EFTA00611654.pdf

Toward a run-free financial system John H. Cochrane' April 16 2014 Abstract The financial crisis was a systemic run. Hence, the central regulatory response should be to eliminate run-prone securities from the financial system. By contrast, current regulation guarantees run-prone bank liabilities and instead tries to regulate bank assets and their values. I survey how a much simpler, rule-based, liability regulation could eliminate runs and crises, while allowing inevitable booms and busts. I show how modern communications, computation, and financial technology overcomes traditional arguments against narrow banking. I survey just how hopeless our current regulatory structure has become. I suggest that Pigouvian taxes provide a better structure to control debt issue than capital ratios; that banks should be 100% funded by equity, allowing downstream easy-to-fail intermediaries to tranche that equity to debt if needed. Fixed-value debt should be provided by or 100% backed by Treasury or Fed securities. ' University of Chicago Booth School of Business, Hoover Institution, NBER, and Cato Institute. I thank the CRSP and the Guggenheim Foundation for research support, and participants at the October 1, 2013 Brookings/Hoover Financial Crisis Conference for helpful comments. http://faculty.chicagobooth.edu/john.cochrane/ 1 Electronic copy available at: EFTA00611654 1. Introduction and overview At its core, ourfinancial crisis was a systemic run. The run started in the shadow banking system of overnight repurchase agreements, asset-backed securities, broker-dealer relationships, and investment banks. Arguably, it was about to spread to the large commercial banks when the Treasury Department and the Federal Reserve Board stepped in with a blanket debt guarantee and TARP (Troubled Asset Relief Program) recapitalization. But the basic economic structure of our financial crisis was the same as that of the panics and runs on demand deposits that we have seen many times before. The run defines the event as a crisis. People lost a lot of money in the 2000 tech stock bust. But there was no run, there was no crisis, and only a mild recession. Our financial system and economy could easily have handled the decline in home values and mortgage-backed security (MBS) values—which might also have been a lot smaller—had there not been a run. The central task for a regulatory response, then, should be to eliminate runs. Runs are a pathology of specific contracts, such as deposits and overnight debt, issued by specific kinds of intermediaries. Among other features, run-prone contracts promise fixed values and first-come first-served payment. There was no run in the tech stock bust because tech companies were funded by stock, and stock does not have these run-prone features. The central regulatory response to our crisis should therefore be to repair, where possible, run- prone contracts and to curtail severely those contracts that cannot be repaired. "Financial crises are everywhere and always due to problems of short-term debt" is a famous Doug Diamond (2008) aphorism, which we might amend to "and its modern cousins." Well, then, let us purge short-term debt from the system and base regulation on its remaining truly necessary uses. When they failed, Bear Stearns and Lehman Brothers were financing portfolios of mortgage- backed securities with overnight debt at 30: I leverage. For every thirty dollars of investment, every single day, they had to borrow a new twenty-nine dollars to pay back yesterday's lenders. It is not a surprise that this scheme fell apart. It is a surprise that our policy response consists of enhanced risk supervision, timid increases in bank capital ratios, fancier risk weighting, macroprudential risk regulation, security-price manipulation, a new resolution process in place of bankruptcy, tens of thousands of pages of regulations, and tens of thousands of new regulators. Wouldn't it be simpler and more effective to sharply reduce run-prone funding, at least by intermediaries likely to spark runs? In this vision, demand deposits, fixed-value money-market funds, or overnight debt must be backed entirely by short-term Treasuries. Investors who want higher returns must bear price risk. Intermediaries must raise the vast bulk of their funds for risky investments from run-proof securities. For banks, that means mostly common equity, though some long-term or other non- 2 Electronic copy available at: EFTA00611655 runnable debt can exist as well. For funds, or in the absence of substantial equity, that means shares whose values float and, ideally, are tradable. Banks can still mediate transactions, of course. For example, a bank-owned ATM machine can deliver cash by selling your shares in a Treasury-backed money market fund, stock index fund shares, or even the bank's own shares. A bank can originate and sell mortgages, if it does not want to finance those mortgages with equity or long-term debt. Banks can still be broker-dealers, custodians, derivative and swap counterparties and market makers, providers of a wide range of financial services, credit cards, and so forth. They simply may not fund themselves by issuing large amounts of run-prone debt. If a demand for separate bank debt really exists, the equity of 100 percent equity-financed banks can be held by a downstream institution or pass-through vehicle that issues equity and debt tranches. That vehicle can fail and be resolved in an hour, without disrupting any of the operations or claims against the bank, and the government can credibly commit not to bail it out. I argue that Pigouvian taxes?' provide a better structure for controlling debt than capital ratios or intensive discretionary supervision, as in stress tests. For each dollar of run-prone short-term debt issued, the bank or other intermediary must pay, say, five cents tax. Pigouvian taxes are more efficient than quantitative limits in addressing air pollution externalities, and that lesson applies to financial pollution. By taxing run-prone liabilities, those liabilities can continue to exist where and if they are truly economically important. Issuers will economize on them endogenously rather than play endless cat-and-mouse games with regulators. 