|
Article Excerpt
WHEN TRADE IS decentralized, and agents are not completely reliable, collateral is useful for effecting exchange. However, collateral is expensive and can be subject to shortages, temporary or permanent. When shortages occur, spot transactions--transactions of goods against collateral--are not always possible. In such a case, payment systems become relevant. (1) As the exchange of goods becomes temporally separated from the exchange of collateral, payment systems become a means of economizing on the amount of collateral needed to enforce the pledges made.
The world has seen a variety of payments mechanisms, under both public and private arrangements, centralized and decentralized. What are the optimal arrangements for such payment systems? What are the mechanisms a payment system employs to economize on collateral? This paper investigates an environment in which we can begin to address these questions. We start by analyzing the development of a private arrangement for effecting payments. We consider the questions of membership in the organization, as well as requirements for monitoring and for the posting of collateral. We then consider a central-bank-sponsored payments arrangement, where the central bank can exploit its taxation powers to back up its promises. We examine the advantages and disadvantages of including the central bank in the payments mechanism, and how its presence causes a system to change.
Our analysis is most relevant for the design of "large-value" or "wholesale" payment systems. Traditionally, these systems are used to settle obligations between banks, as may arise from large-value commercial transactions, from the operation of "small-value" or "retail" payment systems (e.g., checks and credit cards), or from the need to settle financial market transactions. Fedwire (operated by the Federal Reserve) and CHIPS (operated by the New York Clearing House) are the two preeminent large-value payment systems in the United States. (2)
Large-value payment systems typically have a tiered structure. At the core of these systems is either a single institution (a central bank in the case of Fedwire and similar systems) or a relatively small group of institutions (as in the case of CHIPS) with special settlement privileges. A second level of the hierarchy has institutions that may access the system but with restrictions on access such as position limits or additional collateral requirements. At the bottom of the hierarchy are institutions that are not members of the system or are customers of member institutions, both of which clear and settle payments through member institutions. A principal focus of our analysis will be to understand the purpose of this tiered structure.
In ordinary circumstances, the operation of public and private payment systems is, at a practical level, much the same. There are, however, some notable differences in the legal and institutional underpinnings of public versus private arrangements. These distinctions, largely inconsequential during normal times, can give rise to critical distinctions in the functionality of the two types of system during times of duress. The first such distinction deals with "finality." A funds transfer over a public system typically represents an unconditional transfer of a claim on a central bank. As such, it unconditionally discharges an obligation: payment may effectively be thought of as settlement. For example, in the United States, a funds transfer over the Fedwire system automatically and immediately becomes a liability of the Federal Reserve and is virtually always final. Payments made over private systems may not carry the same degree of finality. (3)
A second distinction between public and private systems can arise in the area of credit policy. Intraday credit is an essential component of virtually all large-value payment systems. (4) The demand for such credit largely arises from payment system participants' inability to coordinate incoming and outgoing payments. (5) Both private and public payment systems may grant intraday credit, and in both types of systems, credit risk is commonly controlled through such devices as position limits, monitoring (e.g., bank supervision), and collateral requirements. Membership in a public payment system, however, necessarily carries with it a form of "credit insurance" that has no analog in a private system; that is, while a central bank may limit the availability of intraday credit during normal times, it cannot credibly commit to withhold credit during times of duress. In a crisis, a central bank will always be tempted to enable the settlement of ex post welfare-improving trades. (6) This can lead to the central bank granting credit in circumstances where, during normal times, the granting of such credit would lead to unacceptable level of credit exposure.
In our analysis of private payments arrangements, we consider the effects of three devices in the enforcement of settlement obligations: netting, monitoring, and collateral. Both netting and monitoring are ways of economizing on scarce or expensive collateral. For cost of monitoring sufficiently low, we show that these devices should be applied in roughly in that order. Netting alone will be adequate if all counterparties are known to be sufficiently reliable. If some of the counterparties involved may be too undependable, monitoring of these counterparties may be necessary. This monitoring can be enforced by a tiered structure, requiring that unreliable parties settle through a more reliable member agent. For still less reliable counterparties, posting of collateral will be necessary to ensure settlement.
In most situations these three devices, when combined with sufficient availability of collateral, will enable agents to organize trade efficiently. But as emphasized by Kocherlakota (2001), the efficacy of these devices is ultimately tied to the value of the collateral good. If there is a downward shock to collateral value, then trade may break down even under net settlement, because net settlement provides no inducement to deliver goods when there is no incentive to pay for goods received. Settlement on the books of a central bank, by contrast, can provide an incentive to deliver because the value offered in exchange does not derive from the value of collateral, but instead the taxation powers of the government. As long as the central bank makes credit freely available (and ex post, the central bank will have an incentive to grant such credit), confidence in the value of central bank liabilities will be sufficient to sustain trade. And to the extent that obligations incurred by payment system participants are offsetting, the central bank in equilibrium bears no loss.
