The Risks of Bank Wholesale Funding

Transcription

1 The Risks of Bank Wholesale Funding Rocco Huang Philadelphia Fed Lev Ratnovski Bank of England April 2008 Draft Abstract Commercial banks increasingly use short-term wholesale funds to supplement traditional retail deposits. Existing literature mainly points to the virtues of wholesale funding: market discipline and lower liquidity risk. This paper models a "dark side" of wholesale funding. In an enviroment with a costless but imprecise signal on bank project quality (e.g. market- or sector-wide news), wholesale nanciers can choose not to conduct costly information acquisition and withdraw upon a hint of negative news ahead of passive retail depositors, triggering ine cient liquidations. The model sheds light on the recent credit crisis, explaining why wholesale nanciers did not provide market discipline ex-ante and exacerbated liquidity risks ex-post. The views expressed are those of the authors and do not necessarily represent those of the Philadelphia Fed or the Bank of England. Contact: rocco.huang@phil.frb.org, ratnovski@gmail.com. 1

2 1 Introduction Commercial banks increasingly use short-term wholesale funds in addition to traditional retail deposits to nance their long-term assets. The traditional deposits-based bank funding model has been under strain due to intense competition for household savings among banks and from alternative investment institutions (mutual funds, life insurance products, etc.). To supplement a reduced deposit base, banks increasingly tap into the wholesale money market to attract liquidity surpluses of other nancial institutions, non- nancial corporations, state and local authorities, and foreign entities. Such wholesale funds are usually raised on a short-term rollover basis with instruments such as fed funds purchased, repurchase agreements, large-denomination renegotiable certi cates of deposit, interbank deposits, commercial paper, etc. How does this change in funding structure a ect bank risks? The existing literature mainly points to the "bright side" of wholesale funding: exploiting valuable investment opportunities without being constrained by the local deposit supply; market discipline provided by sophisticated nanciers (Calomiris, 1999); re nancing of unexpected retail withdrawals for banks with established wholesale markets access (Goodfriend and King, 1998). However, recent credit market events revealed a "dark side" of wholesale funding. Banks can use wholesale funds to aggressively expand lending and compromise credit quality, particularly when nanciers exercise insu cient market discipline. Later, at the re nancing stage, there is a risk of wholesale nanciers suddenly withdrawing upon a hint of negative news, triggering ine cient liquidations. When wholesale withdrawals follow a market-wide signal, correlated bank failures may also exacerbate systemic risk. Why, in the recent credit turmoil, did wholesale funds not provide su cient market discipline ex-ante, and seemed to have exacerbated liquidity risk ex-post? This paper attempts to reconcile the traditional view on the virtues of wholesale funding with the recent credit crisis experience. We suggest that wholesale funding is bene cial when informed, but exacerbates ine ciencies and bank liquidity risks when uninformed. 2

3 The paper o ers a model that studies: The incentives for short-term wholesale nanciers to invest in the acquisition of information on bank project quality; The incentives for short-term wholesale nanciers to roll over funding or, alternatively, force bank liquidation at the intermediate stage; The incentives for banks to use short-term wholesale funds; and The optimal creditor seniority of short-term wholesale funds relative to long-term sources of funding (e.g. core retail deposits) that maximizes information acquisition and minimizes premature liquidation of good banks. The analysis starts from a Calomiris and Kahn (1991, CK) benchmark model, which we take to describe the bright side of wholesale funding. We then introduce a novel feature: a costless but noisy public signal on bank project quality. The signal is related to information that wholesale nanciers may obtain from public sources, e.g. liquid secondary markets for standardized assets (e.g. mortgage-backed securities) and credit rating agencies. We show that such a minor and plausible change to the CK set-up can dramatically alter its predictions, and explain a number of critical properties of the recent credit market turmoil. In the model, we consider a bank that nances a risky long-term project with two sources of funds: retail deposits and wholesale funds. Retail deposits are passive and provide a relatively stable source of long-term funds 1. Wholesale funds are relatively sophisticated and have capacity to acquire information on bank project quality. However they are short-term: provided on a rollover basis and have to be re nanced before nal returns realize or the bank will be forced into liquidation. 1 Retail "core deposits" are typically insured by the government; their withdrawals are motivated mostly by individual depositors liquidity needs and are predictable based on the law of large numbers. Although formally demandable, they therefore provide a relatively stable source of long-term funds. However the local deposit base is quasi- xed in size, as it is prohibitively expensive to expand it in the medium term (Flannery 1982; Billett and Gar nkel, 2004). When the deposit supply is not su cient to fund all available investment opportunities, banks can choose to attract, in addition, wholesale funds from sophisticated institutional investors. 3