1.2 Technology The essence of this vision is not novel. Proposals for narrow banking or equity-based banking have been with us about as long as runs and crashes have been with us. The "Chicago Plan," discarded in the 1930s, is only one of many such milestones; Here a second theme emerges: Modern financial, computational, and communication technology allows us to overcome the long-standing objections to narrow banking. Most deeply, "liquidity" no longer requires that people hold a large inventory of fixed-value, pay-on-demand, and hence run-prone securities. With today's technology, you could buy a cup of Pigouvian taxes are designed to discourage undesirable activities, especially externalities such as pollution. Kocherlakota (2010), Jeanne and Korinek (2011), and Perotti and Suarez (2011) suggest Pigouvian taxes to limit debt. Stein (2012) explores their equivalence to a cap-and-trade proposal. My discussion has much in common with Kotlikoff (2010), Chalmley, Kotlikoff and Polemarchakis (2012), Admati and Hellwig (2012). It also builds on Cochrane (2010, 2011). The larger points in this essay build on Cochrane (2013b). 3 Electronic copy available at: EFTA00611656 coffee by swiping a card or tapping a cell phone, selling two dollars and fifty cents of an 500 fund, and crediting the coffee seller's two dollars and fifty cents mortgage-backed security fund. If money (reserves) are involved at all—if the transaction is not simply netted among intermediaries—reserves are held for milliseconds. In the 1930s, this was not possible. We could not instantly look up the value of the 500 (communication). There was no such thing as an index fund, so stock sales faced informational illiquidity and large bid-ask spreads (financial innovation). And transactions costs would have ruled out the whole project (computation, financial innovation). Closer to current institutions, electronic transactions can easily be made with treasury-backed or floating-value money-market fund shares, in which the vast majority of transactions are simply netted by the intermediary. When you buy something, your account loses an electronic dollar and the seller's account gains one, and no security actually changes hands. On the supply end, $18 trillion of government debt is enough to back any conceivable remaining need for fixed-value default-free assets. Three trillion dollars of interest-paying reserves can easily be $6 trillion of reserves. We can live Milton Friedman's (1969) optimal quantity of money, in which the economy is awash in liquidity. This optimal quantity will have financial stability benefit far beyond its traditional elimination of shoe-leather costs. Again, technology has fundamentally changed the game: instant communication means that interest- paying money is now a reality, so we can have the optimal quantity without deflation. Our government should take over its natural monopoly position in supplying interest-paying money, just as it took over a monopoly position in supplying nineteenth-century bank notes, and for the same reason: to eliminate crises, which have the same fundamental source. The quantification of credit risk, the invention of securitized debt, long-only floating-value mutual funds, and the size and liquidity of today's markets mean that financial flows needed to finance home and business investment can come from everyday saver/investors who bear risk rather than hold traditional deposits. So, the most fundamental objection is met: that society "needs" a large stock of money-like assets, more than can be supplied by other means, so banks must try to "transform" maturity, liquidity, and risk, both to supply adequate assets for transaction-type needs and to provide adequate credit for real investment. I treat a wide range of additional common objections below. 1.3 Current policy Our current regulatory response to financial crises is based on a different basic vision that evolved piecemeal over more than a century. In order to stop runs, our government guarantees debts, implicitly or explicitly, and often ex-post with credit guarantees, bailouts, last-resort lending, and other crisis-fighting efforts. But guaranteeing debts gives the borrowers (banks and similar institutions) an incentive to take on too much asset risk and an incentive to fund those 4 EFTA00611657 risks by too much debt. It gives depositors an incentive to ignore bank risks when lending. So our government tries to regulate the riskiness of bank assets and imposes capital requirements to limit banks' debt funding. Then banks game their way around regulations, take on more risk, and skirt capital requirements; shadow banks grow up around regulations; and another crisis happens. The government guarantees more debts, expands its regulatory reach, and intensifies asset regulation. Less heralded, but no less important, this regulatory approach demands strong limits on competition and innovation, even before banks try to capture it. If regulators let new institutions circumvent regulated ones, the problems erupt again. Too big to fail means too big to lose money, and too big to lose money means too big to compete. Thus, Dodd-Frank regulation and its international cousins are not a radical new approach. They are just a natural expansion of a longstanding philosophy. Each new step follows naturally to clean up the unintended consequences of the last one. The expansion is nonetheless