-Sustaining investment and consumption allocation Mechanism, using restrictions motivated by analogous considerations. The spirit of these is that proposals that make later generations strictly worse off than the proposing one are not allowed. Such a restriction may arise spontaneously for market-based mechanisms, or be a result of societal Norms embodied in a legislated Pay As You Go (PAYG) mechanism such as IB. In particular, we restrict the set of proposals by each generation to be one of the three Institutions that were indicated above: Transition to Steady-State of Intergenerational Banking (TIB), Intergenerational Stock Market (SM), and Intragenerational Banking (B) (The first of these takes into account the lack of inherited prior capital investments for the first generation). Some of these proposals may require that an older and a younger generation agree on the "sharing" of the inherited capital stock plus the endowment of the young. Hence, each stage game in our model is one of Proposals and Counter-Proposals by the two overlapping generations, unlike in Esteban and Sakovics [1993] where only young propose transfers to the currently old. Our paper is organized as follows. In Section II, we set out the basic model and delineate the Golden Rule and other steady state allocations under the three alternative institutional arrangements of B, IB, and SM. The dynamics of transition to these steady states, and the ranking of welfare levels attainable under each alternative institutional mechanism ( TIB, SM, B), taking transition into account, are the foci of Section III. Intergenerational Proposal Games over these (transition) Institutions and their subgame perfect equilibria, which we term intergenerationally Self- Sustaining Mechanisms, are described and characterized in Section IV. In Section V we conclude with a discussion of our results in the context of recent literature and with suggestions for further research. The major results that emerge from the analysis in the paper are the following. First, we point out that there exists a feasible stationary consumption path (TIB) such that capital accumulation along that path will lead, within a finite horizon, to the Golden Rule steady state optimal allocation. Furthermore the consumption levels attained along this path ensure that the welfare of each generation involved in the transition strictly exceeds that attainable from any intragenerational mechanism4. Second, we show that there exists a faster one-shot transition to the steady state intergenerational stock market (SM) allocation (Fulghieri and Rovelli, 1993)5. Third,
xx we prove that a transition to SM that benefits the proposing current young, which exists unless they are extremely risk averse, must involve strictly rationing the participation of the concurrent preceding generation in the initial market (only). Fourth, we prove that under reasonable parameter values, there exists an intergenerational conflict in any stage game. Specifically, at any time, the currently young may stand to gain by proposing a (rationed and immediate) transition to the stationary intergenerational stock market allocation (SM), whereas the currently middle-aged are made strictly worse off, relative to the "slowest turnpike" allocation (TIB), from accepting such a proposal. Finally, in the last step we delineate how this conflict may be resolved, in the sequence of equilibria of the intergenerational proposal games across overlapping generations. The equilibrium outcome of the intergenerational proposal games depends crucially on the relative expected utility payoff(s) of the middle-aged generation(s) from their rationed transition to the stock market (SM), versus their fallback option of doing intragenerational liquidity risk-sharing (B), based on their inherited endowment net of prior payout obligations. It turns out that the currently middle-aged generation(s) are better off from their fallback option (B), relative to the constrained (for them only) transition to SM proposed by their “children”, if and only if (i) the proportion of the generation requiring early liquidity (withdrawal) is small, and (ii) agents in the generation are ex ante highly risk-averse (or averse to unsmoothed consumption profiles). If either of these conditions on agents’ ex ante characteristics is not satisfied then, and only then, the stock market (SM) institutional arrangement (which is always suboptimal relative to IB in the steady state) becomes the unique subgame - and “trembling-hand” perfect (Selten, 1975) equilibrium of the intergenerational proposal games6. II. THE MODEL AND STATIONARY ALLOCATIONS II.A Optimal Allocations and Intermediated Outcomes We consider an extended version of the overlapping generation (OLG) model of Samuelson [1958], that was first advanced by Bryant [1980]. Time is indexed by t = 0, 1, 2, ..., ∞. Each generation consists of a continuum of agents of unit measure. An agent born at time point t ≥ 0 is alive at time point (t+1), and possibly also at time point (t+2). Each generation is endowed with one unit of the single commodity, spread equally across its members, at birth; this may be thought of as labor income or prior savings. To simplify matters, we assume that an agent born at t only wishes to consume from his savings at time-point (t+1) if he is an early dier, or at time-point (t+2) if she is a late dier. Such simple ex post corner
xxi preferences, introduced by Diamond and Dybvig [1983], imply that the allocations produced by intragenerational stock markets and banking systems coincide, provided market participation is suitably restricted; see Jacklin [1987], and Bhattacharya and Gale [1987]. Agents become early or late diers with probabilities ε and (1-ε) respectively. These events are independent across the continuum of agents in each generation, so that a proportion (1-ε) of agents born at time t is alive at time (t+2), when they die. Thus, the ex ante expected utility of a representative agent of generation t is given by: t, , V = U( ) + (1-)U( ) ε ε C C t t1 2 (1) where { } C C t t, , , 1 2 are the (certain) allocations of consumption to early and late diers of generation t, respectively. The utility function U(.) is assumed to be strictly increasing and concave, with U’(0) = ∞. There are two investment technologies in the model. The first is long-lived with constant returns to scale: investment of It at time t returns RIt, R > 1 , at time (t+2). The second technology is that of storage, without any loss or depreciation, between time points t to (t+1). The long-lived investment It can also be liquidated at time (t+1) to yield Q Lt+1 , where 0 ≤ Q ≤ 1, and 0 ≤ Lt+1 ≤ It ; we assume Q = 1 , so that early liquidation and storage technologies are equivalent7. The Golden Rule (Phelps, 1961) stationary optimal intergenerational investment and consumption plan thus solves the problem: {} [ ] * , , , , , V ( ) ( ) ( ) , , ≡ + −Max C C ILC C t t t t t t1 2 1 2 1ε ε U U (2) s.t.: ε ε C C t t t t t t R IL L It − − − − + − ≤ + − + − ∀ 1 1 2 2 2 1 1 1 , , ( ) ( ) , , (2.a) 0 L I t t ≤ ≤ ≤ +11(2.b) As shown, for example, in Fulghieri and Rovelli (1993, Proposition 1), the solution to this problem is characterised by full investment of the initial endowment of each new generation, no liquidation of invested capital, and a level of consumption for early and late diers which is constant and equal to R. Furthermore, Qi [1994], Fulghieri and Rovelli [1993], and Dutta and Kapur [1994] give the conditions under which the Golden Rule allocation may be implemented
xxii by an intergenerational banking system. These results are summarised in the following proposition. Proposition 1. (a) (Fulghieri and Rovelli 1993).The Golden Rule (Stationary) Optimal Intergenerational Allocation Rule, which maximizes (2) subject to (2.a) and (2.b), is given by: t *t *,1 *, * I = 1,L = 0,C=C=R,t. t t 2∀ (b) (Qi 1994, Fulghieri and Rovelli 1993, and Dutta and Kapoor 1994). If agents/depositors can be subject to a No Redepositing Condition within (and across) banks, then a single welfare-maximizing bank (or equivalently, many representative competing banks) would, in equilibrium, offer each generation the deposit contract: It in deposits leads to withdrawal rights RIt at time (t+1) or time (t+2), at the choice of the depositor. Furthermore, all agents born at t would deposit all their initial endowment in banks, which