4 In the benchmark case, we verify the CK result that sophisticated lenders have greater incentives to acquire information on bank project quality when assigned higher creditor seniority. Informed nanciers provide market discipline: roll over funds to good projects but force liquidation of bad ones. That enhances social welfare. Note that, in practice, sophisticated lenders indeed typically enjoy seniority thanks to the rst-come, rst-served sequential service constraint. Historically, short-term wholesale funds were often able to exit distressed banks well before retail depositors started to run in large scale (e.g. Continental Illinois, Northern Rock). We then show how a minor and plausible change to CK set-up can signi cantly alter the model s predictions. We consider an additional source of information: a costless but noisy public signal on bank quality that is available to the wholesale nanciers. They may use it when costly monitoring did not produce a perfect signal of bank quality (because of insu cient investment in monitoring or incidentally) 2. Examples of noisy signals include market-wide or sector-wide news that are imperfectly correlated with bank performance. The precision of this noisy signal (i.e., the probability that it is correct; can also be thought of as relevance of the signal to bank fundamentals) is one of the key parameters of the model. In practice, the relevance of available public signal can vary across banks, depending e.g. on asset type. For example, while the prices of mortgage-backed securities (MBS) can provide fairly precise information on the performance of mortgage banks, no similarly relevant public signal may exist for traditional banks that hold mainly small business loans. In any case, costless public signals can only provide imperfect information on bank s true asset quality. For example, while falling energy prices and close business relationship with the failed Penn Square Bank correctly predicted the downfall of Continental Illinois, the linking of Northern Rock to developments in the U.S. subprime mortgage market did not do justice to the bank which in fact had no 2 In the model, wholesale funds nanciers rst choose the level of costly information acquisition (monitoring) e ort. Higher e ort is more likely to yield a perfect signal on bank quality (which, if received, is correct), but the probability of receiving the signal is in any case less than one. If wholesale nanciers do not receive a perfect signal through monitoring, they have to decide whether to use the free but noisy public signal. Using the signal implies liquidating the bank upon a negative one; not using the signal means always rolling over funding regardless of what the signal indicates. 4

5 such exposures. The introduction of a noisy but costless public signal into CK model distorts the incentives of short-term wholesale nanciers in the following ways: It reduces their incentives to perform costly information acquisition activities. It creates excess incentives to liquidate banks at the intermediate date (by not rolling over funding) based on very noisy public signals. Note that the liquidation of good projects is ine cient from a social welfare perspective. Importantly, these distortions become stronger when wholesale nanciers are made more senior claimants to the liquidated assets sharply contrasting to CK results. Senior short-term wholesale nanciers have excess incentives to liquidate because they can withdraw ahead of passive retail depositors, collecting a larger share of a smaller pie. When wholesale nanciers anticipate a high likelihood of a safe exit from an early liquidation, they become less interested in acquiring information on bank project quality in the rst place. The seemingly safe bu er of long-term funds provided by passive retail depositors in fact facilitates early liquidations and discourages information acquisition by sophisticated nanciers. Our results reveal that short-term wholesale nanciers have greater incentives to liquidate strategically basing on a noisy negative signal when: Their creditor seniority (relative to passive depositors) upon liquidation is higher; The share of passive deposits in bank liabilities is higher; Interest rate o ered to wholesale nanciers in case of success is lower; Liquidation value of bank assets is higher; The noisy signal is precise enough (yet not necessarily as precise as to make liquidations optimal from a social welfare standpoint). 5

