FinancialSystemArchitecture ArnoldW.A.Boot;Anjan V.Thakor TheReview ofFinancialStudies,Vol.10,No.3.(Autumn,1997),pp.693-733. StableURL: http://links.jstor.org/sici?sici=0893-9454%28199723%2910%3A3%3C693%3AFSA%3E2.0.CO%3B2-4 TheReviewofFinancialStudiesiscurrentlypublishedbyOxfordUniversityPress. YouruseoftheJSTORarchiveindicatesyouracceptanceofJSTOR'sTermsandConditionsofUse,availableat http://www.jstor.org/about/terms.html.JSTOR'sTermsandConditionsofUseprovides,inpart,thatunlessyouhaveobtained priorpermission,youmaynotdownloadanentireissueofajournalormultiplecopiesofarticles,andyoumayusecontentin theJSTORarchiveonlyforyourpersonal,non-commercialuse. Pleasecontactthepublisherregardinganyfurtheruseofthiswork.Publishercontactinformationmaybeobtainedat http://www.jstor.org/journals/oup.html. EachcopyofanypartofaJSTORtransmissionmustcontainthesamecopyrightnoticethatappearsonthescreenorprinted pageofsuchtransmission. TheJSTORArchiveisatrusteddigitalrepositoryprovidingforlong-termpreservationandaccesstoleadingacademic journalsandscholarlyliteraturefromaroundtheworld.TheArchiveissupportedbylibraries,scholarlysocieties,publishers, andfoundations.ItisaninitiativeofJSTOR,anot-for-profitorganizationwithamissiontohelpthescholarlycommunitytake advantageofadvancesintechnology.FormoreinformationregardingJSTOR,[email protected]. http://www.jstor.org ThuMar1302:40:542008 Financial System Architecture Arnoud W. A. Boot University of Amsterdam Anjan V. Thakor University of Michigan This article builds a theory offinancial system architecture. We ask: wbat is a financial mar- ket, wbat is a bank, and what determines the economic role of each? Starting with basic assumptions about primitives-the types of agents and the nature of informational asym- metries-we provide a theory tbat explains which agents coalesce tof orm banks and which trade in the capital market. It is shown tbat bor- rowers of higher observable qualities access thef inancial market. Moreover, a financial sys- tem in its infancy will be bank-dominated and increasedfinancial market sophistication dimin- ishes bank lending. A primary function of the financial system is to fa- cilitate the transfer of resources from savers ("surplus A. V. 'rhakor would like to thank the Edward J. Frey Chair in Hanking and Finance for financial support. A. W. A. Hoot tlranks the Olin program in Law and Economics at Cornell IJniversity for its hospitality during part of the re- search on this article. The authors would also like to thank Todtl Milbourn, Kathleen Petrie, and Anjolein Schmeits for excellent research assistance, and seminar participants at the University of Michigan, Indiana University, IJniversity of Amsterdam, Erasmus IJniversity, Rotterdam (the Netherlands), IJnivrrsity of Minnesota, the JFI Symposium on Market Microstructure and the Ilesign of Financial Systems at Northwestern University (May 19951, the Nordic Finance Symposium at Vendsnu, Norway (February 1995), Queen's IJnivrrsity, Cornell IJniversity, the 1.ondon School of Economics, the Stock- holm School of Economics, the University of Goteborg (Sweden), McGill University (Canada), and participants at the CEPR meetings in St. Sehastian, Spain (April 19941, and Gerzensee, Switzerland (July 1994), and the Amer- ican Finance Association meeting, San Francisco wanwary 1996) for help- ful comments. The authors are particularly indebted to Ed Kane, Sudipto Bhattacharya, Paolo Fulghieri, Neil Wallace, Mike Stutzer, Franklin Allen (the editor), and an anonymous referee for helpful suggestions. Address correspondence to A. V. Thakor, University of Michigan Business School, 710 Tappan St., Ann Arbor, MI 48109-1234. ?'he Reuiew of'Financiul Studie.~P all 1997 Vol. 10, No. 3, pp. 69.+733 @ 1997 The Review of Financial Studies 0893-9454/97/$1,50 ?'he Review of Financial Studies/v 10 n 3 1997 units") to those who need funds ("deficit units"). In a well-designed financial system, resources are efficiently allocated. The question we address is, what is the configuration of such a financial system? In particular, we examine why bank lending and capital market financing coexist and the factors-such as regulation and the stage of economic development-that determine which dominates. These issues are important for many of the current policy debates regarding the structuring of financial systems. How do banks and cap- ital markets emerge and evolve? What services should be provided by banks and what services by the capital market? How is the resolu- tion of informational problems related to how the financial system is configured? These questions are particularly interesting in the context of Eastern European countries. The financial systems currently in place there are interim arrangements to facilitate