would invest it (It =1) in long-lived investments only, without early liquidation. Thus intergenerational banking (IB) would implement the Golden Rule optimum. Hence in particular: C C R t i IB t i, , * = = , { } i ∈∀ t12, , (c) If redepositing after early withdrawal by agents is not monitorable (indistinguishable from an initial deposit), especially if done across banks, but depositing at one bank by another can be prohibited, then ex ante Bertrand competition among banks would lead to the outcome: depositing It=1 by each depositor of each generation t confers her with withdrawal rights CU 1at time (t+1) or CU 2 at time (t+2), where these satisfy (suppressing the t-subscript): C C U U 1 2 < R < ,(4.a) ( ) 2 C C U U 1 2 ≤(4.b) ε ε C C 1 2 U U + (1-)=R(4.c) Proof. See either Qi, Fulghieri and Rovelli, or Dutta and Kapur. Remark. The Golden Rule optimum is, of course, strictly valid as a solution concept only when t ∈{-∞, ...,-1 , 0 , 1, ..., ∞} since, for example, if t ∈ { 0, 1, ..., ∞}, and I1 = 1 and L1 = 0, then the early diers of generation 0 will obtain consumption C0 1 0 ,=. These issues raise questions about the transition to these stationary optimal allocations, which we discuss in the next section. The notion of a competitive banking system is also not unambiguous in an OLG
xxiii setting, and one may think instead of the resulting allocations as arising from a legislated allocation mechanism. In contrast, the constrained optimal intragenerational allocation, solves the problem: [ ] V U(C) + (1-)U(C) , , , 1 2 ≡ Max ILC C 1 2 ε ε (3) s.t.: 0 C L, 1 ≤ ≤ ε (3.a) 0 (1-)C R(I- L), 2 ≤ ≤ ε (3.b) 0 I 1. ≤ ≤ (3.c) The solution to this problem is characterized by Diamond and Dybvig [1993], who also show that this allocation may be implemented with a single welfare-maximizing bank or several (representative) competing banks who are Bertrand Competitors and obtain clienteles of strictly positive measure. The latter result assumes no interim trading of assets across competing banks at the interim period; see Bhattacharya and Gale [1987]. Proposition 2. (Diamond and Dybvig 1993). The Optimal Intragenerational Allocation Rule, which maximizes in (3) subject to (3.a) to (3.c), is given by: I = 1, ′ ′ U(C) = RU(C), 1 2 CLC R L _ /; _ ( _ )/( ) 1 2 1 1= = − −ε ε Furthermore, in a competitive intragenerational banking system equilibrium, deposit contracts of the form { deposit I at time t implies withdrawal rights C C I B 11 = or C C I B 22 =, at times (t + 1) or (t + 2) respectively} would prevail, leading to expected utility V V B=, and each depositor would deposit I = 1 at one (or many) bank(s), and each bank would invest its deposits in short-term (early liquidation) and long-term investments in the proportions L and 1 - L , respectively. Proof. See Diamond and Dybvig [1983]8. The following implications of Proposition 2 are straightforward. Corollary 1. For all ε > 0 , the expected utility VB of an agent in the optimal intragenerational allocation is strictly less than its level V* at the Golden Rule intergenerational optimum.
xxiv Proof. Since C C _ , _ 1 2 0 > and hence L ∈ (0,1), given U'(0) = ∞, we have ( ) 1 < = L+R(1-L) < R,ε ε C C 1 2 1+ − whereas C C R 1 2 * * = = , from which the conclusion follows using Jensen's inequality. QED. Corollary 2. (Diamond and Dybvig, 1983) If U(Ci ), i ∈ {1,2}, has its Relative Risk Aversion Coefficient uniformly strictly greater than unity (that for logarithmic utility), then 1 < R, C C 1 2 < < implying: ε < L < 1. We should briefly mention the alternatives to the existence of bank-type intermediaries (and "unconditional" withdrawal rights) in this setting. The extreme one is that of intra- and inter-generational Autarky, under which each agent would invest I=1 when young, and consume CA 11= if she is an early dier, and C R A 2= if she is a late dier, with expected utility: A1 A2 A V U(C) + (1-)U(C)≡ε ε , (5) where it is obvious that: A* V V < V,≤ with the first inequality also strict unless U(C) = log(C), since { L = ε, C1 = 1, C2 = R} is always feasible under intragenerational banking. A less extreme comparison is with an institution often analyzed in "standard" financial modelling: the Intragenerational Stock