6 In the presence of a noisy costless signal, short-term wholesale nanciers always have insu cient incentives to monitor the bank. Their incentives are non-linear: increase in seniority for low values of seniority (as in CK), but decrease for higher values of seniority (when higher seniority translates purely into more liquidations). Consequently, the welfare-maximizing seniority of wholesale nanciers has an internal optimum. Deviations from that optimum to either side result in less monitoring and possibly more ine cient liquidations. This result sharply contrasts with the CK benchmark in which higher seniority for the sophisticated funds is always better. Overall, our model predicts that, the optimal seniority given to sophisticated shortterm wholesale nanciers should be lower if: The share of passive retail deposits (that serve as a bu er in liquidations) is higher; The precision / relevance of the costless public signal is higher; Liquidation value of bank assets is higher. These predictions suggest that traditional banks that hold small business loans, for which public signals confer little information, can o er higher seniority to wholesale - nanciers in order to induce greater information acquisition, in line with the predictions of CK. Yet banks with large exposures to sectors with readily available public information (e.g. mortgage banks) must not give too high seniority to short-term wholesale funds. Note that banks that hold securitized assets (e.g. MBS) appear particularly vulnerable to the risk of premature liquidations: trading of assets both provides a public signal on quality and boosts liquidation value. Therefore, CK s insights best apply to the traditional banking business with limited public information on asset quality, while our model can shed light on the new banking business characterized by traded securitized loans and credit rating agencies. To sum up, we show that higher seniority for wholesales funds is not always socially bene cial. In the presence of a costless but noisy signal on bank quality, higher seniority can in fact reduce monitoring and encourage ine cient liquidations. Social welfare is 6

7 constrained-maximized for an intermediate level of seniority, depending on the bank s funding structure (i.e. share of passive retail deposits on the liability side), the precision of public signals on bank project quality (which often depends on the type of assets held), liquidation value of bank assets, and interest rates o ered to wholesale nanciers. This is a novel result that usefully contrasts with CK and bears close resemblance to recent developments in the credit market, as well as some earlier instances of bank failures. It reveals the dark side of short-term wholesale funds, particularly when they are senior claimants. The model can also be applied more broadly, e.g. used to analyze risks in nancial institutions that operate without retail depositors and use wholesale funds only. Indeed, similar con icts of interest exist whenever liquidation payo s are skewed across groups of claimants (for contractual or behavioral reasons): e.g. short-term vs. long-term wholesale funds, or collateralized vs. unsecured wholesale funds. For example, Bear Stearns were nanced by long-term and short-term wholesale funds, with most of short-term funds being moreover collateralized. Our model explains why collateralized short-term lenders had insu cient incentives to acquire information on Bear s fundamentals, and were likely to walk out upon noisy negative news. Had these wholesale funds been unsecured or with longer contractual maturity (and hence lower e ective seniority), their providers would have thought twice before running on Bear Stearns. Note that the expectations of a government bailout can a ect incentives too, because in practice the e ective seniority of wholesale funds is determined not only by private by contractual choices (sequential service constraint; collateralization of wholesale funds by bank assets; clauses that can temporarily suspend redemption) but also o cial resolution rules expected to be applied by regulators and central banks in case of bank failure. The rest of the paper is structured as follows. Section 2 sets up the benchmark model and veri es the Calomiris and Kahn (1991) results. Section 3 introduces the costless but noisy signal on bank project quality and derives main results on the optimal seniority arrangement for short-term wholesale funds. Section 4 studies the incentives of banks to use short-term wholesale funds and choose certain seniority structure. Section 5 provides 7

8 several extensions of the basic model, addressing deposit insurance and systemic risks. Section 6 concludes. [Sections 4-6 TBC] 2 The Bright Side of Wholesale Funding This section sets out a simpli ed version of Calomiris and Kahn (1991) model to demonstrate the "bright side" of wholesale funding. We verify that: Informed "sophisticated" wholesale funds provide market discipline for banks: roll over funding to good projects but force liquidation of bad ones; Seniority of wholesale nanciers at the re nancing stage (e.g. implemented through the sequential service constraint) maximizes their incentives to acquire information on bank project quality. 2.1 Setup Consider a bank with access to a pro table but risky long-term project. The project requires a unit investment at date 0, and returns at date 2: X with probability p (where Xp > 1) or 0 with probability 1 p. The project may be liquidated at date 1 returning L (where L < p < 1). A bank has no own capital and needs to borrow in order to invest. There are two types of providers of funds: "Depositors", who are unsophisticated and thus never get advance information of date 2 realization; they are typically small individual retail savers insured by the government; and "Wholesale nanciers", who are sophisticated and can monitor the bank (at a cost) to acquire information on date 2 realization before date 1. Following CK, we establish that attracting short-term funds from "sophisticated" nanciers increases bank performance and social welfare. The reason is that sophisticated investors can monitor the bank and learn the date 2 realization before date 1, and 8