breathtaking. Beyond massively ramping up the intensity, scope, and detail of financial institutions and markets regulation, central banks are now trying to control the underlying market prices of assets, to keep banks from losing money in the first place. The little old lady swallowed a fly, then a spider to catch the fly, a bird to catch the spider, and so on. Horse is on the menu. Will we eat? 1.4 Comparison The insight that the crisis was a systemic run, that we can fix runs by fixing and removing run- prone financial contracts, and that new financial and communication technology addresses the classic objections, liberates us from this Rube Goldbergian (or Orwellian?) regulatory project. We do not have to fix every actual and perceived fault of the financial system in order to protect against future crises. We do not have to diagnose and correct the sources of the crisis, Fannie Mae and Freddie Mac, the community reinvestment act, so-called predatory lending, no- documentation loans, perceived global imbalances or savings gluts, Wall Street "greed," executive compensation, perceived bubbles (whether thought to be caused by irrational speculation or too-low interest rates), and so on. We do not have to fix credit card fees, disparate- impact analysis, student loans, or hedge fund fees. We don't need to micromanage over-the- counter versus exchange-traded derivatives, swap margins, position limits, the bloated Basel bank regulation mess, the definition of risk-weighted assets, the internal process and regulatory designation of and Moody ratings, the treatment of off-balance-sheet credit guarantees, and on and on and on. The thousand pages of the Volker rule alone can start a nice bonfire. If a crisis is a run, and we can remove or fix run-prone securities, none of these steps is either necessary 5 EFTA00611658 (whew) or sufficient (ouch) to stop a future crisis. A narrower regulatory approach that can stop runs, and hence crises, without requiring these Herculean (or Sisyphean?) tasks, no matter how desirable each one might be, is much more likely to succeed. If financial institutions' liabilities no longer can cause runs and crises, we don't have to try to micromanage institutions' asset choices or the market prices of those assets. Nor do we have to stop entry by new and innovative institutions. Rather than dream up a financial system so tightly controlled that no important institution ever loses money in the first place, we can simply ensure that inevitable booms and busts, losses and failures, transfer seamlessly to final investors without producing runs. Zero cost is not the standard. The financial crisis was, by most accounts, a hugely expensive event. Dodd-Frank regulation and its international cousins are not cheap, either. The challenge is only to show that my vision, which narrowly focuses on eliminating the poison in the well—run- prone assets-stops crises more effectively and costs less than these alternatives. 2. Runs and run-prone assets Demand deposits offer the paradigmatic example of a run-prone contract. If I suspect trouble at the bank, I have an incentive to get my money out first. You, seeing me run, have an incentive to get your money out before someone else. Based on this sim le description, we can sketch the essential characteristics of a run-prone security. Obviously, building on Diamond and Dybvig (1983) here: • The contract promises a fixed value, payable in full on demand or on very short notice. • Failure to pay triggers bankruptcy. Fixed-value short-maturity promises, like "lend me five dollars, I'll pay you tomorrow," invoices, trade credit, and so forth are not run-prone contracts, because one cannot force the firm into bankruptcy for failure to pay immediately. If the firm has the right to delay payment, suspend convertibility, or pay in part, it is much harder for a run to develop. Runs also require specific types of issuers. A run can't develop if the issuing institution can easily sell assets to meet creditor demands or get money elsewhere: • Runs require that the assets of the issuing institution are illiquid and cannot quickly be sold to meet redemption demands. • Runs require that the issuing institution cannot borrow or issue equity to meet redemptions. 6 EFTA00611659  These assumptions are bandied about all the time as facts. But as we think about reforming the financial system, it's important to question them. Really? Why? There are a vast number of unleveraged, deep-pocket investors around sniffing for bargains, including endowments, Warren Buffets, sovereign wealth funds, hedge funds, and so on. If a bank is really "illiquid" but not "insolvent," then just why will these investors not lend or buy equity, especially at a nice discount? Debt overhang is a common story. When a firm's value falls, the market value of its long-term debt falls, so new equity in the first instance just raises the market value of long-term debt. But banks routinely issue equity after losses, and many companies with outstanding long-term debt are able to issue equity. The deals by which the entire firm is sold to new owners over a weekend are really just equity infusions, in which presumably the new owner's overpayment to rescue debt is matched by the greater profitability of a better managed or combined company. Equity issues-especially on terms that force current equity to restore bondholders—also dilute current equity's option value for the firm's recovery and dilute the value of prospective bailouts and debt guarantees. Banks were paying dividends and big bonuses in fall 2008. Why, if they were undercapitalized? Observers suggested that they had to signal strength and retain talent. Lack of desire to issue new equity and bet the farm instead, rather than the impossibility of obtaining new equity, is a distinct possibility. Bear