Market. If individuals still invest on their own, but can sell their long-term investments It at a linear price pI t t+1 at time (t+1), then it turns out that the interim value-maximizing consumption-investment outcome will be the same under intragenerational stock markets as under Autarky; see Bhattacharya and Gale [1987]. The reason is that, under the interim value/wealth maximization criterion used to choose liquidation Lt+1 from It , subject to 0 ≤ Lt+1 ≤ It , Lt+1 ∈ (0 , It) will be chosen by individuals if and only if pt+ = 11.With the ex post corner preferences assumed, this will be the case if and only if { } C C R t t, , ; 1 2 1= = , as shown by Bhattacharya and Gale [1987]. Bencivenga and Smith [1991] and others have emphasized this difference between Banking and Stock Market allocations, suggesting that it should apply even when the setting is intergenerational in nature. However, as argued in Bhattacharya and Gale [1987] and Bhattacharya and Padilla [1996], with intragenerational stock markets investment decisions
xxxi • the new born at time t can always revise both their proposal/program at time t and their proposal (planned) at time (t+1). We restrict the pure strategy set of each overlapping generation in the stage games at times t= 1, 2, ..., ∞ , to be the 3-tuple of allocation mechanisms (or Institutions, in the terminology of Esteban and Sakovics) {B, TIB, SM }, that were characterized in Sections II and III above. We do so for the following reasons. First, as we have suggested earlier, the TIB and SM intergenerational allocation mechanisms have the feature that, if at time t a current (agreed upon) mechanism is started and is continued through times t , t+1 , t+2 , ..., etc., then generations subsequent to the new born at time t are never worse off than the t-th generation. We postulate that reasonable proposal processes (e.g., in a legislature) satisfy this “minimal degree of fairness” property, even though the resulting mechanism need not guarantee, say, ultimate attainment of the Golden Rule optimum in the long run. The spirit of this restriction is identical to that underlying the transfer institutions or proposals described in Esteban and Sakovics [1993] or Boldrin and Rustichini [1995]. Second, while we have not proved that the three institutions / mechanisms which each generation is allowed to propose are the only ones that are (i) intergenerationally stationary for future generations on their continuation paths, and (ii) viable, they do represent the set of major significant alternatives in many aggregate (macroeconomic) settings, such as: generationally autarkic, pay as you go, and market-based pension plan systems. Once we step outside the class of these three mechanisms, we essentially open a Pandora’s Box of proposals that satisfy the second criterion (ii) but not the first (i), that we deem desirable to impose12. The stage games, at each time-point of the infinite horizon intergenerational proposal game, could themselves involve (at least) three alternative extensive forms: A) The middle-aged moving first and proposing, and the young counter-proposing; or B) The new born moving first and proposing, and the middle-aged counter-proposing; or C) Simultaneous proposals by the two generations overlapping at any point in time. Of these alternatives, cases (B) and (C) are relatively more in accordance with the spirit of the transfer-proposal game of Esteban and Sakovics (1993). Furthermore, the equilibrium outcome in case (A) is also somewhat obvious, given that VTIB > VB (Proposition 4) and VTIB > VSM(MA) (Proposition 6) always. It turns out, on the other hand, that (i) the equilibrium outcomes in cases (B) and (C), suitably defined, generically coincide with each other, and (ii)