9 can withdraw early if they expect it to be low. In this case, investors will be able to savage liquidation value L instead of low date 2 realization 0. We also establish that, when a bank is funded by a mix of passive depositors and sophisticated investors, social optimum is achieved when the latter have seniority in a date 1 liquidation as that enhances their monitoring incentives. We analyze, in turn, three types of bank funding: passive depositors only, sophisticated lenders only, and a mix of the two under di erent seniority structures. 2.2 Solution Passive depositors only When bank is funded by passive depositors only, it always continues until date 2. The reason is that, without information, continuation is preferred to liquidation: L < 1 < px. Social welfare is D = px Wholesale nanciers only Now consider the case when funding is received from sophisticated nanciers only. Take the interest rate charged by them to be R. After lending to the bank, sophisticated nanciers choose the e ort with which they will monitor the bank. Monitoring with intensity m (0 m < 1) yields a precise signal of date 2 realization with probability m (and no signal at all otherwise), but has cost C(m) (C(0) = 0, C 0 (0) = 0, C(1) = 1, C 00 (m) > 0). In choosing m, sophisticated nanciers maximize 3 : 3 Note R > 1 > L pr + m(1 p)l C(m) 9

10 which obtains optimal m in: C 0 (m ) = L(1 p) (1) Note that m is a socially optimal amount of monitoring: wholesale nanciers internalize all its costs and bene ts. The social welfare is W = px +m (1 p)l C(m ) 1. Note that W > D as per CK. To complete the solution, one can nd R charged by competitive wholesale nanciers from pr + m(1 p)l C(m) 1 = 0, obtaining R = 1 + C(m ) m (1 p)l p Mix of passive depositors and wholesale nanciers Finally, consider the case when funding is provided by a mix of passive depositors and sophisticated lenders, in proportions D and W respectively (D + W = 1). We introduce parameter s (0 s 1) to describe the seniority of sophisticated lenders at date 1: in case of intermediate liquidation wholesale nanciers get sl. For s = 0 passive depositors are senior, for s = 1 wholesale nanciers are senior. In practice, the seniority of wholesale funds can be implemented through private contractual choices (sequential service constraint; collateralization of wholesale funds by bank assets; clauses that can temporarily suspend redemption) as well as o cial resolution rules expected to be applied by regulators and central banks in case of bank failure. In choosing m, sophisticated investors now maximize 4 : pw R + m(1 p)sl C(m) 4 For simplicity, we keep L < W R. Otherwise substitute by minfl; W Rg 10

11 which leads to optimal m mix : C 0 (m mix ) = sl(1 p) (2) Observe that socially optimal screening m mix = m is achieved for s = 1. This replicates the main CK result: o ering seniority to sophisticated nanciers strengthens market discipline. Social welfare s=1 mix = W. To complete the solution, one can nd R charged by competitive wholesale nanciers from pw R + m mix (1 p)sl C(m mix ) 1 = 0, giving R = 1 + C(m mix) m mix (1 p)sl W p R = 1 + C(m ) m (1 p)l for s = 1 W p Note that the equilibrium interest rate is higher than in case of only sophisticated nanciers, because sophisticated nanciers incur the full cost of screening but receive only a share W in high realization. Note also that R depends negatively on s; as s rises wholesale nanciers receive more in intermediate liquidations and need to be compensated less at date Summary This section has established the traditional "bright side" of wholesale funding. In this benchmark case, attracting funds from wholesale nanciers and giving them creditor seniority at the re nancing stage allows to elicit rst-best monitoring incentives and socially bene cial market discipline. 3 The Dark Side of Wholesale Funding We will now show how a plausible change to the model can signi cantly alter the CK results. We give sophisticated nanciers one more source of information. Even when 11

