The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
3 T w o Faces of Financial Innovation 18 The Recent Credit Crunch: The Neglected Dim ensions 37 Structural Approaches to Vector Autoregressions 58 The Misuse of the Fed’s Discount W in d ow THE FEDERAL A RESERVE JZkRANK of A r ST.IXH IS 1 Federal Reserve Bank of St. Louis R e v ie w September/October 1992 In This Issue . . . Innovation has always been a hallmark o f the financial services indus try. Recent innovations range from new mechanical devices and new contractual arrangements to new operational procedures. In response to the pace and diversity o f financial innovation, researchers have studied the econom ics o f this innovation. One approach focuses on improving financial products by applying the principles o f finance theory to the process o f contract design in securities markets. A second approach fo cuses on the reasons why financial innovation occurs, examining the in centives for people to develop new financial contracts or technologies. In the first article in this Review, "T w o Faces o f Financial Innovation," Mark D. Flood uses tw o case studies to illustrate how financial innova tions are developed to meet a perceived demand for new financial serv ices. The tw o cases presented illustrate h ow market forces can spell failure for product designs that do not incorporate the principles of financial theory and success for those that do. * * * Although the notion o f a credit crunch is not new, its widespread use with reference to the latest recession and recovery has been the con ventional wisdom in many policy circles. In the second article in this is sue, "The Recent Credit Crunch: The Neglected Dimensions,” Kevin L. Kliesen and John A. Tatom take a historical view o f credit crunches and their relevance to the nation’s cyclical perform ance. They explain the origins o f the credit crunch concept and point to the consistent misap plication o f the concept to cyclical credit market developments. For ex ample, they show that movements o f interest rates and o f interest rate spreads in recessions generally have not provided support for the credit crunch hypothesis. The most recent case is no exception. According to the authors, the recent decline in the growth o f credit is mostly attributable to tw o factors that are inextricably linked. First, the demand for credit, especially by business, is cyclical. W hen econom ic ac tivity ebbs in a recession, the demand for credit falls as well. Second, because bank lending to business is typically short term, it is used to finance short-term assets such as inventory. Thus when businesses ex pect their sales to slow or decline in a recession, they no longer need to add to their stocks o f inventory. Accordingly, business inventories and the demand for bank credit to finance them are reduced. Similarly, in the early stages o f an econom ic expansion, when businesses add to in ventory holdings, they typically rely relatively more on internal cash flow and other sources o f financing, than on bank loans. Only later do businesses begin to expand bank borrow ing in line with their burgeoning financial requirements. SEPTEMBER/OCTOBER 1992 2 The authors maintain that in the final analysis, the conventional wisdom espoused by credit crunch theorists is not helpful in assessing recent movements o f interest rates, business loans and business inventories. * * * Vector autoregressive (VAR) models are frequently used to estimate dynamic relationships for econom ic data. Initially the VAR was not in tended to describe structural mechanisms because it represents a reduced form. An econom ic interpretation o f these empirical models, however, does require a set o f structural assumptions. In the third article in this issue, "Structural Approaches to Vector Autoregressions,” John W. Keating reviews the principles o f VAR analysis with particular emphasis on tw o structural VAR approaches. The two approaches impose restrictions on contemporaneous and long-run be havior to identify a structural model from the reduced form. The author provides empirical examples with a standard group o f m acroeco nomic variables to illustrate these tw o approaches. The empirical longrun structural model yields results that are generally consistent with the theoretical model, whereas the results for the contem poraneous model are frequently inconsistent with theory. These findings suggest that long-run structural VAR models are superior; however, additional inves tigation should determine whether this result is robust or simply a spe cial case. * * * In the final article in this Review, "The Misuse o f the Fed’s Discount W indow ,” Anna J. Schwartz, a senior research associate at the National Bureau o f Economic Research, asks not only whether the Fed's discount w indow continues to serve any useful function, but also whether its operation may have unintended adverse effects. The discount rate, for example, draws widespread—but in Schwartz's view, misdirected— attention as an indicator o f the thrust o f monetary policy. Schwartz is also concerned about the many, albeit unsuccessful, attempts to extend the Fed’s lender-of-last-resort function to nonbank firms. Most im por tant, however, is her concern that in recent years the discount window has extended credit to banks that w ere insolvent or near insolvency—a practice that she argues has increased taxpayer losses associated with reimbursing insured depositors at failed banks. Although many analysts will find some o f her arguments controversial, Schwartz raises a num ber o f issues w orthy o f further research. FEDERAL RESERVE BANK OF ST. LOUIS * * * 3 Mark D. Flood Mark D. Flood is an economist at the Federal Reserve Bank of St. Louis. David H. Kelly and James P. Kelley provided research assistance. Two Faces of Financial Innovation I n n o v a t io n h a s ALWAYS been a hallmark o f the financial services industry. Indeed, the history o f finance can be organized as a chroni cle of innovations. W e can trace this history from the introduction o f coinage in the Greek state o f Lydia in the 7th century B.C., through the various ploys to circumvent the Christian and Islamic bans on usury in the medieval era, through the development o f modern systems o f insurance in the 18th and 19th centuries, on up to m ore timely innovations such as foreign cur rency exchange warrants and interest-rate swaps.1 Recent innovations have ranged from the comm onplace — the automatic teller machine (ATM), for example — to the arcane — exchangetraded options to buy futures contracts on municipal bond index funds. Indeed, the range o f innovation, from new mechanical devices to new contractual arrangements to new opera tional procedures, is so broad as to confound concise definition o f the term.2 M oreover, inno 1See Goldsmith (1987) for some descriptions of financial in stitutions and instruments throughout history. 2Schumpeter (1939) defines an innovation as a change in the shape of the production function. He goes on to categorize innovations as being either “ process” or “ product.” Process refers to innovations that permit an ex isting product or service to be provided more cheaply. Product refers to innovations that introduce a product or service that was previously unavailable. The ATM, for ex ample, is a mixture of both types. An ATM provides many routine services, such as accepting deposits and disburs- vations have recently arrived at a frenetic pace; in the narrow field o f exchange-traded futures, for example, Silber (1981) lists 102 new con tracts introduced in the United States in the 1970s alone. In response to the pace and diversity o f finan cial innovation, economists have studied the eco nomics o f innovation. Tw o largely distinct — but not necessarily inconsistent — approaches to the subject have developed over the last 20 years. One approach, motivated by the imperfect suc cess rate fo r new securities, has focused on improving financial products by applying the principles o f finance theory to the process of contract design in securities markets.3 The sec ond approach, motivated largely by central bankers’ concerns about the effect o f innova tions on monetary policy, has focused on the reasons w hy financial innovation occurs. This approach has examined the incentives for people to develop new financial contracts or technolo gies.4 Using tw o case studies, this paper illusing withdrawals, and usually does so more cheaply. At the same time, an ATM can provide services that were previ ously unavailable, such as nighttime and weekend with drawals. 