Stearns' existing equity holders were the ones to object to the deal, not the prospective buyers. Similarly, why can't banks sell assets? Well, they are said to find "fire sales" of unwilling buyers, but again just why are the deep-pocket investors and market-timers, usually facing the buying opportunity of a lifetime, so unwilling? Here, too, banks are often unwilling rather than unable to sell. Assets booked as "hold to maturity" can be counted at cost, not market value. Selling them forces the bank to acknowledge the loss. Selling illiquid assets can force the bank to acknowledge that actual prices are in fact less than even mark to market values, so selling one asset can depress the declared value of others. None of these are genuine economic impediments to asset sales, and could be fixed by changes in accounting and regulation. I will not pursue this line, but it certainly is worth asking just why markets for new equity or bank assets are so bad and what can be done to improve them. There is a tendency to allude to frictions, to take them as gospel, without really questioning their source, and then to design policies around them, or to exploit them, rather than questioning whether we could fix the frictions instead. A key characteristic: • A run requires that if one investor pulls out, the firm is closer to bankruptcy, giving a second investor greater incentive to pull out. 7 EFTA00611660  This is the core externality of run-prone debt. My action to pull out alters your incentives. Externalities do suggest a need for regulation, even once all the unintended disincentives and subsidies have been fixed. If bankruptcy were costless, consisting of a smooth recapitalization in which debt becomes equity the moment firm value is one cent below debt's promises, there would be little incentive to run. • Runs require significant bankruptcy costs. Without bankruptcy costs, runs would also incur little social cost. If an institution shuts down and a bunch of investors lose money, that's just a transfer unless something real is affected. For runs to be a social problem, Diamond and Dybvig (1983) imagine real projects that are abandoned after a run. • Runs are more likely if the institution's assets are nebulous and hard to value. Not for nothing have most runs been sparked by an accounting scandal or fraud. If we knew exactly what the bank's assets were worth at all times, there would be little incentive to run. Even if the assets were illiquid, lenders could always know when the bank was insolvent. The fact that illiquidity and insolvency are essentially indistinguishable in a crisis is a key component of runs. • Runs require that a substantial fraction of the firm is funded by run-prone securities. If an institution is 95 percent financed by equity, there is little chance of bankruptcy, and thus little chance of a run. That's why capital ratios are popular. Alas, the measurement of capital, the measurement of the risk and illiquidity of a bank's asset portfolio, and the effort to find an exact number—one side of which is safe and the other side of which is risky—has not proved successful. 2.1 Shadow-banking runs The concept that the financial crisis was, centrally, a run in the "shadow banking system," and the features of the financial contracts that suffered runs, are well described by Darrell Duffie (2010a, 201013) and Gary Gorton and Andrew Metrick (2012). Duffle shows how the contract structures of the shadow banking system have the same run- inducing features as conventional uninsured bank deposits. As Duffle describes, leaving securities with your broker-dealer is not like leaving your car in the repair shop, where the car transparently belongs to you if the dealer goes bankrupt. The broker-dealer may have used your securities as collateral for borrowing to fund the dealer's proprietary trading, so you cannot 8 EFTA00611661 seamlessly retrieve them after the dealer's bankruptcy. In turn, if you do retrieve your securities from the dealer, the dealer no longer has that collateral, and may have to unwind trades at a loss. Thus, if you remove your securities from a broker-dealer, the dealer is closer to bankruptcy. Derivatives contracts, though senior in bankruptcy, also cause problems for lenders and are a source of cash for the dealer. Most of all, though overnight repurchase agreements would seem exquisitely engineered to protect the lender in case of default, their protection is not in fact perfect. Some jurisdictions treat repurchases as collateralized lending, putting the borrower in line during bankruptcy. Many lenders are not legally eligible to hold the collateral, so those lenders may have to dump it immediately upon receipt. Having to unload a large portfolio of securities on the Monday afternoon of Lehman's bankruptcy is not a picnic. Better to refuse rolling over the loan on Friday. As one reads through Duffie's analysis, an insistent voice springs up, "I can fix that!" Does rehypothecation of securities make broker-dealer relations a run-prone contract? Then hold securities with a custodian. Yes, that may cost a few basis points, but those basis points are, in the end, coming from taxpayers. Being able to use client securities as collateral for proprietary trading is not such a huge social gain that it's worth going through another financial crisis. 2.2 Systemic Runs A run on an individual institution is not a crisis, however. To be a crisis, the run has to affect the financial system and, ultimately, the real economy. We need to understand what makes for "contagion" or a "systemic" run. Gorton and Metrick (2012) fill in this part of the picture. Seeing a run on institution A, investors in institution B question its finances, and are sparked to run there as well. The system as a whole promises more cash than is available, so a simultaneous run threatens systemic insolvency. In Gorton and Metrick's vision, short-term debt is normally an "information-insensitive" security. When the bank is far from default, the value of its debt, especially short-term debt, is essentially the same for a wide range of values for the bank's assets. Debt holders therefore don't need to investigate the company's finances. In turn, this feature means that the short-term debt of companies and banks far from bankruptcy is highly liquid. If I offer to sell you such debt, you don't have to worry that I know something you don't know, because nobody can really have much information about the value of such debt. As a result, bid-ask spreads, which derive from asymmetric information, are tiny. Such debt can circulate as easily as money. But it pays interest, a crucial advantage until the era of interest-paying reserves. 9 EFTA00611662  This is the point of highly rated and, especially, short-term debt. It is designed to be information-insensitive and therefore liquid, bought and sold easily with little investigation of underlying value. Once a bank is closer to bankruptcy, however, its debt becomes information-sensitive. Information about the bank's prospects changes the value of its debt substantially. Now, anyone selling the bank's debt is suspected of having information about the bank's prospects. Buyers are unwilling to take such debt without a lot of due diligence, a steep price discount, and a large bid- ask spread. Traditional buyers may not be willing to buy it at all. Traders and institutions that are not set up to do information-sensitive market-making just bow out. The sudden illiquidity of the bank's debt can lead to a rollover crisis and run long before actual next-morning bankruptcy fears become an issue. This process provides a central mechanism of perceived contagion: how trouble at one bank turns into a systemic run. Creditors, learning that one bank's or mortgage-backed security's assets are suddenly revealed to be bad, start worrying about other similar banks and securities. This shift of attitudes would not make sense if we thought of all active investors as constantly monitoring and forming opinions about the value of a bank, as we think of stock investors. But investors in short-term highly rated debt are not paying attention. That's the whole point. Until they do. Gorton (2010) tells a nice story. We usually assemble ingredients from a salad bar without investigating their individual safety. If somebody says she read a news story about E. coli in some vegetable, rather than investigate just which ingredient is risky and avoid it, it's easier to shun them all and have a hamburger instead. Chamley, Kotlikoff, and Polemarchakis (2012) tell an even more vivid story. "Eight bottles of Tylenol laced with cyanide, sold in a Chicago drugstore, instantly transformed 31 million bottles of Tylenol located in stores all over the globe into toxic assets that could find no buyers." It's just not worth investigating each bottle. The information-sensitivity story applies to securities as well as to bank debt. Long-lived securities, including highly rated tranches of mortgage-backed securities, collateralized debt obligations, and corporate bonds, are information-insensitive and therefore liquid. Cash-like liabilities of special-purpose vehicles, auction-rate securities, or other structures that issue short- term debt to hold long-lived or illiquid securities and thus emulate banks are so as well. As long as the assets are complex, illiquid, hard to value, or subject to large price shifts (important qualifications), securities of this sort can suddenly become information-sensitive and much less liquid. This process leads to the systemic "run on repo" that Gorton and Metrick (2012) document. A bank may have used information-insensitive securities, such as AAA tranches, as collateral to finance borrowing. People who want collateral only want liquid, information-insensitive securities, and they suddenly will not take the previously liquid securities as collateral, or they 10 EFTA00611663 require a large "haircut," i.e., much more collateral than the loan is worth. Now the bank is in a bind. With inadequate collateral (and inadequate equity, and all the above-cited restraints on finding new equity or new sources of borrowing), the bank must sell some of the assets it has financed at just the worst possible moment. The repo haircut works analogously to reserve requirements in generating a collateral multiplier. If the haircut rises substantially, a whole chain of debts must be unwound. The sudden illiquidity leads to a second mechanism of contagion. Suppose Bank B was holding a lot of Bank A debt, or A-type securities. When that debt or those securities suddenly become illiquid, Bank B can no longer count on selling liquid securities to raise cash to pay its creditors in case of a run. So bank or security A's problems can spark bank B's run. The sudden illiquidity means that the run-prone characteristics I listed above are heavily state- contingent. Assets that are liquid and willing equity investors can disappear quickly, so a reasonable plan to pay creditors falls apart when it is most needed. In short, when a systemic run breaks out—when one institution's or asset class's troubles bring into question many others—we see a events that are colorfully, if confusingly, described as flight to quality, fire sales, frozen markets, and illiquidity. There is a dramatic shift in total demand toward government debt and money and away from private debt. 2.3 Real effects Not only must a run be systemic to be a crisis, a crisis only matters if the systemic run has real effects. As the tech bust example made clear, and as much macroeconomic research confirms4,simple wealth effects from a decline in asset values have limited macroeconomic repercussions. In part, declines in values of existing assets are mostly redistributional: if home values fall by half, those planning to sell large houses and downsize lose, but young people can spend half as much on housing, and thus a lot more on other things. Houses are only durable goods, after all, and a decline in value by half is a different event than half of the houses in the country being washed away by a tsunami. There are two central competing stories for why the systemic run of the financial crisis had, apparently, such a large macroeconomic effect. One view thinks it's "financial constraints." The central story here is that banks in the fall of 2008 lost asset values and so were undercapitalized—the decline in asset value lowers the value 4 For example, see Ludvigson, Steindel and Lettau (2002). 11 EFTA00611664 of equity, so the ratio of equity capital to debt is now too low. In this story, banks could not raise new capital, they could not lower capital-draining dividend and bonus payments or acquisitions, they could not sell lending operations to better capitalized investors-all for various complex reasons. And despite their claims to meet regulatory requirements, they were undercapitalized by internal metrics. They resorted to asset sales and certainly were not making additional loans. In addition, banks do not matter. The banking and financial system matters. In this view, better capitalized banks or other financial institutions could not come in and take the new-lending business that existing troubled banks were abandoning and borrowers could not switch to other markets. In short, this theory goes, banks wanted to keep lending at the same rate, people and businesses wanted to keep borrowing at the same rate, but internal or regulatory capital ratios forced banks to stop lending and other banks and institutions could not fill the gap, so the otherwise healthy economy (healthy demand for investment) was starved of funds5. Bernanke (1983) really started the modern credit constraints view. In his analysis, banks in the Great Depression failed, and then the human and organizational capital that knew how to make relationship loans vanished as well. The result was a great wedge between savers and borrowers. However, perhaps partly due to the same Bernanke's actions, there simply was not a wave of bankruptcies at large commercial relationship-lending banks in 2008. There was a large wave of Federal Deposit Insurance Corporation (FDIC)-run failures at smaller banks, but the lending operations were preserved and transferred to new owners. However powerful in the Great Depression, this mechanism is not really a candidate for 2008-2009. Another view focuses on "aggregate demand." The run provoked a massive shift in demand away from private securities corresponding to physical investment and toward government debt, including money (cash and reserves) but also longer maturity government debt. But someone has to hold the existing assets, so this demand shift ends up simply changing prices. Government bond prices rose (interest rates declined) while prices on private securities dropped (risky interest rates, including low grade bonds and commercial paper, rose dramatically). This rise in interest rates, along with a similar decline in stock prices (a rise in the equity premium), a rise in risk aversion, and economic forecasts of poor conditions ahead led to a sharp drop in consumption and investment demand. Tying these ideas together, the aggregate budget constraint says that aggregate nominal demand for goods and services must add up to demand for nominal government debt. The only way to consume less and invest less is to pile up government debt. So a "flight to quality" and a "decline in aggregate demand" are the same thing. The rise in demand for money and short-term government debt, perfect substitutes at zero interest rates, is deflationary, and we saw a short, sharp deflation. Sufficient deflation might provide the sBernanke and Gertler (1989, 1995) are two classic examples of this view. Bernanke, Gertler, and Gilchrist (1999), and Gertler and Kiyotaki (2011) are two good summaries. Gambacorta and Marques- Ibanez (2011) is an excellent survey of the lending channel view of the 2008 crisis. 12 EFTA00611665 government debt people want to hold, but add some sticky prices and you have a theory of a real recession. The two mechanisms are more different than they appear. Fundamentally, the question is whether "institutional finance" or fundamental investor-based finance matters for the connection of asset prices to business cycles. In the former view, fundamental investors and borrowers such as homeowners and firms didn't change views or behavior, but the machinery connecting them broke down. In the latter view, the run fundamentally happened across the economy. Completely unlevered final investors—including endowments, pension funds, sovereign wealth funds, family offices, and so forth—panicked every bit as much as leveraged intermediaries, which is why the former were not there when the latter wanted to sell assets. Firms and households didn't want to borrow as much as before. The controversy over these questions continues, in part because the glass is surely neither completely empty nor full. Many models mix both ingredients-for example, Bemanke, Gertler, and Gilchrist (1996). There is compelling cross-sectional evidence that some businesses and people were credit-constrained in the chaos of fall 2008. But is that the key causal mechanism for the economy as a whole, or was it a distributional sideshow to a recession that would have happened anyway? Many unconstrained businesses contracted dramatically as well—and, moreover, did not expand to take over the business of the unfortunate constrained businesses. A small taste: Chart, Christian, and