12 monitoring did not produce any information about date 2 realization (e.g. due to being insu ciently intensive or incidentally), sophisticated nanciers still obtain a free but noisy signal of date 2 realization in advance of date 1. (Passive depositors have no capacity to observe that signal as they are unsophisticated.) We show that such a seemingly minor twist can generate results that are almost opposite to those of CK. It appears that higher seniority of wholesale nanciers may lead to less monitoring and moreover give them excessive incentives to liquidate (withdraw at date 1) based on a very noisy negative signal. The reason for both e ects is that senior sophisticated investors bene t disproportionately in date 1 liquidations and may prefer getting a larger share of a smaller pie. We show that, as a result, when a bank is nanced by a mix of passive depositors and wholesale funds, social welfare is constrained-maximized for intermediate values of s (rather than s = 1), while the rst-best monitoring e ort of sophisticated nanciers cannot be induced. 3.1 Setup We introduce in the above setup a free but noisy signal that is received by sophisticated lenders before date 1. The signal takes two values, "g" or "b". It is characterized by a precision parameter (0 1, = 0 for uninformative, = 1 for precise). The probability of receiving "g" is p (same as for receiving X at date 2); conditional on that the probability of X realization at date 2 is [p + (1 p)], and that of 0 realization at date 2 is [(1 p) (1 p)]. The probability of "b" is 1 p; conditional on that the probability of X realization at date 2 is [p p], and that of 0 realization at date 2 is [(1 p) + p]. With the available noisy signal, the incentive structure of sophisticated investors has two dimensions. First, as before, they choose monitoring intensity m. Secondly, a new feature, when no information emerges through monitoring, they can choose whether to follow a noisy signal (and liquidate when it is negative) or not. 12

13 3.2 Solution Wholesale nanciers only We start by analyzing the case of a bank funded with sophisticated lenders (i.e., wholesale funds) only. If monitoring did not produce information on date 2 realization, the nanciers have two options. One is to disregard a noisy signal and always roll over funding to date 2 to receive on expectation Rp. Another is to follow a noisy signal and liquidate upon "b" to receive p [p + (1 p)] R + (1 p)l. Compare the always-roll-over strategy and the use-noisy-signal strategy. Note that, by relying on a noisy signal, sophisticated lenders obtain (1 p) [(1 p) + p] L in correct liquidations, but lose (1 p) [p p] (L R) in incorrect liquidations. They prefer to follow a noisy signal for (1 p) [(1 p) + p] L > (1 p) [p p] (L R) > liq = 1 L pr Note that the socially optimal value of liquidation threshold is = 1 L px In equilibrium, even when they are the only lenders, wholesale nanciers have incentives for excessive liquidations: liq <. We can now proceed to the analysis of monitoring decision. When sophisticated investors do not follow the noisy signal, they choose intensity of monitoring m as in (1): C 0 (m ) = L(1 p) When sophisticated investors do follow the noisy signal, they choose the intensity of 13

14 monitoring m liq by maximizing m hpr + (1 p)li + (1 m) hp [p + (1 p)] R + (1 p)li C(m) which obtains: C 0 (m liq ) = p [1 p (1 p)] R In this case, monitoring is below socially optimal: a socially optimal amount of monitoring is given by C 0 (m liq ) = p [1 p (1 p)] X Observe that, unlike m, m liq does not depend on liquidation value L. The reason is that nanciers receive L both in correct and incorrect liquidations. What m liq does depend on however is the high payo R that is lost in the case of incorrect liquidation (positively), and the free signal s precision (negatively). Note also that when sophisticated lenders do follow a noisy signal (p [p + (1 p)] R + (1 p)l > Rp) this implies p [1 p (1 p)] R < L(1 p) so there is a decrease in intensity of monitoring due to the possibility of noisy liquidation: m liq < m liq < m. Yet, in this case, less monitoring is a socially optimal response to more information being available for free. To complete the solution, one can nd R charged by competitive wholesale nanciers. (TBC) Mix of passive depositors and wholesale nanciers We can now proceed to the most interesting case. As previously, we take the respective investment in the bank by passive depositors and wholesale nanciers to be D and W (D + W = 1), and the seniority parameter to be s. If sophisticated investors do not follow a noisy signal, i.e., in the "continuation" scenario, they roll over funding and continue until date 2 to receive W Rp. If they do follow a noisy signal, and liquidate upon "b", they receive p [p + (1 p)] R + (1 p)sl. 14

