Having spent the last month or so poring over writings on last-resort lending,* and especially writings dealing with the recent crisis and its aftermath, with the particular aim of discovering the best means for supplying last-resort credit when it's called for, and for not supplying it when it isn't, I've reached a number of tentative conclusions that seem worth reporting. I report them despite their tentative nature so that I might be convinced sooner rather than later that I'm barking up the wrong tree, and also because, if I'm actually on to something, I might get others to help me flesh-out my ideas.
I hasten to add that I regard any need for last-resort lending as reflecting, not the inherent shortcomings of private financial markets, but the debilitating effects of misguided regulatory interference with the free development of those markets. Some of the least regulated banking systems of the past, including systems that lacked central banks, were also famously crisis-free, or close to it.
Modern banking systems are, sadly, a far cry from those ideal arrangements. It's quite impossible, for that reason, to suppose that we might safely dispense altogether with central bank last-resort lending, without first undertaking other, major reforms. In the meantime, we can strive to reform last-resort lending arrangements, so that they might lower the risk of future crises, instead of making them more likely by contributing to moral hazard, or by otherwise misallocating credit.
A False Dichotomy
Conventional wisdom has it that contemporary central banks must perform two fundamentally distinct duties. According to it, they are, first of all, responsible for implementing monetary policy, meaning that they must manage the aggregate supply of liquid reserves so as to reach various short- and long-term macroeconomic targets. But they must also serve as sources of "last-resort" credit when doing so serves to prevent or contain financial crises.
This established dichotomy of central bank duties has in turn informed a corresponding division of central bank facilities, with one facility or set of facilities serving for the implementation of "ordinary" monetary policy, and the rest devoted to supplying last-resort credit. In the United States, until the recent crisis, ordinary monetary policy was implemented by means of open-market operations conducted with a limited set of counterparties, known as primary dealers, and administered by the New York Fed. Last-resort credit, in contrast, took the form of discount-window loans, administered by each of the twelve Federal Reserve Banks, for which most depository institutions were eligible. Separate ordinary and last-resort liquidity-provision facilities were also standard in other systems.
Although the recent crisis witnessed extraordinary modifications of central bank liquidity-provision facilities, both in the U.S. and elsewhere, and although some of these modifications have become permanent, the conventional dichotomy of duties and facilities has survived, if indeed it has not been reinforced. The most obvious consequence of the crisis consisted of the creation of various new, though mostly temporary, last-resort lending facilities, aimed at supplying emergency credit to institutions that could not or would not get it from established facilities. The new facilities were sometimes open to counterparties to which established facilities were closed; or they were prepared to accept collateral that those facilities would not. In some instances, such as the Fed's Term Auction Facility (TAF), the new facilities dealt with the usual counterparties and collateral, but did so in a manner calculated to avoid the "stigma" attached to ordinary last-resort borrowing.
But for all the ingenuity that went into these novel lending facilities, and all the good they may (or may not) have accomplished, I fear that their establishment may cause the wrong conclusion to be drawn from the crisis, namely, that more or perhaps broader but nonetheless specialized last-resort lending facilities are needed if future crises are to be avoided. The proper lessons to be drawn, IMHO, are dramatically different. They are, first and most fundamentally, that the conventional dichotomy of central bank duties is a false dichotomy ; and, second, that the problem with traditional central banking arrangements is, not that they lack adequate facilities for emergency lending, but that they rely on such facilities at all, instead of having properly designed facilities for the implementation of "ordinary" monetary policy.
To be more specific, the conventional dichotomy, though it may have had some merit in the now-distant past, when implementing "ordinary monetary policy" meant little more than maintaining the gold standard, while last-resort lending was a largely independent matter, is false when applied to modern fiat-money arrangements. A fiat-money issuing central bank has but one fundamental duty to fulfill. That duty consists of supplying cash, meaning currency and bank reserves, in amounts sufficient to meet macroeconomic targets, and doing so efficiently, that is, so that newly created cash is assigned to those parties that are willing to pay the most for it.
Special Last-Resort Lending Facilities are Inherently Inefficient
Does it really matter whether we think of central banks' fundamental responsibilities as consisting of two separate duties, or of a single duty performed efficiently? It matters a great deal, because supposing that central banks have not one but two duties to perform, and then encouraging them to employ separate facilities for each, actually makes the achievement of an efficient allocation of credit, including efficient last-resort lending, highly unlikely, if not impossible. This follows from the fact that, taking its "ordinary" monetary targets, and the amount of new reserve creation needed to achieve them, as given, a central bank operating multiple facilities, each catering to different sets of counterparties or dealing in different sorts of collateral, and offering credit on different terms, must allot specific portions of the credit to be created (some of which may be negative, as when last-resort loans are "sterilized" by open market sales) among the various facilities. Such allocations are bound to be somewhat arbitrary, if not flagrantly so. And even if the allocations were somehow correct, the facilities themselves, in so far as they offer credit on implicitly (if not explicitly) distinct terms, would likely favor certain eligible counterparties over others. Finally, because counterparties do not all compete with one another for the same pool of funds, the ultimate allocation of those funds may be inefficient even when all face similar terms. Think of holding the Olympics at two facilities, with half the teams competing at one and half at the other, and you get the idea.
Instead of suggesting the need for multiple liquidity-provision facilities, the view that central banks have no responsibility save that of implementing "ordinary" monetary policy, but doing so efficiently, points to the desirability of assigning as large a role as possible to the price mechanism as the means for allocating newly-created cash among competing applicants. That goal is best accomplished by means of a single facility for auctioning credit, at which numerous eligible counterparties could compete for available central bank credit on equal terms. Under this arrangement, the central bank, once having set the terms of the auction, would have no other duty to perform save that of determining the aggregate amounts of credit to be auctioned. "Last-resort lending," instead of being a distinct central bank duty, would become an incidental counterpart of ordinary monetary policy, consisting of that part of auctioned credits taken up by liquidity-strapped counterparties that choose to take part in auctions only as a last-resort. In short, there would be last-resort borrowers, but no last-resort lending operations as such.
Achieving "Flexible" Open-Market Operations
So much for the theory. Can the ideal I've sketched-out be achieved in practice? I believe that it, or something very close, can be achieved, and without any great difficulty, both in the U.S. and elsewhere. As the present Federal Reserve System is in many respects among those furthest removed from the sort of system I think desirable, I'll outline the basic steps required to move from the present system to what I'll call a system of "flexible" open-market operations (or Flexible OMOs, for short). Some of the proposals are very similar to ones I originally suggested several years ago, so I encourage readers to consider the arguments I offered in defense of that earlier plan.
First, the Primary Dealer System — the system that confines the Fed's ordinary open-market dealings to a small set of counterparties — should be abolished. Instead, all commercial banks presently eligible for discount window loans, and all Money Market Mutual Funds, should be able to take part along with existing Primary Dealers in the Fed's ordinary credit auctions.
Second, while continuing its traditional practice of confining its outright or "permanent" open-market purchases to U.S. Treasury and agency securities, the Fed should stand ready to accept other sorts of collateral, including all collateral that is presently accepted as security for its discount-window loans, while assigning appropriate "haircuts" to riskier collateral, in its temporary open-market purchases or repos. (A "repo" or "repurchase agreement" is essentially a security purchase accompanied by a commitment of the seller to repurchase the security for an agreed-upon price after a specific — and generally quite brief — period of time.)
Third, the Fed should offer "term" (30 or even 60-day) repos as well as the more usual overnight repos, as the former are more helpful in tiding-over liquidity-strapped firms during financial emergencies.
Fourth, to allow counterparties to bid for credit using different sorts of collateral, the Fed should adopt a version of the "product mix" auction originally developed several years ago by Paul Klemperer, and employed since by the Bank of England in its indexed long-term repo operations (ILTRs). Klemperer's procedure allows bidders to submit multiple mutually-exclusive "sub" bids for a desired amount of credit, each offering different sorts and amounts of collateral. Then, as an article in The Economist explains,
Having received a set of bids for different goods, at various prices and quantities, the auctioneer in Mr Klemperer’s set-up then conducts a proxy auction on bidders’ behalf to see who should get what, and what the price should be. Because nothing is revealed to the bidders and they know they cannot influence this process, their best bet is to tell the truth. What is more, since the auctioneer has price information for a range of quantities, it is possible to see how prices change as supply does.
For details, including explanations of how the auction avoids adverse selection problems, readers should consult the linked sources. The bottom line, though, is (in The Economist's words again) that the auction design serves to "provide accurate information on individual banks’ demand for liquidity and the prices they are willing to pay for it." What's more, the Bank of England has discovered that it can "use the pattern of bids in each auction to assess the extent of stress in the market," and to thereby "inform its decisions on the size and maturity of future operations." In other words, Flexible OMOs serve, not only to make last-resort lending facilities redundant, but to help guide ordinary monetary policy, making it less likely that monetary authorities will err by incorrectly gauging the aggregate demand for liquidity, as Federal officials did, with tragic results, in 2008.
Finally, the Fed should permanently close its discount window, which will have become redundant once Flexible OMOs are provided for.
Flexible OMOs, and Central Bank Discretion
Superficially, the changes I've proposed may appear to award the Fed more powers than it has enjoyed in the past, by allowing more counterparties to engage in open-market operations with it, using previously unacceptable collateral. But the impression is mistaken, for a number of reasons.
First of all, as I've already noted, Flexible OMOs are meant to render all "emergency" lending operations and facilities, whether actual or potential, redundant. That means that they eliminate the rationale, not just for ordinary discount window lending, but also for lending targeted at specific banks deemed too "Systematically Important" to fail, as well as direct lending to non-banks under the Fed's current 13(3) authority. By opening access to the Fed's ordinary credit auctions to numerous counterparties, including all those institutions, whether banks or non-banks, that play a prominent role in the payments system, Flexible OMOs should make it possible for any of these counterparties that is for any reason unable to secure needed liquidity from private sources to apply directly to the Fed for it, and, by outbidding rival applicants, to get it. What's more, by dealing with the Fed's ordinary credit-creation facility, rather than with any facility explicitly devoted to last-resort or "emergency" credit provision, firms will avoid any risk of finding themselves stigmatized, and therefore worse off, than they might be if they refused central bank credit altogether.
Second, by having all counterparties compete for credit offered through a single facility, according to common terms, the reform eliminates opportunities for favoritism that arise when different counterparties must deal with different facilities operating under different rules.
Third, by eliminating distinct last resort lending operations, Flexible OMOs make it unnecessary for authorities responsible for such operations to coordinate their efforts with those of separate central bank authorities charged with conducting ordinary monetary policy operations. The elimination of multiple authorities also reduces the risk of shirking, by placing responsibility for adequate aggregate liquidity provision firmly on the shoulders of a single decision-making authority — here, the FOMC.
Fourth, Flexible OMOs should rule-out any future resort to ad-hoc emergency lending facilities, establishing instead a stable and predictable arrangement for central bank liquidity provision, meant to meet both ordinary and extraordinary liquidity needs. The existence of fixed arrangements for liquidity assistance, combined with the competitive pricing of such assistance, allows prospective borrowers to prepare themselves in advance for potential liquidity shocks, while ruling-out moral hazard.
Fifth and finally, Flexible OMOs simplify central bank decision making, by reducing it to two components, namely (1) the determination of aggregate credit amounts to be auctioned, and (2) the setting, and occasional re-adjustment, of various auction parameters, including collateral haircuts. Credit allocation, including its allocation to solvent firms faced with a liquidity shortage that have sought funding from the Fed only as a last resort, is otherwise automatic. There would be no practical distinction, indeed, between the Fed's conduct during episodes of financial distress and its conduct on other occasions. The only changes would be in the unusual counterparties taking part in the Fed's auctions, the wider range of collateral types offered, and the higher-than-usual interest rates implicit in winning bids.
The relatively automatic nature of last-resort credit provision under a system of Flexible OMOs makes such a system a natural counterpart to rule-based, if not fully-automatic, systems for determining the scale of central bank aggregate credit creation, such as the proposals of Scott Sumner, David Beckworth, and others for targeting nominal GDP.
Although my proposal may seem radical, its various elements are far from being without precedent. As I've noted, the Bank of England already employs product-mix auctions to allocate funds among competing bids involving different sorts of collateral. The ECB, for its part, has always accepted a relatively wide range of collateral in its ordinary (short-term) open-market operations; it also conducts those operations with numerous counterparties. The ECB was, for both of these reasons, able to cope with the first year of the financial crisis without having had to make any changes to its operational framework. The Fed itself has, finally, occasionally and temporarily resorted to unorthodox open-market operations, involving a larger number of counterparties, a wider range of securities, and different repurchase terms. To supply liquidity in connection with Y2K, it extended the term of its its repurchase agreements, while also offering to purchase a wider range of securities. More recently, in September 2013, it established a special overnight reverse repo (ON-RRP) facility, through which it deals, not with its usual set of primary dealers, but with money market mutual funds, GSE’s, and a broader set of commercial banks. More recently still, it began undertaking sizable term (as opposed to overnight) reverse repos using that facility. During the late 1990s and early 2000s, when confronted with what was then a looming shortage of Treasury securities, the Fed also gave serious thought to the possibility of permanently expanding the list of securities it might purchase, both in its repo operations and outright.
What distinguishes my plan for Flexible OMOs from these precedents is that it envisions a single facility only, supplying both ordinary and last-resort credit, and doing so in a way that relies to the fullest extent possible upon market forces, rather than decisions by bureaucrats, to achieve an efficient allocation of liquidity among competing applicants. By allowing a broad set of potential applicants, using a wide range of eligible collateral, to compete for available funds, not only in private markets, but, when necessary, at a single Federal Reserve facility, Flexible OMOs actually serve to minimize the Federal Reserve’s credit footprint, and to thereby prevent it from taking part in deliberate credit-allocation exercises for which fiscal rather than monetary authorities ought to be responsible.
Back to Bagehot
At first blush, the reforms I've proposed may also seem inconsistent with received wisdom regarding the principles of last-resort lending. But they are actually far more faithful to that wisdom, particularly as formulated by Walter Bagehot, than existing arrangements. Consider Bagehot's seminal statement of now conventional last-resort lending principles, as found in Lombard Street:
First. That [last-resort] loans should only be made at a very high rate of interest. This will operate as a heavy fine on unreasonable timidity, and will prevent the greatest number of applications by persons who do not require it. The rate should be raised early in the panic, so that the fine may be paid early; that no one may borrow out of idle precaution without paying well for it… .
Secondly. That at this rate these advances should be made on all good banking securities, and as largely as the public ask for them. The reason is plain. The object is to stay alarm, and nothing therefore should be done to cause alarm. But the way to cause alarm is to refuse some one who has good security to offer… No advances indeed need be made by which the Bank will ultimately lose. … If it is known that the Bank of England is freely advancing on what in ordinary times is reckoned a good security — on what is then commonly pledged and easily convertible — the alarm of the solvent merchants and bankers will be stayed. But if securities, really good and usually convertible, are refused by the Bank, the alarm will not abate, the other loans made will fail in obtaining their end, and the panic will become worse and worse.
Allowing that there is only a trivial difference between repos and securitized loans, there is after all little difference between what Bagehot recommends and what Flexible OMOs would provide for. Indeed, they make for a more certain commitment to the principle of making last-resort credit available both "largely" and at suitably "high" rates, for the auction procedure itself assures that, in times of extraordinary need, high rates are bound to prevail. If you ask me, it's not my proposed reform, but the dizzying array of emergency lending facilities seen in the course of the recent crisis, with all the opportunities for inefficient credit allocation they entailed, that would have struck Bagehot as odd.
Such are my thoughts, so far, on reforming last-resort lending. Like I said, I may be barking up the wrong tree. For that reason, I especially welcome critical comments pointing out how my proposed reform might go wrong. If those comments come with alternative suggestions for improving on what I've proposed, all the better.
*Apart, that is, from a portion spent pouring port in Porto, Portugal.