Kehoe (2008) document that bank lending overall did not decline early in the crisis—it declined later, when the recession was well under way. Highly rated nonfinancial companies were able to issue lots of commercial paper at low rates. They remind us that most investment comes from retained earnings and that 80 percent of corporate debt does not come from banks anyway. Ivashina and Scharfstein (2010) counter, however, that much new lending was simply borrowers drawing down credit lines and that banks in worse shape cut lending more than banks in better shape. But this is a long way from showing that the financial system as a whole refused to lend to clearly profitable new investments. Dell'Ariccia, Detragiache, and Rajan (2008) show that sectors more dependent on external finance fall more during the recession following a financial crisis. But sectors not at all so dependent also fall, and external financial dependency is endogenous, which one can never fully control for. Cochrane (2011) argues for the view that a coordinated rise in risk premiums, even in completely un- intermediated markets and even after intermediation frictions washed away, was the major characteristic of financial markets. All of this matters, for several reasons. Again, if crises do not cause recessions or other real economic damage, we don't care. Well, crises do cause real damage, but why they cause economic damage illuminates what a crisis is in the first place, what steps we might take to avoid crises, how costly crises really are, and thus how much cost we should tolerate from anti-crisis policies, which put-out-the-fire policies are important to contain economic damage, and what structural reforms we might undertake to limit economic damage. 13 EFTA00611666  If the essential link from run to macroeconomy is undercapitalized intermediaries, then the final form of the TARP—recapitalizing intermediaries-should have solved the macroeconomic problem. That it did not do so is one piece of evidence in my mind against this view, but one can always argue that even more capital was needed. If the essential link from run to macroeconomy is a flight to quality, then lender-of-last-resort institutions and massive exchanges of government debt for private debt are more promising fire extinguishers. The lender-of-last-resort theory has always preached to accommodate shifts in money demand versus other assets by large lending, with the limitations of "against good collateral" and "at a penalty rate" open to question. One can argue that, in our crisis, the government should simply have exchanged trillions (more) of government debt for trillions of private securities that people did not want to hold. Of course, the question of "good collateral" and the moral hazard of this extreme central-banker's put option will limit the idea. Both views, I think, pose an unresolved challenge to the project of an expensive regulatory response. If, as Keynesians believe, aggregate demand is the only link from anything to recessions, and it can easily be managed with sufficient fiscal and monetary stimulus, then in fact we only need to fix macro policy and leave the banks alone; crises really are not by themselves a social problem. (I don't see any Keynesians advocating this position, but it is a logical consequence of belief in aggregate demand and stimulus.) If the main link from crisis to real activity is a sort of clogging of the arteries of existing banks, with layers and layers of additional frictions required to keep still-optimistic investors from getting money to still-optimistic firms, then one could suffer runs far more easily with faster and more effective recapitalization. In both views, we should be paying a lot more attention to the hazily-described "frictions." If price stickiness is the bottom of recessions, then why are economists who write such models not outraged at the government's efforts to make prices and wages stickier, and why are they not campaigning to fix price stickiness? If inadequate capital and frictions in the way of private recapitalization are the key problem, fixing capital-raising frictions would seem to be at the top of the agenda, rather than amassing larger options for ex-post bailouts and recapitalizations. Aside from that weakness, however, both views point to large macroeconomic effects and important benefits from stopping systemic runs in the first place. Those who think undercapitalized banks led us to five years of GDP falling $1 trillion below potential should be in the front of the queue demanding much, much more capital to begin with. 14 EFTA00611667 2.4 Stock again Common equity is the paradigmatic example of a corporate liability that is immune from runs. When an equity-funded company is in trouble, you can try to sell stock, and stock values can crash. But you cannot run to the company and demand your money back and you cannot drive the company into bankruptcy should it fail to pay. When you sell stock, you do not do anything to push the company closer to insolvency. Seeing your investment crash, your neighbor, invested in another company, can't demand his money back and cause that company to fail either. One might feel that stock price crashes represent fire sales, irrational fads, or otherwise socially suboptimal phenomena. Stock market crashes can come with important shifts in investment and other economic outcomes. But stock market crashes are not runs and they are not crises. Investors bear risk for their returns directly and inescapably. My decision to sell—even if unwise, even if it provokes a price decline—does not pose an externality to you. If anything, the opposite is true: my fire sale is your buying opportunity. An exchange-traded fund (ETF) is a paradigmatic example of a run-proof intermediary. When asset values fall, the liability values of an exchange-traded fund fall automatically and the fund itself cannot go bankrupt. For this reason, the tech boom and bust of the late 1990s is a good comparison. The decline in tech stock value was similar to the decline in subprime mortgage-backed security value by September 2008. But there was no crisis and only a mild recession in the early 2000s. Why? Because tech losses were held in stock, and when stock falls you can't run. Housing losses were held in fragile, run-prone securities. Yes, the Great Depression was heralded by a stock market crash. But financial turmoil came from defaults by intermediaries who had borrowed to invest in stock and by the subsequent bank runs and bank failures. These are failures of debt, not of equity. Long-term debt occupies an intermediate spot. It does make fixed promises, but only at periodic intervals. On suspecting bad news, like equity, there is nothing you can do immediately to avoid losses. Long-term debt leads to crises when it isn't quite long enough, when it needs to be refinanced in large chunks. The Greek debt crisis came when Greece needed to refinance long-term debt, not when it could not borrow for one year's spending. 15 EFTA00611668 2.5 Things that don't matter The view that the financial crisis was, at heart, a systemic run is as—or more—important for clarifying things that are not important in the quest to avoid another crisis as it is for clarifying what is important. The source of losses does not matter. So fixing the (deplorable, in my view) federal interventions in housing finance, or fixing the various abuses, predations, and irrational behavior others see in housing, is neither necessary nor sufficient to stop crises. The next panic may start with losses on sovereign debts: perhaps a new European crisis or a US state debt and pension crisis. The whole long argument over whether supposed global imbalances can fuel a savings glut, and whether policymakers are wise enough to detect and prevent such things, is pointless. If such events exist but savings are held in floating-value non-runnable assets, then, yes, big price drops (buying opportunities!) and exchange rate changes can happen—but not a crisis. The "dominoes" or "interconnectedness" theory is a popular alternative view of a crisis: A defaults on its debts to B, so B defaults on its debts to C, and so forth. Much regulation is written against the dominoes theory—for example, the new limits on single-party exposures. But dominoes were not a major factor in our crisis or in previous crises. Companies build buffers against dominoes. If A's default is to cause B, C, and D to default, A's loss of value must exceed the combined capital and borrowing capacity of A, B, C, and D. But what happened to us was that seeing A fail, investors ran on B, C, and D, who had little unhedged direct exposure to A. The systemic run view argues that lots of "interconnectedness" regulation is fairly pointless. 3. Stopping Runs 3.1 An end to run-prone financing If the problem is runs, and runs are identifiable features of certain kinds of contracts issued by certain kinds of intermediaries, then the focus of regulation should rather naturally be to fix run- prone contracts where possible and to strongly discourage their use when they can't be fixed. For commercial banks, the answer is pretty simple: equity, lots more equity. How much? Well, more is better, and "enough so that it doesn't matter" or "enough that we never, ever hear again the call 'recapitalize the banks"' are good answers. 100 percent is perfectly workable. More is obviously better, because more capital puts banks further from bankruptcy and further from a run to begin with. Less obviously, all of the dynamic problems sparking runs are ameliorated by capital. If a bank has 2 percent equity capital, loses I percent of the value of its assets, and (as the story goes) cannot quickly issue more equity to rebuild capital, then it has to 16 EFTA00611669 sell half of its assets to restore its capital ratio. If the bank has instead 50 percent equity capital and loses I percent of the value of its assets, it only needs to sell 2 percent of its assets to get back to a 50 percent capital ratio. And if the bank is 100 percent equity-financed, it doesn't have to sell anything. If you worry about fire sales, you should like equity. Similarly, if a bank has 2 percent equity capital and loses 1 percent of the value of its assets, it is one more such loss away from bankruptcy. Debt overhang will surely loom over any equity issues, recapitalization, or effort to sell the company. If a bank has 50 percent equity capital and loses 1 percent of the value of its assets, it is so far from bankruptcy that issuing more equity involves essentially no transfers to debt holders and can be accomplished seamlessly. If a bank is 100 percent equity-financed, it doesn't have to issue any more equity at all in response to losses. If you worry about the difficulty of issuing or hanging on to equity, you should like more equity to start with. Many recent proposals specify debt that converts to equity under some circumstances. Examples include French et al. (2010a) and Hart and Zingales (2011). Some convertible debt has been issued. More simply, banks could be required to buy put options, giving them the right to issue new equity at predetermined prices. However, critics such as Admati and Hellwig (2012) point out, why bother? Why bother with the fig leaf that the convertible security is debt rather than equity or really an equity option? The answers mostly do not come from fundamental economic problems: to keep the tax-deductibility of interest payments, for example, or to address accounting rules. Well, if these are why we have runs, they're not that hard to fix. In addition, rumors that a conversion option might be triggered