Friday, August 17, 2007

Central banks as market makers of last resort 2

My earlier blog with Anne Sibert on modern central banks as market makers of last resort (MMLR) rather than Bagehotian lenders of last resort (LOLR), has prompted a number of thoughtful and stimulating comments, the longest from Simon Cox of the Economist. I will try to deal with most of them here.

(1) What are the ‘welfare economics’ and efficiency arguments for the central bank acting as MMLR?

The Bagehotian LOLR intervenes for two reasons.

(i) Bank runs are unnecessary. Typically, the formal models of bank runs show that there exist (at least) two equilibria: one in which there is no bank run and at one in which there is no bank run. Without a central bank acting as LOLR (or some other device like compulsory bank holidays, in which the sequential service constraint (the first-come, first-served right of depositors to withdraw their money) is suspended), there can be a ‘coordination failure’ causing depositors to focus on the wrong equilibrium, the ‘run’. The LOLR eliminates the run equilibrium.

(ii) Bank runs are costly. They may cause banks to go bankrupt; they may impair the payment system (when checkable bank deposits are an important means of payment) and they may interfere with the ability of the banking system to extend new loans or roll over existing ones, thus impairing the supply of trade credit, working capital etc. etc. Bankruptcy is socially costly. It is not just a reshuffling of ownership claims. Real resources are wasted on lawyers, auditors, accountants and bailiffs. ‘Going concern value’, including goodwill may be destroyed. Established relationships between borrowers and loan officers may be interrupted and social capital diminished.

The efficiency argument for the MMLR function in an economy where most credit is extended through the financial markets is similar to, but even stronger than that for the LOLR in a bank-dominated credit system. That is because a market, an arrangement for bringing together and matching willing buyers and sellers, has many of the features of not just a public good, but of a public good with network externalities.

Conventional pure public goods are non-rival and non-excludable. Markets don’t, on the whole, have the non-excludability property: it is possible without incurring too high a cost, to exclude a would-be trader from operating in a market. Markets have, however, the non-rivalness property on steroids. With a non-rival good or service, my ability to consumer or enjoy it is not diminished if someone else consumes or enjoys it. With a matching mechanism like a market, the usefulness of the market to any would-be buyer or seller increases when the number of alternative buyers and sellers increases. Such network externalities mean that, if the market is also costly to establish and run, it is unlikely to function efficiently as an unregulated private, profit-maximising venture. This is why many markets are regulated either by a trade association including buyers or sellers, or by some agent of the state.

A disorderly market – one in which the matching mechanism has broken down – inflicts real costs on the would-be buyers and sellers who fail to make a match. At the very least, the gains from trade are foregone. If the sale of an asset in the market in question is the only available means to raise cash to forestall default or even bankruptcy, all the arguments about socially costly bankruptcy made in the context of the LOLR and bank failure apply here also. A ‘materiality test’ clearly applies. If the market in question is not significant, in the sense that even a complete lock-down of the market would not drive would-be market participants into bankruptcy, there is no case for a MMLR. There has to be a risk of substantive (‘material’) avoidable economic costs being incurred. Logically, that does not necessarily require that a run on an asset or asset class is spilling over into other asset markets. The asset in question could be significant enough in its own right for intervention to be warranted. I don’t believe that the case for a MMLR is significantly strengthened if the malfunctioning of certain credit markets threatens depository institutions like commercial banks. I no longer consider depository institutions of special interest in well-developed financial markets. Their checkable deposit liabilities no longer play a crucial role in the payment mechanism, and their role in extending loans and other forms of credit to businesses and households have been duplicated by non-bank institutions. With securitisation their loan asset portfolio is also more liquid than it was before (although not as liquid as the banks and their counterparties believed until recently).

Clearly, when there threatens to be a significant impact on employment and output, the ‘material damage’ test is likely to be met. But it is not necessary. Even in a world with continuous full employment, the real resource costs associated with widespread unnecessary bankruptcy and default could be sufficient reason to intervene.

(2) Would the MMLR not create moral hazard?

Of course it would, just like the LOLR. Moral hazard is almost unavoidable whenever the government intervenes anywhere. The practical issue is how moral hazard can be minimized and whether, when the moral hazard is minimized, the benefits of MMLR intervention exceed the costs. The observation that a government intervention creates moral hazard is not the end of the argument, but the beginning of an intelligent discussion.