xxxii they depend on the relative magnitude of VTIB versus VSM(NB), as well as on VSM(MA) versus VB . We now proceed to describe the agents’ strategies and payoffs more precisely, and also to define the notion of Self-Sustaining Mechanisms arising from these proposal games across current and future generations. TABLE 2. Payoffs in (First Stage) Proposal Game MIDDLE-AGED NEW BORN BTIB SM B { V B, V B} {V(c), V B} {V(c), V B} TIB {V(c), V B} {V( ′ c ), V TIB } {V(c), V B} SM {V(c), V B} {V(c) , V B } { V SM (NB), V SM (MA)} In Table 2, we describe the pure strategy payoffs in the simultaneous moves case (C) of the proposal stage scheme, across generations 0 and 1, postponing the discussion of subsequent stage games until the subsequent result (Proposition 8). Notice that the continuation payoff, V(c) or V( ′ c ), of the new born in the event of disagreement takes into account the optimal reswitching strategy of these agents. It is easy, using Table 2, to write the payoffs in the extensive form game trees of the stage games in cases (A) and (B) above. Mixtures of stage game pure strategies (uncorrelated) by players are allowed in all the cases. We define the continuation payoff for the new born, at any stage game at time t, as: V(c) = [VTIB] , if {TIB, TIB} is in [ SSM (t +1) Ht+1 = (1, 0) ] (22.a) and : V(c) = [ VSM (MA) Ht+1 = (1, 0) ], otherwise, (22.b) where Ht+1 is the vector of real capital stock (one-period old and maturing) which the current (as of time t) new born generation will be holding at time (t+1), and SSM (.) represents the notion of Self-Sustaining Mechanisms, to be defined below. The payoff V( ′ c ) is
xxxiii analogously defined, with Ht+1 now equaling the inherited capital stock vector along the TIB path, starting at t = 0. We conclude this subsection with the following definition of equilibrium outcomes in the proposal game: Definition. A Self-Sustaining Mechanism (SSM) is a sequence of outcomes of the intergenerational stage games at times t = 1, 2, ..., ∞ , such that the outcomes and equilibrium strategies (given Ht ) at each time t, {SSM (t)}, satisfy the criteria that, ∀ t ≥ 1: (1) SSM(t) ∈ NE(t), Nash Equilibrium in stage t ; (2) { } SSM t ( )ττ= ∞∈ NE(τ), Subgame Perfect Equilibrium across stages (Selten, 1975); (3) { } SSM t( ) ∈ SPE(t) in the stage game at time t ,in cases (A) and (B), or: { } SSM t( ) ∈ TPE(t), the Trembling Hand Perfect Equilibrium (Selten, 1975) set at t , in case (C) (simultaneous-move) of the stage game. IV.B Equilibrium outcomes: Only the meek shall inherit the earth We can now state and prove the result suggested in the previous section (Propositions 6 and 7). We do so assuming that V(c) = V( ′ c ). While this assumption on V(c) is critical for an unique outcome in a subset of case (C), and it may appear to be unduly restrictive, in Proposition 8 below we shall prove that, under reasonable conditions, it is naturally satisfied13. Theorem. Self-Sustaining Mechanisms in the first stage Intergenerational Proposal Games are characterized as follows: a. In case (A), middle-aged move first, SSM(t) = { } TIB TIB,. b. In case (B), new born move first, then either of the following equilibria may result: b.1. SSM(t) = { } TIB TIB, if either: (b.1.1) VSM (NB) < VTIB ; or: (b.1.2) VSM (NB) ≥ VTIB and VSM (MA) < VB; b.2. SSM(t) = { } SM SM, if: VSM (NB) > VTIB and VSM (MA) ≥ VB;
xxxiv b.3. Both { } SM SM,and { } TIB TIB, ( ) ∈ SSM t if: VSM (NB) = VTIB and VSM (MA) ≥ VB. c. In case (C), the following equilibria may result: c.1. SSM(t) = { } TIB TIB, if either: (c.1.1) VSM (NB) < VTIB ; or: (c.1.2) VSM (MA) < VB . c. 2SSM(t) = { } SM SM, if: VSM (NB) > VTIB and VSM (MA) ≥ VB. c.3 Both { } SM SM, and { } TIB TIB,∈ SSM (t) if: VSM (NB) = VTIB and VSM (MA) ≥ VB. Proof. Part (a) follows from Propositions (4) and (6). In part (c.2), the fact that { } SM SM, is uniquely in SSM(t) follows from using the Trembling Hand Perfect equilibrium concept and from the maintained hypothesis that V(c) = V( ′ c ); see below the discussion of Proposition 8. The remainder of parts (b) and (c) are obvious, on examination of Table 2 and equations (22). Remarks. In economic terms, the above Theorem, together with Propositions 4 , 5 and 6 above, implies the following: • The very first generation will never start a stock market in long lived capital (SM), nor an intragenerationally autarkic bank (B). • The slowest turnpike transition path (TIB) to steady state (Golden Rule) intergenerational banking would survive (continue) at time t if either it is a dominant strategy for both new born and middle aged at time t or, at least, the middle aged at time t can credibly resist a switch to the (steady state, non-periodic) stock market (SM) at time t, by threatening to revert to their intragenerationally autarkic banking allocation (B). • In particular, the most desirable steady state outcome (the Golden Rule supported by intergenerational banking, IB) will be reached when people are sufficiently risk averse and
xxxv are not too likely to require early utilization of their savings (or liquidity), so that Only the Meek Shall Inherit the Golden Rule steady state! For the TIB path to be maintained, we must have that it is also an equilibrium outcome of the intergenerational proposal games at all dates along the accumulation path, until the switch to the Golden Rule steady state can actually take place. We conclude this section by showing that if { } SM SM,is not a Nash equilibrium outcome in the initial stage game, it will continue not to be so for all intergenerational stage games along the accumulation path in a wide class of cases. We do this in the following Proposition 8, which serves to justify the assumption that V(c) = V( ′ c ), which was used in the proof of the Theorem above. Then, given V(c) = V( ′ c ), the proof of the Theorem also applies in all stage games). Furthermore, the same equilibrium outcome would arise in all current and subsequent stages. The thrust of our argument is the following. Assume TIB to be the unique Nash equilibrium outcome in the first stage intergenerational proposal game. Then in cases (b.1.1) and (c.1.1) of the preceding Theorem it will clearly continue to be the chosen equilibrium over time, since the terms of choice between TIB and competing mechanisms will be invariant across stage games, even as capital accumulation proceeds along the transition path. In the other cases, where the new born might prefer a move to SM, but the middle-aged can credibly resist such a proposal in the first stage of the game, by credibly threatening to revert to the intragenerationally autarkic allocation B, then the terms of this choice will change as capital accumulation proceeds along TIB. We can still prove, for the class of homothetic (instantaneous) utility functions, that this credible threat to revert to B will persist along the accumulation path. We can thus state that: Proposition 8. If TIB is a (Nash, and SPE or TPE) equilibrium allocation in the first stage of the intergenerational proposal game, it will continue to be an equilibrium allocation at all subsequent stage games, for t ≥ 2, for all homothetic UCC ( ) ,=−> −1 10 ρ ρρ. Proof. See Bhattacharya, Fulghieri and Rovelli [1997]. Remark. Given Proposition 8 and our main Theorem, and definitions (22.a, b), V( ′ c ) = V(c) holds true for continuation payoffs if: [ ] [ ] VHVMA H BSM 1 1 1 0 1 0= > =(,) ( ) ( ,). Alternatively, suppose [ ] [ ] VHVMA H BSM 1 1 1 0 1 0= < =(,) ( ) ( ,) , but still {SM, SM } is
xxxvi not the unique Trembling-Hand Perfect SSM at time t=1 in case C (simultaneous proposals) because, by hypothesis, V( ′ c ) > V(c). But then {TIB, TIB} must also be in SSM (t+1) because V( ′ c ) > V(c) will continue to hold also at time t = 2, in light of the proof of Proposition 8. But then it contradicts the definition of V(c) in (22.a, b) to assume that V( ′ c ) ≠ V(c) at t=1. V. CONCLUDING REMARKS In this paper we have attempted to extend and synthesize the extant and emerging theoretical frameworks pertaining to (a) static liquidity sharing (consumption smoothing) mechanisms, (b) intergenerational (OLG) tradeoffs and modifications to these, in and out of steady states, and (c) non-cooperative intergenerational proposal games related to Sustainable Mechanisms or Institutions in these settings. Our results show that, if currently young generations are constrained to be “minimally altruistic” towards their descendants (i.e., not to do relative harm to them) in their “feasible” (allowed) proposals, then transition to the long-run optimal Golden Rule of Phelps [1961] outcome may be obtained as the unique non-cooperative equilibrium outcome among generations, but only if the agents involved have preferences embodying a significant degree of relative risk aversion, or curvature in the (additive) utility functions for consumption at different