15 Note that the former (continuation) payo is reduced from Rp to W Rp when W < 1, while the latter (liquidation) payo remains the same. This is the consequence of seniority rules at date 1: payment in liquidation depends not on the amount of funding initially provided but on seniority. Consequently, the incentives of senior providers of wholesale funds become biased towards liquidation. They choose to follow a noisy signal for (1 p) [(1 p) + p] sl > (1 p) [p p] (L W R) sl > mix = 1 pw R or s > s (1 ) pw R = L Note that mix can be smaller (more reliance on the noisy signal) or larger (less reliance) than depending on the values of s and W. When s > W R=X we get mix < and wholesale nanciers rely on the noisy signal more than it is optimal in the rst best, because of bene ts of seniority in early liquidations. This can trigger liquidations based on too noisy information. For s < W R=X the opposite holds: mix > and wholesale nanciers discard the noisy signal too often because they are junior and do not bene t enough from early liquidations even if they are rst best e cient. When sophisticated investors do not follow the noisy signal, they choose intensity of monitoring m mix as found in (2): C 0 (m mix ) = sl(1 p) When sophisticated investors follow the noisy signal, they choose the intensity of monitoring m mix+liq by maximizing m hpr + (1 p)sli + (1 m) hp [p + (1 p)] R + (1 p)sli C(m) 15

16 which obtains: C 0 (m mix+liq ) = p [1 p (1 p)] W R We can now analyze the impact of seniority on equilibrium outcomes. We consider two cases. Case 1: Liquidation based on noisy signal is not socially optimal This corresponds to <. Then, when wholesale nanciers are very junior, they have no incentives to use the intermediate signal (as in the rst best) but also have little incentives to monitor. As s increases, incentives to monitor increase, up to the point s when wholesale nanciers get incentives to liquidate prematurely. Observe that when s increases further beyond s, the incentives of wholesale - nanciers to monitor start falling again (note that s enters expression for C 0 (m mix+liq ) indirectly: higher s leads to lower R). Therefore point s (approached from the left) achieves both correct use of the noisy signal (disregard it) and maximum possible monitoring incentives. It is an important result that maximum monitoring is achieved for the intermediate value of s. Case 2: Liquidation based on noisy signal is socially optimal This corresponds to >. (This is a somewhat less relevant case.) When wholesale nanciers are very senior, they use the intermediate signal, but (as explained in Case 1) perform relatively little monitoring. As s decreases, incentives to monitor increase, up to the point s, where wholesale nanciers stop using the noisy signal. As before, point s (now approached from the right) achieves the correct use of the monitoring signal (follow it) and maximum possible monitoring incentives). 3.3 Summary We have shown that, in the presence of the costless but noisy signal on bank quality, the incentives of wholesale nanciers become distorted from the benchmark CK case. 16

17 In particular, wholesale nanciers will have lower incentives to monitor and incentives to over-rely on a the noisy signal, triggering undesirable liquidations of good projects. This demonstrates the "dark side" of bank wholesale funding. 4 Incentives of Banks to Use Wholesale Funds 5 Systemic Risk 6 Conculsions References [1] Billett M., Gar nkel, J., 2004 Financial Flexibility and the Cost of External Finance for U.S. Banks, Journal of Money, Credit and Banking, 36(5), [2] Calomiris, C, Kahn, C, 1991, "The Role of Demandable Debt in Structuring Optimal Banking Arrangements.", The American Economic Review, (81) June: [3] Calomiris, C., 1999 Building an Incentive-Compatible Safety Net, Journal of Banking & Finance, 23(10) [4] Goodfriend, M., King, R.G., Financial Deregulation, Monetary Policy, and Central Banking. Fed. Reserve Bank Richmond Econ. Rev. May/June, 3-22 [5] Flannery, M., 1982 Retail Bank Deposits as Quasi-Fixed Factors of Production, The American Economic Review, 72 June: