信用额度、银行存款还是CBDC——现代支付系统中的竞争与效率-36页_269kb.pdf

Credit lines, bank deposits or CBDC?Competition & efficiency in modern payment systems*Monika PiazzesiStanford & NBERMartin SchneiderStanford & NBERMarch 2020AbstractThis paper studies the welfare effects of introducing a Central Bank Digital Currency(CBDC). Its premise is that CBDC is a new product in the market for liquidity where itcompetes with both commercial bank deposits and credit lines used for payments. If thecentral bank offers CBDC but not credit lines, then it interferes with the complementaritybetween credit lines and deposits built into modern payment systems. As a result, increas-ing CBDC may reduce welfare even if the central bank can provide deposits more cheaplythan commercial banks.*Email addresses: piazzesistanford.edu, schneidrstanford.edu. An earlier version of this paper was pre-pared for the June 2019 Annual Macroprudential Conference in Eltville, under the title Central bank-issueddigital currency: The future of money and banking? We thank Nobuhiro Kiyotaki, Narayana Kocherlakota,Guido Lorenzoni, Lorenzo Rigon, Carolyn Wilkins and conference participants for helpful comments.11 IntroductionShould central banks provide reserve accounts to everyone? A number of concrete proposalsfor central bank digital currency (CBDC) are now being discussed by policy makers as well asthe general public. For example, the governor of the Swedish Riskbank has put the probabilityof issuing an e-krona within the next decade at greater than 50%.1 Moreover, in the June2018 Vollgeld referendum, Swiss voters assessed (and, for now, turned down), a proposal tointroduce non-interest paying CBDC. Along with this general interest, an emerging literatureis weighing the pros and cons of CBDC.This paper is a theoretical study of the welfare effects of CBDC. Our approach is to viewCBDC as a new product in the market for liquidity, broadly defined: CBDC competes not onlywith deposits, but also with credit lines offered by commercial banks for liquidity purposes.To illustrate the importance of credit lines for payments, Figure 1 plots deposits at all UScommercial banks together with those banks outstanding credit card limits. Since credit cardlimits provide only a lower bound on credit lines used for payments for example, the figureleaves out credit lines provided to asset management firms by their custodian banks themessage is that credit lines matter.Our focus on the broad market for liquidity leads to a skeptical assessment of CBDC. Inparticular, introducing CBDC would interfere with current payments technology that exploitscomplementarities between deposit taking and credit lines. Most CBDC proposals imply theemergence of a system with two types of banks: commercial banks that offer both deposits andcredit lines, and a central bank that offers only the former. Our model shows that such hybridsystems entail costs that are not present in the current system when banks jointly providecredit lines and deposits. As a result, introducing CBDC may lower welfare even if the centralbank can offer deposits at lower cost than commercial banks.Our conclusions are derived from three assumptions. First, banks face leverage constraints:to be credible as payment instruments, deposits require backing by bank assets. Second,holding assets inside banks, or firms more generally, is costly: delegation of asset holdingsgives rise to asset management costs. Finally, the provision of payment instruments eithercredit lines or deposits requires banks themselves to have access to liquidity to executecustomers payment instructions. A good payment system minimizes costly asset holdingsthat are required as collateral to back promises or as liquid assets to meet liquidity needs.Banks that jointly offer credit lines and deposits economize on both collateral and liquidassets. Indeed, when a customer makes a payment by drawing down a credit line, the bank-1 GDPdepositscredit card limitsFigure 1: Deposits and credit card limits at US commercial banksing system creates a matching deposit account. The loan serves as collateral for these newdeposits. At the same time, no liquid assets are needed to handle the payment instruction:the bank creates liquidity on its books. If instead deposits and credit lines are provided byseparate banks, then more assets are needed to facilitate payments: loans have to be fundedand deposits have to be backed. Moreover, banks that provide credit lines need to hold liquidassets to manage deposit outflows that result from customer payments to banks that provideonly deposits.Our model builds on the standard neoclassical growth model. We add frictions that requirepayment instruments supplied by banks as well as collateral constraints and balance-sheetcosts. We show that a model with those frictions is equivalent to a model with a less efficientproduction technology. A switch to the wrong payment system thus works like a negativetechnology shock: it has real effects on consumption, investment and the allocation of laborthat lower welfare. We also add idiosyncratic shocks to preferences and production that makethe liquidity needs of an individual agent not perfectly predictable.We then compare three types of payment system. We first show that a system in which3banks jointly provide credit lines and deposits is superior to one where they only offer de-posits. The gain from credit lines is especially large if many agents have liquidity needs thatare difficult to predict. The advantage of credit lines is that assets on bank-balance sheetsand the associated costs only reflect actual transactions. Deposits held by agents with un-predictable liquidity needs instead have a precautionary component that requires more assetsand hence costs. The gain from credit lines is also larger if there is greater liquidity demandfrom agents who are not natural savers, for example, firms that themselves incur balance sheetcosts.To assess CBDC, we consider a hybrid system where the central bank offers deposits, andcan do so more cheaply than commercial banks. We show that such a system is superioronly if the cost advantage of the central bank is sufficiently large. As commercial bankscompete with the central bank for funding, they reduce their supply of deposits. As a result,credit lines become more expensive. They nevertheless continue to be used since they are stillcheaper than deposits for some bank customers. If the benefit from cheaper deposits is small,it is outweighed by the higher cost of credit lines and welfare declines. This force is strongerif liquidity needs are harder to predict as well as when there is more liquidity demand fromagents who are not natural savers.Our model is motivated by a liquidity-centric view of banking. In fact, the only way inwhich banks add value in our model is by providing liquidity. In particular, they do nothave a special ability to lend, except by extending on-demand loans when credit lines aredrawn. The role of other bank assets in the model is only to provide backing for deposits.A liquidity-centric view fits well with evidence on bank portfolio composition. Indeed, inmost countries banks hold not only loans, for which they might have a special ability to lend,but also securities. Moreover, a sizeable share of loans tends to be mortgages that are easilysecuritized. In our model, it makes sense for banks to hold securities even though they areworse at holding them than households, because securities help back deposits and thus deliverpayment services.We restrict technology and preferences so that banks required collateral is small relative tothe capital stock of the economy. In other words, it is never important to accumulate capital inorder to provide enough collateral to back payment instruments. Instead, society determineswhat share of capital is optimally held inside banks and holds the rest outside. This approachalso fits well with data on sectoral wealth in modern economies. Indeed, banks typically holdonly a share of fixed income securities outstanding in the economy, with another sizeable shareof both government and private debt held either directly by households or indirectly throughinvestment intermediaries such as pension funds. Moreover, in most countries, business andhousing equity are held almost entirely outside banks.4We emphasize that our assumption that bank lending is small relative to capital does notmean that banking is irrelevant for investment. In our model, capital accumulation dependson the cost of payments because capital-goods producers require liquidity in order to produce.Our assumption on balance-sheet costs imply that capital-goods producing firms find it par-ticularly burdensome to hold deposits and prefer credit lines. Interpreting capital broadly asphysical plus intangible capital, we view this feature as a stand-in for the demand for paymentservices from nonbank financial institutions that typically favors credit lines over deposits. Theimplication is that introducing CBDC can have a stronger negative effect on investment rela-tive to consumption since it distorts effective prices faced by firms more than those faced byhouseholds.A premise of our analysis is that it is beneficial for society to minimize the amount ofassets held inside banks and firms. Here we build on a large literature that has discussedthe costs of delegated portfolio management. At the same time, delegated asset managementmay also have benefits, for example cheaper diversification or savings of transactions costs.Our approach assumes that such benefits can be achieved more easily through investmentintermediaries such as mutual funds that are funded with equity. They do not require assetholdings inside banks that also issue debt or firms that also engage in production. More-over, the delegated monitoring problems that arise in leveraged banks and producing firms which are arguably more complicated than those of investment intermediaries induce coststhat outweigh any benefits that can be realized through investment intermediaries. It is thusoptimal to minimize assets inside banks and firms, and to think of the household sector asconsolidated with investment intermediaries.To zero in on the key interaction of credit lines and deposits, our model abstracts from anumber of other interesting considerations on CBDC. First, we define CBDC narrowly as adeposit contract and do not consider the option of anonymity that would make CBDC closerto physical currency. In terms of the money flower taxonomy of monies introduces by Bechand Garratt (2017), we study CBDC that is widely accessible (as opposed to restricted) as wellas account-based (as opposed to token-based). In fact, we abstract from physical currencyaltogether and require that all payments are made with deposits or credit lines. As a result,we do not engage in the discussion of how CBDC might alter a potential lower bound oninterest rates.Second, we study an entirely real model and do not consider the determination of the price5level or how the transmission of monetary policy might change if a CBDC is introduced.2 Thisapproach is guided by our focus on long-run welfare from the design of the payment system.We thus formulate policy as the elastic supply of real balances at a certain spread between theinterest rate on CBDC and a safe claim that is not liquid. In practice, one would expect thecentral bank to fix both the price and quantity of nominal CBDC. In the long run, the pricelevel would then adjust to deliver the quantity of real balances desired by the economy.Third, we work with frictionless capital and insurance markets. In particular, householdshave access to a complete set of contingent claims to insure against preference shocks andbanks can issue equity at no cost at all times. In addition to making the model analyticallytractable, these assumptions also clarify that our mechanism does not rely on net-worth con-straints in banks. Moreover, banks in our model are not needed to facilitate consumptionsmoothing when assets are illiquid, as in the literature that studies banks as a mechanismfor liquidity insurance. Their purpose instead is to provide immediacy of payments in goodsmarket transactions that financial markets cannot provide. Future work might explore theinteraction of our mechanism with other financial frictions.Finally, we do not explicitly model credit risk and aggregate shocks. All lending is de-terministic and there is no default. While this simplification precludes us to talk about someinteresting features of the data such as risk premia or credit spreads, the loss for our study ofbalance-sheet positions is smaller. In particular, one interpretation of balance-sheet costs in-curred by banks is bankruptcy costs in lending. Suppose for example, that banks hold claimsto capital not by owning capital directly, but by making loans that carry idiosyncratic risk anddeadweight costs of default. On average, banks then expect to lose a certain share of the returnon capital. Households who hold capital directly or through investment intermediaries do notincur the same costs since they hold equity claims.Our approach builds on the theoretical and empirical literature on liquidity provision.Strahan (2010) provides an overview and discusses the trend towards greater use of creditlines. Berger and Sedunov (2017) argue that including off-balance sheet measures such asloan commitments is important to measure the role of banks in an economy and providecross-country evidence. Sufi (2007) shows that credit lines help firms avoid costs of holdingcash. Consistent with this idea, our model builds in the assumption that firms have a strongerpreference for credit lines over deposits, since they are not natural savers.2Piazzesi, Rogers and Schneider (2019) study the transmission of CBDC in a New Keynesian model. Theyargue that the transmission of interest-rate policy works is similar a floor system as currently implemented bymany central banks, but different from traditional corridor systems because of a lower elasticity of broad moneysupply. Barrdear and Kumhof (2016) also study CBDC in a New Keynesian setup and derive welfare gains froma reduction in transaction costs.6Holmstrm and Tirole (1998) show how credit lines can help allocate liquidity when in-dividual agents needs for liquidity are not perfectly predictable, an important theme in ouranalysis. Kashyap, Rajan and Stein (2002) provide theory and evidence on complementaritybetween credit lines and bank deposits. They show that, at the level of the individual bank,liquidity management is cheaper when liquidity needs that is, outflows of funds implied bythe two products are imperfectly correlated. They provide supporting cross-sectional evidenceon banks selection of products to offer (see also Gatev, Schuermann and Strahan (2009).In our model, complementarity is due to both collateral and liquidity constraints. In fact,in the version of the model without a central bank, liquidity management plays no ro