points of their stochastic lifespans, and also relatively low likelihood of liquidity or early withdrawal needs. By embodying such a reasonably restricted notion of intergenerational “autonomy”, we have improved on the essentially planning-theoretic modeling of, for example, Allen and Gale [1995], and also enriched (in context) the methodology of Esteban and Sakovics [1993], obtaining quite different results visà-vis the universality of long-run efficiency attainment. Our model may also be usefully compared to that of Boldrin and Rustichini [1995]: with no long-lived capital and, hence, no transfer of property rights to capital across generations taking place, they find that whether or not PAYG Social Security (intergenerational transfer) systems are voted into existence (and maintained) depends on the return to capital and the dynamics of stochastic population growth (a high rate of growth will favor the adoption of a PAYG system)14. Keeping in mind that both our transition path to intergenerational banking (TIB) and also the intergenerational banking steady state (IB) share an essential similarity with
xxxvii PAYG systems, our results point to the fact that whether or not a PAYG outcome will be chosen, in an environment with long lived capital stock and stochastic lifetimes, may also depend on the extent of agents’ aversion to risk and random needs for early withdrawal or liquidity. A synthesis of these two types of modeling, with roles for both endogenous trade or exchange of property rights to capital and for aggregate shocks to state variables pertaining to (i) agents’ (preference) characteristics, (ii) demographic changes, and (iii) return to capital, should help provide important and policy-relevant answers to intriguing economic questions regarding the nature and efficiency of equilibrium financial intermediation and social security systems.
xxxviii ENDNOTES Section I 1The background assumption is that there are many small uninsurable risks, to endowments (such as accidents), health, or family size, for which it is extremely costly to develop separate insurance markets, in part due to private observability of outcomes, which could be monitored by insurers only at a prohibitive cost. The lack of such markets gives rise to uncertain indirect utility functions for agents, over their intertemporal withdrawal patterns from their invested savings. 2In Bhattacharya and Gale [1987] it is shown that the opening of an interim stock market in the Diamond - Dybvig [1983] model leads to over investment, relative to the ex ante optimal level, in the long-term technology, when agents have relative risk aversion greater than unity and they make investment choices individually. See Bhattacharya and Padilla [1996] for further discussion of these contrasting results, which arise from the dominance of different incomplete markets effects, within and across generations. 3The obvious reason is that each subcoalition of current and future generations has an incentive to deviate from the Golden Rule allocation, by denying the anticipated positive oldage transfer to the immediately prior generation, when each generation's endowment pattern is tilted towards its youth. 4 Hence, the reliance on intra-generational mechanisms in an OLG context, as in Bencivenga and Smith [1991] and others, appears to be not very well justified. 5 In addition, we also show the existence of a steady-state periodic stock market equilibrium. 6Hence, our results differ from those of Esteban and Sakovics [1993] in their simpler “transfer games” context (with no endogenous real investments), in which - as the cost of making new proposals goes to zero - they obtain nearly efficient steady-state outcomes as unique equilibria. Section II 7 The assumption that Q = 1 is used by Diamond and Dybvig [1983] to prove the existence of Panic Bank Runs, of early withdrawals for storage at (t+1) by agents who do not wish to consume from savings until time (t+2). However, simple measures, such as suspension of convertibility, can eliminate such Runs, so we ignore this problem. This assumption considerably simplifies the analysis of transitions to optimal Steady-State allocations. 8This intragenerational optimal allocation (B) is the unique equilibrium outcome of such a contract if banks can suspend early withdrawal rights once a proportion ε of their depositors has been withdrawn (that is, a fraction L of their investments has been liquidated). All agents
xxxix would then deposit all their initial endowments in such banks, and not invest in real investments on their own. They would also withdraw early if and only if they are early diers. Section III 9One may think of a 40 year period as made up of ages 40 to 80, so that a (disabled) individual may take early retirement at the age of 60, or continue working until 80 and then enjoy an increased level of consumption, with both dying soon after retirement. 10 In thinking of VB as the outside option of the currently middle-aged, when the next generation proposes a transition to SM, we are implicitly assuming that intragenerational solidarity among the middle-aged is present (or, more plausibly, that times of death are unresolved for them) when the next generation proposes. 11 In contemplating these transitions, we are not allowing the middle-aged generation to do both the constrained trading with the young, and also further within-generation risk-sharing as in (B); otherwise, VSM(MA) would always exceed VB. Our rationale is that we think of the SM-mechanism as being more individually decentralised than either B or TIB, which require intragenerational coordination at least. If, on the other hand, the middle-aged could renegotiate (B) among themselves, the economy would shift to SM (following a proposal from the new born) whenever ( ) VNB V SM TIB >. Section IV. 12 For example, generation 1 could offer to buy ε units of real, one-period old investments from generation 0 at time t = 1, at unit price, and invest the remainder of its endowment (1-ε) in real investments, thus guaranteeing its early and late diers consumption levels of {C1,1 = R , C1,2 = R }. Generation 0 would not have a better outside option (as this would involve physical liquidation) and generation 2 would be strictly worse off than generation 1. 13 In other words, we are analyzing cases in which the new born at t do not “go along with” the middle aged, on {TIB, TIB}, just to acquire Ht+1 > 1 to do better in a move to SM, i.e. for [VSM(MA) Ht+1 ] , at the next stage (t+1) game. Section V 14 In common with our model and Esteban and Sacovics [1993], Boldrin and Rustichini [1995] also assume that a proposed social security transfer to the old embodies in it the same proportional transfers to the current young when they would be old (unless this is then modified by a new proposal by the then young agents).
xl REFERENCES Allen, F. and D.Gale, 1995, A welfare comparison of intermediaries and financial markets in Germany and the US, European Economic Review, 39, 179-209. Bencivenga, V. and B. Smith, 1991, Financial intermediation and endogenous growth, Review of Economic Studies, 58, 195-209. Bhattacharya, S., P. Fulghieri and R.Rovelli, 1997, Financial Intermediation versus Stock Markets in a Dynamic Intertemporal Model, Discussion Papers Series, CEPR (London, UK). Forthcoming. Bhattacharya, S. and D. Gale, 1987, Preference shocks, liquidity and central bank policy, in: W.A. Barnett and K.J. Singleton, eds., New Approaches to Monetary Economics (Cambridge University Press, Cambridge, UK). Bhattacharya, S. and J. Padilla, 1996, Dynamic banking: a reconsideration, Review of Financial Studies, 9, 1003-1032. Boldrin, M. and A. Rustichini, 1995, Equilibria with social security, Unpublished. Bryant, J., 1980, A model of reserves, bank runs, and deposit insurance, Journal of Banking and Finance, 4, 335-44. Diamond, D.W. and P.H. Dybvig, 1983, Bank runs, deposit insurance, and liquidity, Journal of Political Economy, 91, 401-419. Dutta, J. and S. Kapur, 1994, Liquidity preference and financial intermediation, Discussion Paper in Economics, No. 17, Birkbeck College (London, UK). Esteban, J. (1986), A characterization of the core of Overlapping Generations economies, Journal of Economic Theory, 39, 439-456. Esteban, J. and T. Millan (1990), Competitive equilibria and the core of Overlapping Generations economies, Journal of Economic Theory, 50, 155-174.