Consider the example of a number of hedge funds that have made highly leveraged bets on CDOs, and the banks that provided the hedgefunds with the means to build up their highly leveraged portfolios. The hedgefunds are exposed to the CDOs; the banks are exposed to the hedgefunds. Then a CDO panic hits and the market for CDOs freezes. Some of the funds may have made bets whose success depends on the persistence of low risk-free rates and low credit risk spreads. These bets will go wrong. Hedgefunds that have made enough bets if this kind are insolvent and should fail. So should the banks that have lent to them. However, there may be other hedgefunds that have made bets whose success depends only on the existence of an orderly market for CDOs at prices commensurate to the true risk profile. They could fail also, if markets seize up, and so could banks exposed to them. If the central bank comes in as MMLR and names a bid price (P1) for CDOs, those illiquid hedge funds may well survive. Moral hazard would be avoided if the bid price were not so high that the insolvent institutions are bailed out also.

Given the uncertainties prevailing in times of financial crisis, it is clear that moral hazard will be hard to avoid completely when the central bank acts as market maker. The central bank should therefore not step in at the first sighting of a blip on the liquidity crisis radar screen. There must be some ‘materiality threshold’ below which the central bank does not intervene. It is hard to define that threshold in the abstract and in general, or even to list a comprehensive set of operational criteria that would be the building blocks of such a threshold. But the same applies to classic Bagehotian LOLR operations. That’s what the art of central banking is about. If it were easy, we would have nine monkeys doing it.

(3) On what terms should the MMLR enter the market?

(3a) What should be the bid-ask spread (P2-P1)?

In competitive private markets, the bid-ask spread should reflect the transaction cost of buying and selling the same instrument. In most liquid financial markets that is tiny, especially for large volume transactions. When dealing with the bid-ask spread of the MMLR, I propose something quite different. I would use the bid-ask spread to impose part of the penalty that should be faced by institutions caught with their liquidity pants down.

I would base my spread proposal on what is currently implied by the ECB’s and the Bank of England credit and deposit facilities (and on the Fed’s credit facility, or discount window)

The ECB’s Marginal Lending Facility currently charges a 5.00% rate, 1.00% above the ECB policy target rate, the Main Refinancing Operations Minimum Bid Rate, which stands at 4.00%. There is also a deposit facility which currently stands at 3.00%, 1.00% below the ECB’s policy target rate.

The Bank of England also defines its money market operations through three rates (or a central rate and a symmetric zone around the central rate). In addition, there are the Bank’s standing facilities. “Standing deposit and (collateralised) lending facilities are available to eligible UK banks and building societies. They may be used on demand. In normal circumstances, they carry a penalty, relative to the Bank's official rate, of +/- 25 basis points on the final day of the monthly reserves maintenance period and of +/- 100 basis points on all other days.”. The policy target rate, the Repo rate currently stands at 5.75%. The Bank of England stands ready to lend, to eligible counterparties and against eligible collateral, any amount at a rate 1.00% above the Repo rate (currently that would be at 6.75%). There is also a rate, 1.00% below the Repo rate (currently that would be 4.75%) at which eligible counterparties can deposit funds with the Bank of England.

From the Bank of England’s website, we can infer some of the tensions between its monetary policy objectives and its desire not to be the liquidity port of first call in any credit crunch. Objective 1 of the Bank of England’s sterling money market operations is “: Overnight market interest rates to be in line with the Bank's official rate, so that there is a flat money market yield curve, consistent with the official policy rate, out to the next MPC decision date, with very limited day-to-day or intra-day volatility in market interest rates at maturities out to that horizon.”

The Fed’s discount window has three different facilities and associated rates; the benchmark primary credit rate currently stands at 6.25%, 1.00% above the Federal Funds target rate; the secondary and seasonal credit rates exceed the primary rate. The ECB’s Marginal Lending Facility currently charges a 5.00% rate, also 1.00% above the ECB policy rate, the Main Refinancing Operations Minimum Bid Rate, which stands at 4.00%. Financial instruments eligible for collateral in discount operations (or repos) are valued at their market prices and a haircut (‘liquidity discount’) is applied to them. The Fed does not have a deposit facility, presumably because bank reserves in the US are non-remunerated.

So, my proposed spread P2-P1 would be equivalent to a 2.00% yield differential between the central bank’s collateralised lending rate and its deposit rate.

Note that the Bank of England has applied this philosophy also during the liquidity kerfuffles of the past week. Unlike the ECB and the Fed, which injected liquidity at lending rates equal to their policy rates, the Bank of England was willing to let the overnight interbank rate go up by almost 100bps. There would be unlimited liquidity (against suitable collateral) but only at a 100bps penalty relative to the policy rate.

I like the Bank of England’s approach, even if it conflicts with Objective 1 of its sterling money market operations. The ECB and the Fed made credit available without a penalty to any and all comers, including those who did not need it. This lays the foundations for the next credit boom.