transition to systems with lesser em- phasis on the central planning of capital allocation [Checchi (1993)l. Although reform discussions have focused largely on the creation of financial markets [Mendelson and Peake (199311, the more spectacular initial developments are likely to be in banking. For example, privately owned commercial banks were uncommon in Communist Europe un- til recently. Since then, however, banks have evolved rapidly [Perotti (1993) and Van Wijnbergen (1994)l. These developments point to a key aspect of financial system architecture: the determination of the roles of the banking system and the financial market. Despite its importance, the research on this topic is still only emerg- ing. Allen (1993) provides a qualitative assessment and sketches a preliminary framework for analysis. That article links financial sys- tem design to the complexity of decision making within firms seeking capital and provides a perspective on the disparate evolutions of fi- nancial markets in Europe and the United States. Hhattacharya and Chiesa (1995), Ikwatripont and Maskin (19 95), von Thadden (1995), and Yosha (1995) examine the comparative allocational efficiencies of "centralized" (bank-oriented) credit markets versus "decentralized" (market-oriented) credit markets. Somewhat different approaches are taken by Allen and Gale (1995, forthcoming) who suggest that bank- oriented systems provide better intertemporal risk sharing, whereas market-oriented systems provide better cross-sectional risk sharing, and Sabani (1992) who argues that market-dominated economies will restructure financially distressed borrowers less than bank-dominated economies. These contributions notwithstanding, there are unanswered ques- tions. For example, how is the informativeness of market prices af- fected by financial system design? If unfettered by regulation, what determines the design of the financial system? And how does this de- Financial System Architecture sign affect the borrower's choice of financing source? Does financial system design have real effects? This article is a modest first attempt to address these issues. We explain how financial institutions and markets form and evolve when economic agents are free to choose the way they organize themselves. Rather than taking the roles of institutions and markets as given and then asking how borrowers make their choice of financing source [e.g., Berlin and Mester (1992), Besanko and Kanatas (1993), Iliamond (1991), and Chemmanur and Fulghieri (1994)1, we start with assump- tions about primitives-endowments and informational frictions-and endogenize the roles of banks and financial markets. The distinction we make between a bank and a market is that agents within a bank can cooperate and coordinate their actions, whereas agents in a market compete;' we assume nothing more about what banks and markets do. We begin by positing three types of informational problems: (i) in- complete information about future projects that is of relevance for firm valuation and real investment decisions within firms, (ii) postlending (asset substitution) moral hazard that can affect payoffs to creditors, and (iii) uncertainty about whether postlending moral hazard will be encountered. Part of the primitives are economic agents who special- ize in resolving these informational problems, with each individual agent being atomistic in impact. Our first major result is that problem (i) is most efficiently resolved in an "uncoordinated" market setting where individual agents compete with each other, and problems (ii) and (iii) are most efficiently resolved through coordinated action by agents coalescing to form a bank. The scope of banking vis-a-vis the financial market is thereby determined endogenously in an unregu- lated economy in which the financial market is characterized by many agents and a rational expectations equilibrium price formation process that noisily aggregates information contained in the order flows for securities. A key attribute of the financial market, and one that delin- eates its role from that of a bank, is that there is valuable information feedback from the equilibrium market prices of securities to the real decisions of firms that impact those market price^.^ This information loop provides a propagation mechanism by which the effects of finan- ' Perhaps an even more basic distinction is that agents can he anonymous in a market but not in institutions. This may be a way to rationalize the possibility of coordination within a hank and the lack of it in an anonymous, competitive market setting. Allen (1993)suggests that an important role of the stock market is to provide decision-makers in firms with information they would not otherwise have possessed. Holmstriim and Tirole (1993) examine the role of the stock market as a monitor of managerial performance. They show that a Finn's stock price incorporates performance information that cannot be gleaned from the firm's current or future profit data, and that this information is useful in stnicturing managerial incentives. The Review of Financial Studies/v 10 n 3 1997 cia1 market trading are felt in the real sector. Hank financing does not have such an information loop. Hence, real decisions are not impacted t,y the information contained in bank credit contracts. However, banks are shown to be superior in resolving asset substitution moral hazard. Thus, in choosing between banks and financial markets, one trades off the improvement in real decisions due to feedback from market prices against a more efficient attenuation of moral hazard. The relative levels of credit allocated by banks and the financial market depend on the efficacy of the bank's monitoring and the "de- velopment" (i.e., sophistication or level of financial innovation) of the financial market. We let the latter be reflected in the information ac- quisition cost for those who wish to become informed. We show that the cost of information acquisition affects the informativeness of equi- librium security prices, and therefore the relative scopes of banks and the financial market in credit allocation. In describing these scopes, our article explains: why banks emerge even when every agent in a bank could trade on his own in the market; why financial markets develop even when there are no restric- tions on banks' activities; why a financial market equilibrium in which prices convey infor- mation can exist only if prices do not have too much or too little informati~eness;~ why borrowers prefer either the financial market or banks based on differences in observable borrower attribute^;^ how financial market trading affects firms' real decisions; how the state of development of the financial market can impact the borrower's choice of financing source. The rest of the article is organized as follows. Section 1 contains a description of the basic model. Section 2 analyzes the formation of banks and the financial market. Further analysis is contained in Sec- tions 3 and 4. Section 5 examines model robustness issues. Section 6 explores the implications of the analysis for financial system design. Section 7 concludes. All proofs are in the Appendix. ' This is in contrast to the existing literature in which the value of information acquisition is non- decreasing the noisiness of the process hy which information is aggre~ated[e.g.,Grossman and Stiglitz (1980)l. Since banks resolve moral hazard in our model, the bank's decision to grant a loan does not trigger an ahnormally positive reaction in the borrower's stock price as found empirically by James (1987), Lummer and McConnell (1989), and Shockley and Thakor (1996, forthcoming). Of course, if our model were to be altered to introduce uncertainty about whether the borrower would have a project available, then the hank's decision to grant a loan would signal good news. See Hoot and Thakor (1996, forthcoming) for such an approach. Financial Sy.~temArchitecture Figure 1 A Schematic of the types of projects (without payoff enhancement) and contractible returns 1. The Basic Model 1.1 Production possibilities for firms 1.1.1 Preferences and types of projects. There is universal risk neutrality, and the riskless interest rate is zero. The economy consists of firms each with a project that needs a $1 investment. As shown in Figure 1, each firm has a stochastic investment opportunity set that contains two projects: good and bad. The contractible end-of- period return for the good project has a probability distribution with a two-point support: with probability r] the end-of-period return will be Y > 0, and with probability 1 - r] it will be 0. The contractible end-of-period return for a bad project will be 0 with probability 1, but this project offers the borrowing firm's manager a noncontractible private rent, N, from investing in the project [see, e.g., O'Hara (199311. Let rj Y > N, so that the borrower prefers the good project with self- financing. 