3Some examples are Allen and Gale (1988), Chen and Kensinger (1990), Johnston and McConnell (1989), and Silber (1981). 4Some examples are Goodhart (1986), Kane (1984), Podolski (1986), Rasche (1988), and Simpson (1984). SEPTEMBER/OCTOBER 1992 4 trates h ow financial innovations arise to meet a perceived demand for new financial services. The contrasting experience o f the tw o cases shows how market forces can spell failure for product designs that do not attend to the princi ples o f financial theory and success for those that do. W H Y IN N O VA TIO N OCCURS In general, individuals do not innovate out of a spirit o f magnanimity. Indeed, w e shall assume, as we do with other econom ic behavior, that financial innovations are created in anticipation o f material gain. Most theories o f the incentives to innovate can be understood in terms o f a cost-benefit analysis: new potential profits are the incentives to innovate. These arise when a change occurs that makes possible either a reduction in costs, an increase in revenues, or both. For simplicity, such changes are usually treated as occurring exogenously to the finan cial services industry, even when this depiction is not entirely accurate. On the cost-reduction side o f the cost-benefit in terpretation, exogenous technological change is the force most often cited as producing the potential cost reductions that can induce innova tion. As w e shall see below, the transaction costs and the costs o f market illiquidity are two factors that frequently affect the production and success o f financial innovations. Advances in computing pow er, for example, have lowered the cost o f such accounting-intensive products as brokered deposits and mutual funds. Other products relying on rapid calculation and deci sion, such as portfolio insurance and index arbi trage transactions, have similarly been made feasible by increases in com puter speed. The ATM, w hich reduces bank operating costs by ef ficiently executing much o f a teller’s drudgery, was made possible by gains in both computing pow er and miniaturization. Some o f these innovations can also illustrate the other side o f the incentive to innovate — the potential for increased revenues. Activities like index arbitrage w ould be inconceivable without some form o f reliable, high-speed calculation; computers can thus be seen either as reducing 5See, for example, Kane (1984) or Miller (1986). 6See Marton (1984), p. 239. Alchian (1977), pp. 30-32, dis cusses the role of trial and error in the process of innova tion, and he equates success with survival. He notes (p. 31) that “ the available evidence [that the necessary condi RESERVE BANK OF ST. LOUIS FEDERAL the potential cost o f the activity from infinity or as increasing potential revenues above zero. More commonly cited as forces that can generate potential sources o f new revenue are government policy and inflation.5 The tax code, in particular, offers numerous incentives. Tax payers innovate to exploit loopholes and avoid assessments (for example, tax-free municipal bond mutual funds). If taxpayers are successful in using financial innovations to low er their tax bills, government may seek to re-legislate to close loopholes and expand the scope o f taxa tion. Completing the cycle — in what Kane (1977) has called the "regulatory dialectic" — taxpayers find new incentives in the revised tax code and innovate again. Broad-based m acroeco nomic factors can also motivate innovation. For example, inflation com bined with deposit in terest rate ceilings in the late 1970s and early 1980s to produce new kinds o f bank deposits — super-NOW accounts and money-market mutual funds — w hich function essentially as interestbearing checking accounts. M ore recently, the Coffee, Sugar and Cocoa Exchange experiment ed unsuccessfully with consum er price index (CPI) futures, w hich w ere to hav^ allowed inves tors to hedge against inflation. W H Y PA R TIC U LAR IN N O VATION S SUCCEED Creators o f financial innovations are obviously interested in whether their innovations will suc ceed in the marketplace. Success, o f course, is not automatic. For example, as Marton reports, "some exchange officials privately admit that they put out new futures products pretty much the way a cook tests his spaghetti strands — by flinging them against the wall to see if any of them stick.”6 Because innovation is not costless, however, it is important that innovators have a m ore systematic approach to innovation than primitive trial and error. To examine the successfulness o f specific in novations, w e must first have some way of measuring success. Because w e have assumed that people innovate in hopes o f profit, a mea sure o f success should either measure the innotions are met under which trial and error converges to a profit-maximizing equilibrium] seems overwhelmingly un favorable.” For our purposes, defining success as survival is not precise enough to be useful, since even failures will survive at least briefly before their demise. 5 vator’s profits directly or, at least, be correlated with those profits. In many cases, a measure of popularity can proxy for success. For example, an exchange that introduces a new security might measure success by dollar trading volume in the new contract. Given a measure for suc cess, innovators can set about designing a pro duct to maximize that measure. Some o f the factors leading to a successful in novation are illustrated in the following exam ple. The economist Milton Friedman, anticipating a devaluation o f the British pound in the early 1970s, discovered he could not speculate on his beliefs, because no bank would allow him to sell the pound short. Hearing o f his plight (and presuming that his predicament was not unique), Leo Melamed and the Chicago Mercan tile Exchange (CME) launched the International Monetary Market (IMM). The IMM trades, among other things, foreign currency futures, which al low investors to speculate on devaluations (or appreciations) o f the pound. This innovation has been successful because the CME found a trade that investors wanted to make, but could not, and it devised an instrument that allows them to do so cheaply and reliably (an exchangetraded foreign currency futures contract).7 subtleties can play a role in determining the success or failure o f a new financial product. These issues play a prominent role in our first case study. Nonetheless, some rules o f thumb do exist for identifying likely successes in futures markets: contracts based on commodities that are easily standardized, have large price volatility, and have enough suppliers and demanders to create a liquid market.10 These rules are far from fool proof, however. Moreover, they do not general ize automatically to other types o f innovations. Indeed, codification o f the elements o f success ful innovation as a collection o f rules is almost certainly impossible; it is also beyond the scope o f this paper. Instead, w e illustrate with tw o case studies the more general ideas o f the in centives to innovate and the application of financial principles to real innovations. CASE 1, A FAILURE: CAN AD IAN COIN FUTURES The first innovation we consider is a futures contract on bagged Canadian silver coins, in troduced by the IMM. This innovation was a failure. After 13 months o f meager trading, the IMM discontinued the unpopular and unprofit able contract. W hy did this contract fail? There is an answer that is consistent with both our theoretical rationales for successful financial in novation and with the facts: A good cross-hedge existed in the much more liquid silver futures market. There, hedgers could achieve similar results at low er cost. A m ore general application o f this recipe for success must answer several questions. For ex ample, how is an innovator to know w hich as yet non-existent security investors want to trade? A full answer to this requires considera ble insight into investor demand. Famous economists may offer suggestions, but this process is not always reliable. Milton Friedman, for example, also advocated the failed CPI fu tures contract.8 A Description o f the Instrument A secondary problem is knowing whether in vestors can already make a certain type o f trade. While the answer might seem readily obtainable, the existence o f substitutes is not always obvi ous. In the futures markets, for example, hedg ing a com modity position with a futures contract on a close substitute (cross-hedging) may be adequate or even superior to hedging directly.9 Identifying these possibilities and how they might be used can be quite subtle. Such On October 1, 1973, the IMM opened trading in a new futures contract on Canadian silver coins. The purchaser o f a contract promised to pay a certain future amount in U. S. dollars (USD) at a specific future maturity date; in ex change, the purchaser would receive future delivery o f five bags o f Canadian silver coins (dimes, quarters or half-dollars), with each bag w orth 1,000 Canadian dollars (CAD) at face value. Different denominations could not be 7See Miller (1986), p. 464. Similar speculation was possible in the interbank market for forward foreign exchange. In dividual investors, like Friedman, are generally excluded from this market, however. Innovations like this, that expand the possibilities for ex change, are generally Pareto-improving. That is, they can make everyone involved better off. See, for example, Flood (1991). 