(3b) What should be the difference between its bid price (P1) and the risk-free price of the security (P)?

How would the central bank determine the right bid price (P1)? (Honest answer: beats me, as there’s no market ….). It would have to use all its knowledge of the fundamental determinants of the value of these securities and then punt, make a bid and live with the profit or loss. That would mean recruiting a sizeable number of staff with quite different qualifications from that typical of current central banks. In addition to the macroeconomic and monetary economics skill sets familiar to the traditional central bankers, the central bank, in order to act as market maker of last resort, would have to have people with empirical asset pricing skills; so quantitative finance skills would be very important. In addition, since main-stream finance assumes perfect competitive markets, there would be a need for people with skills in 'micro-market structure', capable of analysing markets and price setting in markets that don't behave like the textbook model. Behaviour finance skills would also be required.

I would, however, insist that everyone, quantitative traditional finance expert, micro-finance expert and behavioural finance expert be firmly grounded in a thorough understanding of modern dynamic macroeconomic theory and its applications. Every worker in the MMLR vineyard should be able to think intelligently about the relationship of market movements to fundamentals. Without that, we end of in Quant land, which is a very dangerous place for anyone, let alone a central bank, to be.

Which reminds me. I think it is Larry Summers who defined finance theory as the science that explains why one 12 ounce can of Pepsi should sell for the same price as two six ounce cans of Pepsi (a straightforward implication of ‘no-arbitrage’ - the only behavioural content of finance theory).

Empirical finance then shows why this relationship does not hold in practice. When done properly, finance is just a subset of partial equilibrium microeconomics, explaining the behaviour of any one asset in terms of the behaviour of the remaining assets, but never explaining the lot. Useful, but limited. Quants strip financial economics of its last bit of economic content. There is no longer any discussion of fundamentals or other factors driving prices, because past prices themselves are the only fundamental considered. What quants do is, fundamentally, no more than high-cost data mining. Not surprisingly, its affectionados make large amounts of money by shifting large amounts of money in response to small perceived anomalies in orderly markets when nothing dramatic happens; they also lose large amounts of money when there is a big change in the economic environment or when markets become disorderly.

Central banks would have to have staff who understand and can manipulate and stress-test the mathematical/statistical models used currently by financial engineers and the private institutions that employ them to ‘mark to model securities that cannot be ‘marked to market’. Central bank staff will have to get their hands dirty and deal regularly with a whole range of actual and potential counterparties, dealing in a whole range of financial instruments, both exchange-traded and OTC, to be able to understand the price discovery and price setting mechanisms in these markets.

Central banks would have to work much more closely with the rating agencies. The rating agencies will, of course have to be reformed radically given their history of rating failures not just in the subprime markets, but in a whole string of structured products and fancy derivatives, including CDSs. It is clear that rating agencies should not be engaged in any other business other than the rating business. Conflict of interest is just too pervasive if this guideline is ignored. Rating agencies should no longer be paid by the issuers of the financial instruments being rated. They should either be paid by the industry as a whole or by the holders of the instruments. There should also be much greater entry into the rating agency Walhalla; a triopoly is just too few.

Finally, central banks would have to take much more credit risk onto their balance sheets by acting as MMLR for financial instruments that are below investment grade.

It is key that the efficient functioning of the financial markets be more important to the central bank than the profit and loss account or balance sheet probity of the institution. A central bank that does it job properly may have to be recapitalised from time to time. Central bankers hate few things more than going cap-in-hand to the Treasury or ministry of finance to ask for a capital transfer. They also fear this might undermine their independence. We should, however, always remember that central bank independence is not of interest in itself, but only to the extent that it contributes to superior outcomes.

(4) How does the central bank know the difference between a disorderly market and one in which some news has reduced the fair value of banks' assets to a lower price than the banks find acceptable? Perhaps if the central bank did nothing, liquidity would be available at some firesale prices. If so, does the LOLR function represent a subsidy to the banks and prevent the emergence of private sector vulture lenders?

Good question. The first part of the answer is: see the answer to (3b). If liquidity were to be available at firesale prices that accurately reflected a fundamentally low fundamental probability of survival of the bank (or jointly of the bank and the issuers of the instruments the bank is offering as collateral), then there would be no reason to intervene. (We would also have to believe that the private sector vulture lenders/funds were to be equally efficient managers of the portfolio they are acquiring as the banks, or at least as efficient as the publicly funded vulture fund known as the MMLR central bank I am proposing!).

We must believe, just as in the case of the LORL, that there is a clear conceptual distinction between illiquidity and insolvency and that the central bank has some reasonable chance of making that distinction operational.