1.1.2 Project availability and payoff enhancement possibility. Project availability is stochastic. With probability 8 E @,8) c (0,I), the firm finds itself in the "low flexibility" (LF) state in which it has only the good project availal~leW. ith probability 1 8, the firm finds - itself in the "high flexibility" (HF) state and has both the good and the bad project available. The Retiiew ofFinancial Studies// 10n 3 1997 Borrower - - 1 Low Flexi ility State / High Flexibility State \ JBadProejc lt Figure 2 A schematic of the typesof projects (with payoff enhancement) and contractible returns We assume that the firm can possibly enhance the return of the K K]. good project at a private cost of K = > 0, where K E {O, This investment is unobservable to outsiders, and it enhances the project return l~ya E (0, I), conditional on a favorable realization of an "environmental" or "market" random variable v E (0, 11,w~ith Pr(v = 1) = y E (0, 1) as the proba1,ility that the a priori uninformed assign to the event that v = 1. Let a > K.Note that v is specific to each firm rather than being an economywide variable. Thus, if v = 1 + and the firm invests K = i?, then the good project pays off Y a with probability q and a with probability 1 -q. If a borrower invests K = 0, then the good project's return is Y with probability q and 0 with probability 1 - q, regardless of v. If a borrower invests either K = K or K = 0 in a bad project, the contractible project return is 0 with probability 1, regardless of v.Thus, the improvement in the project return depends on borrower-specific investments as well as the realization of exogenous uncertainties like market demand (see Figure 2). It does not matter much if we assume that K is ohsewable to outsiders. With K unobservable, the firnr underinvests relative to first best, whereas with K ohsrrwble, there is no underinvest- ment. Financial System Architecture We assume that the realization of v can only be observed by those who become informed at a cost; we will say more about this later. For now, it suffices that the borrower cannot observe v, but believes that Pr(u = 1) = y. This belief is common knowledge. If the borrower is uninformed about v, an investment K = I? > 0 is suboptimal. That is, we assume ya -I? < 0, where ya is the expected project enhancement if the borrower is uninformed about v. We assume q > y and that K ya < c ya. (1) Further, we assume that exogenous parameter values are such that there exists an interest factor (one plus the interest rate) rO< 1 satis- fying y[Y +a - rOl-l? = N. Given this, the firm prefers the bad project with external financing if the interest factor exceeds rO,and it prefers the good project with external financing if the interest factor is less than rOS.ince r0< 1 and we had earlier assumed that 7 Y > N, the firm always prefers the good project with self-financing and the bad project with external financ- ing. From the lender's standpoint, therefore, there is asset-substitution moral hazard only in the HF state. We view the parameter 8 as the commonly known prior probability assigned by the market to the event that a randomly selected borrower will be in the LF state and hence pose no moral hazard. Each potential borrower is character- ized by its observable 8 E (8,8).Let H be the cumulative distribution function over the cross-section of 8s. 1.2 'Qpes of securities We limit financiers to debt contracts. This is primarily because bank lending is typically done through debt contracts, and we want to have comparability between the bank and capital market financing cases. Thus, the capital market financing in our model is through bonds. Of course, information acquisition in bond markets is probably smaller than that in the stock market. It is therefore important to note that our analysis understates the information acquisition benefits of financial markets, but is qualitatively unaffected if debt is replaced by equity (see Section 5). 1.3 Sequence of events in lender-borrower interaction The firm first makes an irreversible decision about whether to borrow from a bank or the financial market. At this stage the only information that it has is about its 8, and this information is common knowledge.6 The Review ofFinancia1 Studie.7 / 10 n 3 1997 LJ t=o . Each firm's B is Firm learns whether it The firm borrows $1 The firm's project common knowledge. is in the LF or the HF to invest in its project. payoff is realized. state, after which v is Each firm makes an privately learned by The firm infers v Lenders are paid irrevocable decision informed traders. from its financial off if the project of whether to borrow market price and payoff permits it. from a bank or in the Different types of decides nhether to financial market. traders anonymously invest K or 0 in project submit their purchase payoff enhancement. Traders decide orders for the firm's . whether to become securities to the market If the firm is in the monitoring agents, maker in the financial HF state and the lender informed agents or market. can monitor the project remain uninformed choice, it will ensure discretionary traders. A price for the firm's that the firm chooses The informed agents debt is determined the good project. Each invest M to acquire either in the financial agent in the lender their information. market or by a hank. coalititon incurs monitoring cost M Banks are formed. regardless of the state the borrower is in. The financial market organizes for trading. Figure 3 A description of the sequence of events in the economy Subsequently, the firm learns whether it is in the LF or the HF state, and after this v is realized and learned by those who choose to become informed about it. The lender (either the bank or the financial market) offers a price of credit that the firm can either accept or respond to with a take-it-or-leave-it counteroffer. Moreover, based on the lender's actions (the offered credit price or the market demand for the firm's security in the financial market), the firm makes its inference about the realization of v. Next, the firm makes its initial choice of project if it is in the HF state; in the LF state, this choice is trivially the good project. If the lender can observe this initial choice and monitor, it can force a change to the good project in the event that the firm had initially chosen the bad project. This leads to the firm's final project choice decision and its investment of $1in the project. Moreover, at this time the firm also makes its decision regarding investing K E (0,I?) for project payoff enhancement (see Figure 3). We assume that the firrn commits to a financing choice at the outset to avoid the situation in which financial market investors produce information about a firm that ends up borrowing from ;I bank. Although in equilibrium each firm's choice of financing source is unambiguously linked to its 0. and this H is commonly known at the outset. W11;lt we wish to avoid is the firm lrarning about v from its market price (which is based on investors erroneously believing that the firm will horrow in the financial market) and then borrowing from a hank. Wnunciul System Architecture 1.4 vpes of agents in the economy The structure for the financial market is as follows.' There are two types of investors/traders in the market: liquidity traders and discre- tionary agents. The aggregate asset demand, t,of the liquidity traders is random and exogenously specified by the continuously differen- q, tiable probability density function f(t)= A - where A is a pos- itive constant. Thus, the support of f(t)is [O, 2/Al. A discretionary agent can become an "informed" or a "monitoring" agent at a finite cost M > 0. This investment M either generates a signal that perfectly reveals the v for the firm in question or enables the agent to monitor the firm's investment choice between the good and the bad project. The discretionary agent must decide before investing M whether she wishes to be an informed agent and receive the signal or become a monitoring agent. If the discretionary agent does not invest M, she can be an uninformed discretionary trader who can either invest in the capital market or in bank deposits. We will first focus on agents who become informed about v. Each submits a demand order dl. Let us conjecture that the equilibrium strategy of an informed trader is to set dI = 1if the signal says v = 1 and dl = 0 if the signal says v = 0; we will validate this conjecture later. Each trader is very small but oft > 0 Lebesgue measure on the real line. We will focus on the llmiting case in which t + 0 so that each trader is atomistic, and all traders lie in a continuum. Let !2 be the (Lebesgue) measure of informed traders, with each submitting a demand of 0 or 1. The total informed demand is therefore DI = Qd1. Liquidity traders' demand is not information driven and is based on exogenous factors outside the model. All demand orders are sub- mitted to a market maker, and Informed and liquidity traders are ob- servationally identical to the market maker. Thus, the market maker + observes only the total demand, D = Dl t, and not its individual components, DI and t.The supply of the (debt) security is fixed at $1. We assume that there is a sufficient number of "professional" market makers, so that the market is competitive. The market maker receives all the orders for a given security and takes the position in the security required LO clear the market at a price that yields her zero expected profit, conditional on the information in the order flow. Thus, the market maker takes a long position in the security if supply exceeds demand and a short position if demand exceeds supply. The debt security In question is a bond issued at par, and the price set by the market maker is the bond's coupon rate (or interest rate). ' Th is structure is similar to that in Boot and Thakor (1993a), hut richcr in that agents can also choose to monitor and there is information feedhack from thc financial market to thc firm. 701