The liquidity provided by public payment systems has a downside, however. Although confidence in the liabilities of the central bank can sustain trade during crises, that same confidence can undermine the incentives of payment system participants for mutual monitoring. This is of concern if one believes the public sector is worse at monitoring, or is less inclined to act upon the basis of information received. As a result, this disadvantage must be given due weight in the consideration of the relative merits of public versus private systems. A policy of limiting central bank settlement facilities to a select group of agents may be seen as an informed compromise between the informational superiority of purely private systems and the greater liquidity of their public counterparts.
1. LITERATURE SURVEY
Central to the analysis below is the notion of delegated monitoring (Diamond 1984), specifically, the delegation of monitoring within a payment system. The study of strategic interactions that may arise between monitoring incentives on the one hand, and the settlement of outstanding obligations on the other, was introduced by Rochet and Tirole (1996) (RT), and has been extended by Fujiki, Green, and Yamazaki (2008) (FGY). RT show how a central bank's "too-big-to-fail" (TBTF) policy may dilute banks' incentives to monitor their exposure with other banks, including those that may arise in a payments context. FGY, by contrast, are concerned with providing a more fundamental justification for public sector involvement in the payment system. To this end, they show that such involvement can be seen as a feature of (generalized) core allocations in an economy that incorporates private information and other trading fractions. Our approach is generally closer to that of RT in the sense that we will take certain aspects of public sector involvement in large-value payment systems (including TBTF) as parametric; we then consider potential interactions of such involvement with monitoring incentives. Details of the model environment are closer to FGY, however, in the sense that monitored information may be conveyed by explicit reports, and does not have to be inferred from observed behavior.
Another element of our analysis is the notion that limited enforcement frictions can sometimes be overcome by substituting public obligations for private ones. This idea is by now a very familiar one, following such papers as Woodford (1990), Holmstrom and Tirole (1998), Kocherlakota (2001), and Koeppl and MacGee (2001). What is different below is that we explore how such substitution may relate to the efficacy of other devices for overcoming limited enforcement, as are commonly employed in large-value payment systems.
By virtue of its subject matter, our paper is also connected to many other papers in the burgeoning literature on the design of payment systems. Two of the most relevant papers are Freixas, Parigi, and Rochet (2000) (FPR) and Holthausen and Ronde (2002) (HR). FPR analyze the interplay between patterns of settlement obligations, their potential for creating systemic risk, and the efficacy of various policy interventions designed to prevent its spread. Our setup is similar in the sense that certain alignments of preference shocks (and their resulting settlement obligations) can give rise to scenarios with an elevated potential for settlement failures. Our analysis, however, is less concerned with the desirability of public sector bailouts per se. Instead, we focus on how bailouts may interact with other means for lessening settlement risk, in particular, tiering arrangements.
HR also look at issues of membership in large-value payment systems that utilize net settlement. They show that under limited liability, there exists an incentive for overly broad membership in these systems since member institutions may not internalize the costs of potential settlement failures. Nonetheless decisions on membership in these systems may best be left up to the private sector, if the private sector enjoys a significant informational advantage over regulators. Our analysis yields a complementary inference: regulators may wish to exploit the private sector's informational advantage by restricting access to public settlement systems (see the discussion in Section 7 below).
More recent papers have advanced the analysis of tiering to dynamic and computational environments. Chapman and Martin (2007) extend the Freeman (1996) model of payments to show that tiering may improve the efficiency of central bank liquidity provision. Chapman, Chiu, and Molico (2008) develop a dynamic model in which tiering allows for efficiency gains through revelation of agents' trading histories. An empirical model of endogenous tiering is presented in Adams, Galbiati, and Giansante (2008), based on data from the CHAPS (UK) payment system.
2. THE MODEL
There are three time periods 0, 1, and 2; agents establish a payment system in period 0, trade in period 1, and settle and consume in period 2. There are a large and equal number of agents of three types: A, B, and C (below we will sometimes use A to mean "an agent of type A," etc.). We will regard these agents as centralized during periods and 2, but separated on different "islands" during the trade period. Conceptually, it is most natural to imagine that each trader is a two-agent household and that at the trading round the two agents separate, one meeting with an agent of each of the other two types. Each agent is also a potential member of various payment arrangements, which will be described in more detail below. (7)
There are four goods: an indivisible endowment good unique to each type (goods A, B, and C) and a collateralizable, numeraire good (good N). Each agent is endowed with one unit of his type's good (collectively known as "eponymous" goods) and L units of the numeraire good.
The numeraire will always be a desirable consumption good for all agents. Numeraire good can only be transferred to another agent in period 2. Each of the eponymous goods is also a potential consumption good. Agents always value their own eponymous good at [alpha], relative to numeraire, where < [alpha] < 1. Agents' preferences over others' goods are determined by preference shocks whose structure we describe below. The eponymous goods...
|