(5) Should CDOs and other instruments that were not very liquid in the first place, should be made liquid (but at a price!) by the central bank?

CDOs were never very liquid in the first place. Those who viewed new-style special purpose vehicles used in securitising bank loans, or hedge funds as providers of liquidity confused cheap credit and liquidity. Old-style banks with illiquid loans on the asset side of their balance sheets provided credit, not liquidity. At times they provided cheap credit (when loan rates were low and collateral requirements or other covenants ‘lite’), but they did not provide liquidity. Neither did hedgefunds. Should the MMLR make liquid (make a market for) things during bad times that weren’t very liquid even during good times? I would argue that, under the right circumstances, yes, but at a price. The same problem of liquifying the normally illiquid applies of course to the LORL during bank runs – the LORL effectively makes illiquid bank loans liquid. If the discount at which these securities can be sold to the central bank is hefty enough, moral hazard can, I believe, be minimized.

(6) Should the MMLR accept as counterparties only suitably prudentially regulated and supervised institutions?

Another interesting question is whether access to the central bank as market maker should be restricted to institutions that accept a sufficient degree of regulatory oversight from the central bank or some other regulatory/supervisory institution.

I think it would certainly be desirable to make access to the MMRL facility contingent on prior acceptance of whatever degree of prudential regulation and oversight the central bank and the rest of the regulatory community would deem appropriate. That would create two classes of institutions, unregulated ones without access and regulated ones with access. If the central bank can commit itself to let the unregulated ones go under, should the occasion arise, this arrangement might work (a bit like FDIC-insured for deposit-taking institutions. The alternative would be to make prudential regulatory oversight mandatory for a much wider class of institutions, defined by what they do (functionally) rather by what they call themselves.

I would prefer to create two classes of institutions, across the hedgefund, investment bank etc. universe: those who accept prudential regulation and supervision and have access to the discount window and the MMLR, and those who don’t. No doubt, the ‘uninsured/unregulated’ institutions would, during a crisis, try to get ‘insured/regulated’ institutions to borrow more than the ‘insured/regulated’ institution needs in order to on-lend to the uninsured/unregulated institutions, but that would be good news, not bad news.

(7) What kinds of securities should the MMLR either purchase outright or accept as collateral for loans or in repos?

The brief answer is anything the markets are not willing to trade but which, in the opinion of the central bank, still has fundamental value. During this past week, the central banks missed a huge opportunity by sticking to their ‘normal’ times criteria for collateralisable assets when they injected liquidity into the market. The Fed made much of the fact that it enlarged the set of securities it accepted from US Federal government securities and securities issued by US Federal agencies (Fanny May, Freddie Mac etc.) to included mortgage-backed instruments guaranteed by US agencies. Big, big deal! Not! If the Fed had been had been willing to quote a price at which it would accept as collateral lower tranches of securities backed by subprime mortgages not guaranteed by any Federal agency, I would have been impressed and the liquidity crunch would have been tackled right at the point where it hurt. Instead, the Fed, like the ECB, used a blunderbus approach to provide liquidity to all and sundry, including those who did not need it.

The ECB, which injected in total probably around €140bn into the markets, has put itself outside the targeted, well-focused liquidity provision game by not accepting anything rated below the A category as collateral in repos.

So too much liquidity was pumped in overall, and too little in the right places. Until the world’s leading central banks during liquidity crunches accept complete junk as collateral in repos (appropriately, indeed punitively priced and subject to a sobering haircut), they are not doing their jobs properly.

© Willem H. Buiter 2007


lombard_lombard said...

A practical question:
How can a central bank manage to recruit such experts with the meager government-sector salary? Anyone with such a skill set as you described below would certainly get offers from the private sector worth between 2-3 times the Bank of England salary for economists. Maybe the central bank can make exception to its salary scale system to pay such high-quality expert with competitive market salary during financial crisis; but to become a reliable market maker of last resort that is ready all the time, the central bank needs to store such experts in-house during normal times too; I don’t think any central banks can afford that financially.

"In addition to the macroeconomic and monetary economics skill sets familiar to the traditional central bankers, the central bank, in order to act as market maker of last resort, would have to have people with empirical asset pricing skills; so quantitative finance skills would be very important. "

Keith said...

One quick rule of thumb springs to my mind on setting the penalty discount rate for owners of distressed subprime-backed securities:

That penalty interest rate had sure as heck be higher than the retail rate on 30-year fixed prime 20% mortgages.

After all, it seems like all of our subprime holders would trade their current assets for prime mortgages if they could.

Of course, the Fed just cut the discount rate to half a point below the retail rate on 30-year fixed prime mortgages.

So it looks more like a complete bailout than a LOLR/MMLR saction.