9A useful concept in this regard is “ redundancy.” If the price of a good is always a fixed multiple of the price of another good, so that the price changes for the two are al ways perfectly correlated, then one of the goods is said to be redundant. '°See Black (1986), pp. 5-12. 8See Friedman (1984). SEPTEMBER/OCTOBER 1992 6 mixed within a bag, and the coins w ere to have been minted before 1967, with the exception that a bag could contain up to 2 percent 1967 or 1968 coins containing at least 50 percent sil ver.11 words, we can explain failure by establishing the existence o f an efficient cross-hedge for the futures contract. It turns out that, subject to some caveats, we can demonstrate that the coinage contract was indeed redundant. The contract was presumably intended to af ford commercial banks holding vault inventories o f silver coinage the opportunity to hedge their positions. In the mid-1960s, both the United States and Canada had phased out the use o f sil ver in their coinage. In an application of Gresham's Law — "Bad m oney drives out good” — both banks and private investors hoarded sil ver coins for the bullion content, using instead the new, non-silver coinage as a transactions medium.12 In doing this, it would be helpful to under stand theoretically w hy the price o f the cross hedging instrument should be correlated with the price of the coinage futures contract. With this in mind, we examine the dual role (as either silver or currency) o f the coins more closely. The hoarded Canadian silver coins might be valuable for tw o reasons. First, they w ere Cana dian currency and, therefore, could be ex changed for goods and services in Canada. Second, they contained substantial amounts of silver, a mineral with numerous industrial uses.13 Therefore, their USD value could fluctuate with changes in the USD/CAD exchange rate or with changes in the USD price o f silver. Banks wish ing to protect themselves against such price fluctuations could simply sell their stores o f the coins, or they could sell the new futures con tracts, which would lock in a future sale price for their inventory. As w e shall see, however, indirect hedges w ere also available. Redundancy W e are interested in factors contributing to the failure o f the IMM’s futures contract on the coins. One way to explain this failure w ould be to find some other security that provides the same risk allocation m ore cheaply. In other 11See IMM (1973a), p. 22. Because the payoffs described here are all in U. S. dollars, it may be helpful to imagine that the hedging institution is a U. S. bank. Regardless, the nationality of the parties involved should not affect in any way the pricing relationships discussed below. 12The new coins consisted of a copper core clad in a copper-nickel alloy; they contained no silver. The Board of Governors of the Federal Reserve (1970) ruled that vault in ventories of U. S. silver coins held by commercial banks for their own account could still be counted as part of required reserves, even though they were being hoarded and there fore would not circulate. Coins to which the bank did not have “ the full and unrestricted right” — for example, coins held for safekeeping in the name of a speculating deposi tor — would not satisfy reserve requirements. 13IMM (1973a), p. 20, suggested that the numismatic value of the coins might also be a factor. The possibility that bagged coins might have significant numismatic value is http://fraser.stlouisfed.org/ FEDERAL RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis At any time, the coins could be used for one of tw o purposes: as a stock o f raw silver for in dustrial purposes, or as a medium o f exchange for transaction purposes. The coins clearly could not be used for both purposes simulta neously: industrial use w ould require melting the coins; monetary use would require not melting them. Instead, if the coins w ere rem oved from bank vaults and put to use, w e should expect them to go to the m ore valuable o f the tw o uses. These relationships are presented in figure 1. This surface is a plot o f the value o f coins as a function o f the USD/CAD exchange rate and the USD value o f silver. Given coordinates for the USD value o f silver and the USD value o f Cana dian dollars, the height o f the surface at that point is the value o f the bagged Canadian coins. The graph formalizes the notion that the silver coins are worth the larger o f tw o values: their value as silver or Canadian currency. The axes are scaled to correspond to contract specifica tions in the futures markets. The units on the CAD axis give the USD value o f CAD 100,000, the contract size for a foreign exchange futures contract. Similarly, the silver axis gives the USD value o f 5,000 troy oz. o f silver, the contract size for a silver futures contract at the Chicago remote, however. Collectors grade coins according to quali ty (e. g., proof, uncirculated, very fine, etc.). With rare ex ceptions, coins of recent minting are priced above face value, only if they are rated as proof or uncirculated. The fact that the coins subject to the futures contract were poured loose into bags implies that they could not be rated proof or uncirculated. Indeed, the “sweating” of coins, i. e., shaking them loose in a bag to rub off silver shavings, is a traditional means of debasing a currency. The ABA (1965), p. 8, for example, claimed that it was “obvious that it will be years and years before these coins have any value as collectors’ items.” 7 Figure 1 Theoretical Relationship Between Spot Prices USD per 5 bags o f coins 200000 150000 100000 USD per CAD 100,000 50000 USD per 5,000 oz. silver Board o f Trade (CBOT). The surface is a graph o f the function: Vct = max {0.05Vdt, 0.579Vst}, where \£, is the USD value o f the five bags o f coins at time t, Vdt is the USD value o f CAD 100,000, and Vst is the USD value o f 5,000 troy oz. o f silver.14 While this equation describes the relationships among spot prices for the three commodities, it 14The coefficients, 0.05 and 0.579, are simply the proportions of CAD 100,000 and 5,000 troy oz. of silver, respectively, present in five bags of coins. The contract requires that the coins are worth CAD 5,000 at face value, or 5 percent of CAD 100,000. The contract also specifies that “ The coins must bear a minting date of 1966 or earlier, except that an individual bag may contain up to 2% 1967 or 1968 coins containing at least 50% silver ... The gross weight of each bag, including bag, seal, and tag shall not be less than 50.75 pounds [avoirdupois] for each [CAD] $1,000 face amount;” emphasis in the original, IMM (1973b), p. 2. (Coins minted prior to 1967 were 80 percent silver by weight.) Deducting 0.75 pounds for the weight of the bag, and assuming that banks applied Gresham’s law, that is, they chose their coins to minimize does not address the price o f the coinage fu tures contract directly. Tw o facts are relevant to our investigation at this point: First, futures contracts for both Canadian dollars and silver w ere actively traded at the time; second, the price o f any futures contract must converge to the corresponding com m odity spot price on the maturity date o f the futures contract. The latter condition holds because, at maturity, arbitrage ensures that a contract for future delivery ef fectively becom es a contract for spot delivery.15 the silver content of each bag within the constraints of the contract, we find that each bag (CAD 1,000 face value) contained 39.7 lbs., or 578.9 troy oz. of silver. Thus, five bags contained 57.89 percent of 5,000 troy oz. of silver. 15For example, if the price of a maturing futures contract sig nificantly exceeded the spot commodity price, then an ar bitrageur could make unlimited, riskless profits by selling futures, buying spot and delivering the commodity. There is some margin for price discrepancies: this arbitrage is only profitable if the price difference exceeds the costs of transaction and delivery. Still, only one arbitrageur is need ed to enforce the arbitrage; the delivery and transaction costs for the least-cost arbitrageur should be the only rele vant ones in this context. SEPTEMBER/OCTOBER 1992 8 Figure 2 Daily Spot Prices of Silver and Canadian Dollars U.S. $ per 100,000 Canadian dollars U.S. $ per 100,000 Canadian dollars $ U.S. per 5,000 oz. Silver Given these relationships, w e can restate our equation (approximately) in terms o f the matur ing futures contracts: VcT ~ max {0.0 51^T, 0.579Kr}< w here VcT is the USD value o f the coinage fu tures at the maturity date, T, VdT is the USD value o f CAD futures, and Kt is the USD value o f the silver futures. Assuming that delivery and transaction costs are negligible makes this equation exact: l£x = max {0.05l£T, 0.5791{X}. This equation holds only at the maturity date of the futures contracts, however. At maturity, there is no uncertainty about which is larger, 0.05ViT or 0.579!£t . Before maturity, l£T and are uncertain.16 As it turns out, this latter complication is largely academic. In hindsight, over the life o f 16We can still establish certain relationships between the prices, however. For times t < T, the possibility that the coins might be more valuable as currency than as silver provides a price floor for the coinage futures in terms of the silver futures: Vc, > 0.579Vst. Similarly, we have a price floor for the coinage futures in terms of Canadian dollar fu tures: Vct > 0.05Vdt. Finally, because holding both silver and Canadian dollars at maturity must always be prefera- FEDERAL RESERVE BANK OF ST. LOUIS the coinage futures contracts, the coins w ere never m ore valuable as currency than they w ere as silver. Figure 2 graphs daily observations (scaled as in figure 1) o f spot silver prices and the USD/CAD spot exchange rate for October 1973 through June 1974. The line in the graph corresponds to the crease in the surface o f fig ure l . 17 The fact that all points fall to the right o f the line means that max{0.05Vdt, 0.579Vst} was always 0.579VS, (that is, that 0.579Vsl > 0.05Vdt). More significantly, it appears that, even b e fore the fact, investors considered an outcome to the left o f the line to be highly unlikely. If the probability o f such an outcom e is negligible (that is, if the probability is small that 0.05Vdt > 0.579VJ, then it is safe to use the ap proximation: max{0.05Vdt, 0.579Vst} ~ 0.579Vst. M oreover, if this approximation always holds in the spot markets, it should also hold for futures ble to holding only the more valuable of the two, we can readily establish a price ceiling: Vct < 0.579Vsl + 0.05Vdt. 17The crease in figure 1 is the set of points satisfying both Vc = 0.05Vd, and Vct = 0.579Vst. Thus, it is the line defined by 0.05Vdt = Vct = 0.579Vst, or Vdt = (0.579/0.05)Vst. This is the line graphed in figure 2. 9 Figure 3 Daily Futures Prices for Silver and Canadian Coins Price of Canadian coinage futures Price of silver futures 0 0 0 contracts; this implies w e can use the approxi mation Va ~ 0.579V„. The upshot o f this is that, as long as investors could safely ignore the possibility that the coins would be m ore valuable as currency, a silver futures contract was a good cross-hedge for the coins. In other words, the value o f the coins would be determined solely by their silver con tent, rather than their potential use as Canadian currency — the value o f the coins should move in tandem with silver prices. Conversely, if all the points w ere to lie on the left side o f the line in figure 2 instead o f the right, the opposite condition would hold: the CAD futures con tract would be the appropriate cross-hedge to June contracts O O O December contracts consider, and w e could disregard the silver contract. The usefulness o f silver futures as a crosshedge is confirm ed in figures 3 and 4. Figure 3 plots daily price observations for silver futures and Canadian coinage futures maturing in June 1974 and Decem ber 1974. The bold line through the origin is the theoretical relationship under redundancy: Va = 0.579 The other tw o lines are regression lines (lines o f best fit) for the tw o different maturities.18 Given our regression results, w e can be more specific about the effectiveness o f a cross-hedge. The coefficient o f determination, or R2 statistic, from such a regression is a standard measure of 18These regressions show that the behavior of silver futures prices and that of coinage futures prices are very similar, implying that these two potential hedging instruments are essentially indistinguishable from a pricing standpoint. An alternative would have been to demonstrate their hedging effectiveness directly, by regressing spot coin prices first on coinage futures prices and then on silver futures prices. A spot price series for Canadian silver coins was not avail able, however. SEPTEMBER/OCTOBER 1992 10 Figure 4 Daily Prices for December 1974 Silver and Canadian Coin Futures Price of Canadian coinage futures the effectiveness o f a hedge. The R2 statistic can vary from 0.00 for a useless hedge to 1.00 for a perfect hedge. For the regressions on the June and Decem ber contracts, the R2 statistics w ere 0.973 and 0.933, respectively.19 The implication is that silver futures m oved closely together with coinage futures, and thus could serve as an excellent cross-hedge: Investors could have achieved almost identical results with the silver contract as with the coinage futures. This is confirm ed in figure 4, which shows the same relationship plotted as time series o f the two prices. Silver and coinage futures prices moved in near lockstep.20 19We can compare these numbers to the hedging effective ness of CAD futures. As expected, a regression of the price of coinage futures on the price of CAD futures yields lower R2 statistics, implying that silver futures were a better cross-hedge. The Ft2 was 0.568 for the June contracts and 0.715 for December. The use of the R2 originated with Ederington (1979), pp. 163-64. The standard measure is based on regressions of returns on returns, rather than prices on prices, as are shown in the figure. Because the coinage futures had no recorded price on most days (due to no trading), however, it was not possible to construct a daily return series. An imperfect alternative is to calculate returns from one price observation to the next, producing a time series of returns with irregular holding periods. The R2 statistics for these FEDERAL RESERVE BANK OF ST. LOUIS Price of silver futures Performance o f the Instrument Having established the existence o f a service able cross-hedge for the Canadian coinage fu tures contract, w e now examine the contract from the perspective o f the innovator. From this perspective, the contract is successful to the extent that it profits the innovator. The in novator in this case was an organized futures exchange, the IMM, ow ned by the members of the exchange. The membership o f a futures ex change consists o f its traders, w h o benefit from an increase in trading volume through higher regressions are 0.385 for the June contract and 0.419 for the December contract. In comparison, Johnston and McConnell (1989) report on Government National Mortgage Association (GNMA) Col lateralized Depository Receipt (CDR) futures contracts. Dur ing the most successful years of the contract, 1980-82, the hedging effectiveness (R2) of the futures contract for the underlying GNMA securities over five-day holding periods ranged from .85 to .94. During the contract’s leanest years, 1983-85, the R2 ranged from .54 to .62 percent, making it a less effective hedge than a Treasury bond futures contract. 20Because of low volume in the coinage futures market, there were many days with no bids or offers submitted and, therefore, no posted settlement prices. 11 turnover or through greater liquidity.21 For our purposes, w e can proxy for profits by measur ing trading volume in the contract. Illiquidity can be costly to traders in two ways. First, traders may be forced to search ac tively for counterparties with w hom to trade; searching consumes time and resources. Second, traders w ho are delayed by the search process face price risk: the price o f their commodity can change before they locate a counterparty. In general, a trader would require some re muneration before bearing such a risk willingly. One o f the primary econom ic functions o f an organized exchange is to bring traders together in the same place, to obviate such search costs. There is a recursive catch in the econom ic logic here, however: traders may not be attracted to a market if it is not liquid, but a market may not be liquid unless traders are attracted to it. Trading volumes for the June and Decem ber coinage contracts, however, w ere never large. On most days, few er than five contracts changed hands. There w ere many days with no trades. Of 183 trading days for the June contract, there w ere 66 days when only a bid quote or only an ask was available; neither was available on 78 days. For the Decem ber contract, 92 o f 310 trading days had only one side o f the market present, and on 181 days neither side was available.22 Trading volume peaked at 45 con tracts on October 31, 1973; in comparison, for the CBOT’s silver futures, several thousand con tracts changed hands on a good day in 1973.23 By October 1974, trading in the Canadian coinage futures had dwindled to almost nothing. The last recorded trade came on October 25, 1 9 7 4 , when a single Decem ber 1 9 7 4 contract changed hands. In hindsight, w e have a theoretical explana tion for the contract’s failure. Futures contracts for Canadian silver coins should have been at 21See Black (1986), pp. 19-21. There is no universally accept ed definition for the term “ liquidity.” It is usually associated with an absence of search costs. See Demsetz (1968), Tinic (1972) and Logue (1975) for further discussion. tractive to owners o f silver coins, w ho wanted to hedge the value o f their coins against price fluctuations, and speculators, w h o w ere willing to bet they could predict those price changes m ore accurately than the rest o f the market. A comparison o f prices reveals that existing silver futures w ere a close substitute for the coinage futures as a hedging/speculating tool. Moreover, the long-established silver contract traded in a much more liquid market. Thus, hedgers and speculators had in the silver contract most of the benefits o f the coinage contract as a hedg ing instrument, without most o f the drawbacks associated with illiquidity. In this context, then, the silver futures should be seen as uniformly preferable to the coinage futures. CASE 2 , A SUCCESS: M A R K E T INDEX M UTUAL FUNDS W e turn now to a successful innovation: mar ket index mutual funds. The size o f the particu lar fund chosen as an example has grow n steadily since its introduction in 1976. Although an index fund is little m ore than a repackaging o f other securities, with little decision-making discretion left to the fund's management, such funds can succeed by reducing the costs o f transacting for individual investors. A Description o f the Security On August 31, 1976, the Vanguard Group in troduced the "500 Portfolio,” a stock market mutual fund whose specific objective is to track the Standard and Poor’s 500 stock market index (S&P 500). In the language o f the fund’s prospectus, "the 500 Portfolio seeks to replicate the aggregate price and yield perform ance of the Standard & Poor’s 500 Composite Stock Price Index. ... The 500 Portfolio invests in all 500 stocks in the S&P 500 Index in approxiin October 1973. Again, in comparison, open interest for the CBOT’s silver contract reached tens of thousands of committed contracts during this time. 22Figure 4 includes prices for the coinage contract for many days on which trading volume was zero. The exchange posted settlement prices on many days when either bids or offers appeared but no trading occurred. 23A related measure of the coinage contract’s lack of popularity is open interest (the sum of all traders’ net long positions in the contract, also called committed contracts). The same picture of a thin and illiquid market emerges. Open interest in the June 1974 contract peaked in May 1974 at 56 committed contracts. For the December 1974 contract, the peak of 37 committed contracts was reached SEPTEMBER/OCTOBER 1992 12 mately the same proportions as they are represented in the Index.”24 There are good reasons w hy w e might expect a market index mutual fund to be unattractive to investors. First, it is inflexible. For example, an investor in an index fund w ho wanted to divest any stake in, say, petroleum stocks would have to sell her entire stake in the fund. In con trast, an investor holding the same stocks directly could sell petroleum stocks without affecting her other positions. Second, a mutual fund intro duces a middle man: The fund must be com pen sated for its investment management services, which, for an index fund, are essentially papershuffling. The allocation o f assets in an index is set by a predetermined rule, which an investor could follow on his own. On the other hand, there are also reasons w hy investors might find an index fund attrac tive. First, despite the inflexibility o f rule-based indexes, some investors might be attracted to a specific index, so that the inflexibility o f that in dex would not be a binding constraint for those investors. The standard capital asset pricing model, for example, concludes that all investors should hold a value-weighted portfolio o f all available assets to achieve the best risk-return trade-off.25 Although the S&P 500 does not con tain all assets, it is well diversified, it is valueweighted, and it is widely known. Perhaps for this reason, the S&P 500 has becom e an indus try benchm ark.26 Assuming there is a special interest in the S&P 500 as an investment portfolio, w e still must explain the popularity o f a mutual fund as a preferred means o f holding that portfolio. Gorton and Pennacchi (1991) construct an elabo rate rationale for security baskets as a way for uninform ed traders to exploit the inflexibility of 24Vanguard Index Trust (1992), p. 9. For those unfamiliar with index investments, it is important to note that a buy-andhold investment strategy is incompatible with indexing. The composition of an index portfolio (i. e., the number of shares of each stock) depends on the prices of all the stocks in the index. Therefore, in contrast to a buy-andhold investment, the composition of a theoretically exact in dex portfolio changes every time the price of any one of its component securities changes. A dynamic investment strategy is required for an index portfolio. Vanguard is only one of many companies to offer an S&P 500 index mutual fund for investors. Its use as an ex ample should not be interpreted as a recommendation for or against this or any other mutual fund. There are also other ways of investing in the S&P 500 index, including in dex futures and index options. http://fraser.stlouisfed.org/ FEDERAL RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis a market index to diversify themselves against losses to insider trading. A plausible, although more prosaic, reason is that mutual funds might be a way to spread the fixed costs o f stock mar ket trades over many investors, thus lowering the average cost faced by each investor. These costs can be significant. Since stock prices generally change many times in one day, the number o f transactions required to maintain a theoretically exact index portfolio o f 500 stocks is potentially enormous. As long as there is a non-zero fixed cost per trade, the sum o f these fixed costs o f maintaining the index will be proportionately large. Redundancy W e saw in the first case that the prices of Canadian coinage futures contracts w ere tracked almost exactly by the prices o f silver fu tures; the coinage futures w ere redundant. We now perform a similar analysis on the index fund and demonstrate that it too is redundant: the value o f the fund is closely tracked by the value o f the index. In this case, however, the managers o f the fund have actively pursued ex actly this correlation as their stated purpose. Redundancy here helps explain the innovation’s success. Figures 5 and 6 are analogous to figures 3 and 4, respectively. Figure 5 plots daily price observations for the Vanguard 500 Portfolio and the S&P 500. W e see that the fund price and the index value move tightly together. Upon closer examination, the prices seem to be con fined to a collection o f line segments radiating from the origin and rotating dow nw ard as one moves further out. The line segments are peri odically bum ped downward, as maintenance fees, operating charges and dividend and capital gains distributions tend to be concentrated in 25See, for example, Fama (1976), chapters 7 and 8. 26There are other commonly used benchmark indexes, for example, the S&P 100, the Wilshire 5000 Index, or the Dow Jones Industrial Average. 13 Figure 5 Daily Values for the S&P 500 Index and the 500 Portfolio Net Asset Value (U.S. $) of the 500 Portfolio Net Asset Value (U.S. $) o f the 500 Portfolio S & P 500 Index Figure 6 Daily Values for the S&P 500 Index and the 500 Portfolio S&P 500 Index Net Asset Value of the 500 Portfolio SEPTEMBER/OCTOBER 1992 14 cumulative year-end adjustments.27 The line seg ments also tend to move successively outward over time, because o f the general upward trend in stock prices over time. Even with these dis continuities, however, a regression o f five-dayholding-period mutual fund returns on S&P 500 returns yields an R2 statistic o f 0.978; according to the standard measure, the mutual fund is a nearly perfect hedge for the S&P 500.28 Figure 6 shows the same relationship, but with the 500 Portfolio values and index values plotted as time series. Over relatively long peri ods o f time, the value o f the index grows at a slightly higher rate than the value o f the mutual fund. This long-run discrepancy reflects the fact that the index is a theoretical portfolio — one that assumes transaction costs are zero.29 The mutual fund, on the other hand, charges each shareholder a fixed quarterly account mainte nance fee, plus operating expense charges equal to .2 percent o f the value o f the fund over the course o f each year. Over time, these fees ac cumulate and com pound to produce a dis crepancy. Performance o f the Security W e turn now to the question o f the innova tion’s success. For this, w e need a measure. The innovator in this case was the fund’s manager, the Vanguard Group. W e therefore define a success as an innovation that profits them. For an index fund, unlike many other form s o f in termediation, the manager profits only through^ its fees. The fund receives periodic fees from its shareholders, while incurring the costs o f fund management: accounting, buying and selling stocks, receiving and disbursing payments, etc. Data on management costs w ere unavailable, 27Because of differences in the way the index and the mutu al fund account for dividends and capital gains, tracking the net asset value of the mutual fund tends to understate its performance relative to the index. In particular, the in dex generally assumes that dividends, stock splits, etc., are reinvested, adjusting the index accordingly, while the mutu al fund gives investors the option of collecting their divi dends and capital gains. Unfortunately, sufficient information was not available to make accurate compensat ing adjustments to the net asset values of the fund. “ Performing the same analysis with daily returns, the R2 statistic is 0.859. The drop in significance is attributable to the fact that returns for one-day holding periods are smaller than for five-day holding periods, relative to noise factors such as rounding errors. 29As noted in footnote 27, the mutual fund time series presented here tends to understate its performance relative to the index. If we were able to compensate for disburse Digitized forFEDERAL FRASER RESERVE BANK OF ST. LOUIS however, so that profits could not be measured directly. Instead, we assume that the fund’s profits are increasing with revenues over the entire range considered here. Under this assumption, profits increase with the size o f the fund. W e there fore use the size o f the fund as our measure of success.30 Figure 7 shows the size o f the fund both nominally and in constant (CPI-adjusted) 1983 dollars. Almost since its inception in 1976, the value o f the fund has grow n exponentially. In nominal terms, the fund currently contains $4,346 billion, or $3,176 billion in constant 1983 dollars. One plausible explanation o f the success of an index mutual fund is transaction costs. Like many expenses, the cost to an individual inves tor o f a stock purchase can be divided into a variable and a fixed portion: price per share times number o f shares plus a brokerage com mission. If large numbers o f investors wish to hold the index, and if the size o f a brokerage commission is fixed, or at least insensitive to the quantity transacted, then the investors can pool their transactions to reduce the total amount of commissions paid. A mutual fund is one way to achieve this pooling o f investments. Although the mutual fund falls short o f the index in the long run, the relevant comparison is not with a theoretical entity, but with those alternatives that are available on a practical ba sis. While everyone acknowledges the presence o f transaction costs, some might argue that these costs are too small in modern financial markets to make a difference. Gorton and Pennacchi, for example, suggest that “investors can costlessly replicate [these composite securities].”31 ments, the compensated mutual fund price path would lie above the net asset values shown here, but still below the curve for the S&P 500. The average annual return calculat ed for the mutual fund on a compensated basis was 13.9 percent over the period 8/31/76 to 12/31/91, compared with 14.4 percent for the index. See Vanguard Index Trust (1992), p. 16. 30A full discussion of the relationship between the size of the fund and Vanguard’s profits, and therefore of Vanguard’s incentives to innovate, would require consideration of is sues of returns to scale and market structure that are be yond the scope of this paper. 31Gorton and Pennacchi (1991), p. i. They go on to state (p. 2) that “ the popularity of such composite securities seems puzzling since consumers, on their own, can apparently accomplish the same resulting cash flow by holding a diversified portfolio of the same securities in the same proportions.” 15 Figure 7 Size of the 500 Portfolio Billions of dollars Billions of dollars To demonstrate the potential role o f low er transaction costs in the success o f the mutual fund, consider the perform ance o f an index portfolio managed directly by an individual in vestor trading through a brokerage house and facing realistic brokerage commissions. To make this example plausible, w e consider pre-tax returns on an S&.P 500 portfolio for an investor making a modest number o f trades each day and able to transact at standard brokerage com mission rates. In particular, w e consider an ini tial net investment o f $1 million in the S&P 500 index made on September 1, 1976. Assume that our investor is able to track the index (approxi mately) by making 10 trades per day at an aver age cost o f $30 per trade.32 Brokerage commis sions on this portfolio thus absorb $300 per day. o f the fund. The results are straightforward: over the course o f almost 15 years, the value of the index held directly has fallen to about $860,000 in nominal terms, while the mutual fund has posted a 119.2 percent gain. If w e w ere to adjust for inflation and dividends, the poor perform ance o f the direct portfolio would be even m ore striking. The result is that inves tors w ho would otherwise have to trade at stan dard brokerage rates can hold the index more cheaply as a mutual fund. Indeed, given such transaction costs, it appears likely that an index portfolio managed by an individual investor would have negative e?c ante returns. If this w ere the cheapest means o f doing so, no one would want to hold an indexed portfolio. Figure 8 compares the results for this portfo lio with those for the mutual fund over the life The foregoing has presented tw o case studies of financial innovations. One innovation, Canadi- 32Brokerage commission schedules are complex and vary considerably from one broker to another. Some of the fac tors that can affect the commission on a trade are volume discounts, odd-lot trading premia, and negotiated commis sions for very active customers. The average $30 fee used here is based on a commission rate of 1 percent of the to tal price of each transaction, an average share price of $30 per share, and trading in single round lots of 100 shares. CONCLUSIONS SEPTEMBER/OCTOBER 1992 16 Figure 8 An Index Portfolio Subject to Transaction Costs Direct Portfolio Value (Millions of dollars) an silver coin futures, failed. The other, a mar ket index mutual fund, has succeeded. Together, the tw o case studies represent an experiment of sorts. W e have examined tw o innovations with a com m on feature: Both w ere redundant in the technical sense that their price movements were closely tracked by the price movements o f other securities. One might say that the first failed b e cause it was redundant, while the latter has succeeded for the same reason. This conclusion is m ore compelling if w e state it as the following m ore general proposition: Given tw o securities with redundant prices (that is, tw o that are perfect hedges for one another), investors will be drawn to the one with the low er transaction and liquidity costs. If there is no investor clientele for which a redundant security is the cheaper to employ, then that security will fail. W hen stated in these terms, the conclusion seems obvious. Nonetheless, this proposition is not universally observed. It requires an explicit acknowledgement o f the fact that transaction and liquidity costs can be significant factors in the financial marketplace. This runs contrary to http://fraser.stlouisfed.org/ FEDERAL RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis Net Asset Value of the 500 Portfolio the com m on assumption in financial economics that capital markets are “perfect,” which im plies, among other things, that transaction costs are zero. While such an assumption may be ap propriate in certain applications, a full under standing o f the behavior o f financial markets and innovations requires an appreciation o f these various hindrances to exchange. Thus, to the extent that financial frictions such as trans action and liquidity costs represent real resource drains on an econom y, successful financial inno vations should normally be regarded as welfareenhancing. By replacing cum bersom e or ineffi cient modes o f exchange, successful innovations can make everyone involved better off. REFERENCES Alchian, Armen A. Economic Forces at Work (Liberty Press, 1977). Allen, Franklin, and Douglas Gale. “ Optimal Security De sign,” Review of Financial Studies 1 (Fall 1989), pp. 229-63. American Bankers Association. “ Facts About the New U. S. Coins,” (ABA, 1965). 17 Black, Deborah G. “ Success and Failure of Futures Con tracts: Theory and Empirical Evidence,” Monograph Series in Finance and Economics, Salomon Brothers Center for the Study of Financial Institutions (New York, 1986). Board of Governors of the Federal Reserve. “ Interpretation of Regulation D: Currency or Coin Held Principally for its Numismatic or Bullion Value,” Federal Reserve Bulletin (De cember 1970), p. 942. Johnston, Elizabeth Tashjian, and John J. McConnell “ Re quiem for a Market: An Analysis of the Rise and Fall of a Financial Futures Contract,” Review of Financial Studies 2 (1989), pp. 1-23. Kane, Edward J. “ Good Intentions and Unintended Evil: The Case against Selective Credit Allocation,” Journal of Money, Credit and Banking (February 1977), pp. 55-69. . “ Microeconomic and Macroeconomic Origins of Financial Innovation,” in Financial Innovations: Their Impact on Monetary Policy and Financial Markets (Kluwer-Nijhoff, 1984), pp. 3-20. Chen, Andrew H., and John W. Kensinger. “ Creating Contin gent Liabilities: Master Craftsmanship in Financial En gineering,” in Game Plans for the ‘90s: Proceedings of the 26th Annual Conference on Bank Structure and Competi tion, May 9-11, 1990 (Federal Reserve Bank of Chicago, 1990), pp. 199-223. Logue, Dennis E. “ Market-Making and the Assessment of Market Efficiency,” Journal of Finance (March 1975), pp. 115-23. Demsetz, Harold. “ The Cost of Transacting,” Quarterly Journal of Economics (February 1968), pp. 33-53. Marton, Andrew. “ How much is too much?” Institutional In vestor (August 1984), pp. 238-50. Ederington, Louis H. “ The Hedging Performance of the New Futures Markets,” Journal of Finance (March 1979), pp. 157-70. Fama, Eugene F. Foundations of Finance (Basic Books, 1976). Flood, Mark D. “An Introduction to Complete Markets,” this Review (March/April 1991), pp. 32-57. Friedman, Milton. “ Financial Futures Markets and Tabular Standards,” Journal of Political Economy (February 1984), pp. 165-67. Goldsmith, Raymond W. Premodern Financial Systems, A Historical Comparative Study (Cambridge University Press, 1987). Goodhart, Charles. “ Financial Innovation and Monetary Con trol,” Oxford Review of Economic Policy (Winter 1986), pp. 79-102. Gorton, Gary, and George Pennacchi. “ Security Baskets and Index-Linked Securities,” NBER working paper No. 3711 (May 1991). International Monetary Market of the Chicago Mercantile Ex change, Inc. A Guide to Silver Coin Futures Trading (IMM, 1973a). Miller, Merton H. “ Financial Innovation: The Last Twenty Years and the Next,” Journal of Financial and Quantitative Analysis (December 1986), pp. 459-71. _______ Financial Innovations and Market Volatility (Basil Blackwell, 1991). Podolski, T. M. Financial Innovation and the Money Supply (Oxford: Basil Blackwell, 1986). Rasche, Robert H. “ Monetary Policy and Financial Deregula tion in the United States,” Kredit und Kapital (Heft 3, 1988), pp. 451-68. Schumpeter, Joseph A. Business Cycles: A Theoretical, Histor ical, and Statistical Analysis of the Capitalist Process, Volume I (McGraw-Hill, 1939). Silber, William L. “ Innovation, Competition, and New Contract Design in Futures Markets,” Journal of Futures Markets (Summer 1981), pp. 123-55. _______ “ The Process of Financial Innovation,” American Economic Review (May 1983), pp. 89-95. _______ “ Towards a Theory of Financial Innovation,” in Wil liam L. Silber, ed., Financial Innovation (Lexington Books, 1975), pp. 53-85. Simpson, Thomas D. “ Changes in the Financial System: Im plications for Monetary Policy,” Brookings Papers on Eco nomic Activity (1:1984), pp. 249-65. _______ “ Chapter 15: Canadian Silver Coins,” Rules of the International Monetary Market of the Chicago Mercantile Ex change (CME, 1973b). Tinic, Seha M. “ The Economics of Liquidity Services,” Quart erly Journal of Economics (February 1972), pp. 79-93. _______ International Monetary Market Year Book 1973-1974 (IMM, 1974). Vanguard Index Trust. Prospectus: March 15, 1992; Revised April 10, 1992 (Valley Forge: The Vanguard Group, 1992). SEPTEMBER/OCTOBER 1992 18 Kevin L. Kliesen and John A. Tuttun Kevin L. Kliesen is an economist and John A. Tatom is an assistant vice president at the Federal Reserve Bank of St. Louis. James P. Kelley provided research assistance. This article was written while Tatom was a visiting economist at the Austrian National Bank. Tatom received useful comments in seminars on this paper at the Bank of the Netherlands, the Swiss National Bank and the Institute of Advanced Studies in Vienna. The Recent Credit Crunch: The Neglected Dimensions C o n v e n t io n a l w i s d o m h a s i t t h a t a credit crunch occurred in the U.S. econom y in 1990-92, causing, or at least contributing to, the latest recession and jeopardizing the strength of the recovery. M uch has been written about the causes and consequences o f the credit crunch and about remedies for it.1 behavior o f interest rates, interest rate spreads and commercial bank business loans. This paper suggests that recent movements in short-term interest rates and changes in relevant interest rate spreads cast doubt upon the conventional credit crunch view. While the term is widely used, the precise definition o f a credit crunch is not widely agreed upon. Credit crunches have in comm on, however, a slowing in the growth o f—or an outright decline in—the quantity o f credit outstanding, especially business loans at commercial banks. Analysts w ho espouse the credit crunch theory typically have a more specific definition in mind. In their view, a credit crunch arises from a reduction in the supply o f credit. Accordingly, this article uses the term "credit crunch” to refer only to a reduction in the supply o f credit. It addresses the credit crunch hypothesis by examining the existence and implications o f competing potential sources of a decline in credit, including the recent W H A T IS A CREDIT CRUNCH? ’ The Chairman of the Federal Reserve System, Alan Greenspan, has expressed concern over the slowing in credit growth and the extent to which it was induced by bank regulators’ attempts to raise bank capital. See Greenspan (1991). Other Federal Reserve officials who have expressed concern over the credit crunch during this period include LaWare (1991), Forrestal (1991) and Syron (1991). Concern over a potential global credit crunch has been raised by the Bank for International Settlements (1991) and Japan’s Economic Planning Agency [see Reuters (1991a)]. Concern for a national crunch has also surfaced in France [see Reuters (1991b)]. Also see O’Brien and Browne (1992). FEDERAL RESERVE BANK OF ST. LOUIS The traditional notion o f a credit crunch originally involved the process known as disintermediation— a decline in savings-type deposits at banks and savings and loans that result in a decline in bank lending.2 Episodes o f disintermediation occurred w hen market interest rates, especially rates on Treasury bills and commercial paper, rose above Regulation Q interest rate ceilings at these financial institutions. As this occurred, depositors withdrew their funds from banks and savings and loans to invest at higher open market rates, and bank credit, especially for business loans, fell. 2See Kaufman (1991) for a discussion of the origin of this term in his work with Sidney Homer and for a brief sketch of the history of credit crunches. For detailed analysis of previous credit crunches, see Wojnilower (1980). 19 The phrase "credit crunch” was coined in mid-1966 when the Federal Reserve’s monetary policy becam e more restrictive; the Fed wanted to slow the growth o f demand for goods and services in order to fight inflation.3 In 1966, the Fed’s actions to slow the growth o f money and credit were reinforced by allowing short-term interest rates to rise above the Regulation Q ceiling rate on bank deposits. As a result, depositors withdrew their funds from regulated deposits to seek higher market rates. The reduction in bank deposits, in turn, limited banks’ ability to lend and their supply o f credit. What made this event significant was the Fed's refusal to accommodate the rise in short-term market interest rates by raising the Regulation Q ceiling rates at banks and savings and loans, as it had done in the past.4 Since financial deregulation in the early 1980s ended interest rate ceilings, such regulatory-induced disintermediation can no longer occur. A more encompassing view o f a credit crunch is based on any non-price constraint on bank lending, not simply on disintermediation. In its recent application, the source o f this constraint has presumably been the response by bankers to increased regulatory oversight and their own reaction to recent deterioration o f bank asset values and profitability. Increased savings and loan failures, as well as increased capital require ments, may also have played a part.5 This broader definition o f a credit crunch has been summarized by the Council o f Economic Advisers (1992): A credit crunch occurs when the supply of credit is restricted below the range usually identified with prevailing market interest rates and the profitability of investment projects, (p. 46) 3As a result of this policy action, interest rates rose and credit became more scarce. For example, in the third quarter of 1966, the interest rate on three-month Treasury bills rose above 5 percent for the first time in more than 30 years; the 5.04 percent average rate during the quarter was up sharply from the 4.59 percent rate in the previous quarter. The growth of the money stock (M1) had already slowed from a 7.3 percent annual rate in the two quarters ending in the first quarter of 1966 to a 4.3 percent rate in the second quarter of 1966. This was followed by a decline at a 1.2 percent rate in the third quarter and a 1.2 percent rate of increase in the last quarter of 1966. See Burger (1969) and Gilbert (1986) for discussions of this episode. 4According to Burger (1969), p. 24, the Fed previously accom modated the rise in rates in July 1963, November 1964 and December 1965 by raising the Regulation Q ceiling. This behavior is also discussed by Wojnilower (1980). One flaw with the disintermediation view is that a decline in intermedi ation through banks does not reduce the total supply of A credit crunch, either through disintermediation, overzealous regulators or banks' unwillingness to lend for some other reason, is therefore usually thought o f as a supply phenomenon. The flow o f credit, however, results from the interaction o f both credit supply and credit demand. In other words, while supply consider ations could certainly result in a decline in credit flows, a reduction in the demand for credit could produce the same result. Fortunately, econom ic theory indicates a criterion for assessing which o f these is the dominant source o f a change in the quantity o f credit. CREDIT CRUNCHES: TH EO RY AND EVIDENCE Figures la and lb illustrate the typical analysis o f the credit market aspects o f a credit crunch using the supply and demand framework for credit flows. In this framework, the quantity o f credit demanded varies inversely with the cost o f credit— that is, the interest rate—given other factors that influence overall demand for credit. Conversely, the quantity supplied o f credit increases with the interest rate, given the other factors that influence credit supply decisions.6 We will examine two scenarios. The first illustrates the credit crunch hypothesis, namely a reduction in the quantity o f credit supplied resulting from reduced bank willingness to lend. The second scenario offers an alternative. In this case, a reduction in the demand for bank credit occurs. As will be explained below, this could be the result o f a decline in business demand for credit associated with a reduction in inventory investment.7 C ase I: S u p p ly -in d u c e d d e c lin e in c r e d it . In figure la, an equilibrium in the credit market credit unless some of the funds removed from banks are not tunneled into other credit instruments like Treasury bills or commercial paper. 5The latter idea is closely associated with Syron (1991), who has termed the reduction in lending as a ‘‘capital crunch.” Greenspan (1991) has also supported elements of this argument, as has the Council of Economic Advisers (1992). Also see Bernanke and Lown (1991) for a different perspective on this hypothesis. 6Figures 1a and 1b are drawn conditionally on an expected rate of inflation. That is, the standard assumption that expected inflation is not a cyclical phenomenon during business reces sions is employed. Therefore, in this framework, a change in a nominal variable is also a change in a real variable. 7Figures 1a and 1b are not meant to imply that bank credit determines the level of economic activity. For alternative explanations about the linkage between bank credit and economic activity, see the shaded insert on p. 24. SEPTEMBER/OCTOBER 1992 20 Figure 1a Decline in the Supply of Credit Case I: Reduced willingness to lend Interest Rate Figure 1b Decline in the Demand for Credit Case II: Reduced loan demand by businesses Interest Rate FEDERAL RESERVE BANK OF ST. LOUIS 21 Figure 2 Quarterly Change in Domestic Nonfinancial Debt as a Percent of GDP Percent Percent r7 7 -i i 1960 62 i i 64 i i 66 i i 68 i 70 72 74 76 78 80 82 84 86 88 90 1992 Periods of business recession are indicated by the shaded areas. exists at interest rate i0 with a flow o f credit equal to C0. If one of the other factors influencing the supply o f credit shifts—for example, if there is a reduced willingness on the part o f banks to supply credit at a given interest rate—then the initial supply schedule (S0) will shift leftward to Sr As a result, the market interest rate will rise to i, to eliminate the shortage o f credit and thus the quantity o f credit will fall to Cr Case II: Dem and-induced decline in credit. The quantity o f credit can also fall because the demand for bank credit declines. This is shown in figure lb. As before, the quantity supplied and quantity demanded for credit are initially in equilibrium at C0 and i0. As the demand for credit falls, an excess supply o f credit develops. Accordingly, the interest rate will decline to a new equilibrium level (ij)—the 8lt is possible for a reduction in the supply of credit to occur in conjunction with a decline in the interest rate, but this still requires that the demand for credit declines (shifts to the left) by more than the supply of credit. Thus, the dominant point where quantity demanded and quantity supplied are once again equated.8 Evidence: Credit Flows The central feature o f a credit crunch—a decline in the growth o f credit—has occurred in every period that has been identified as a credit crunch. Such periods also tend to be recessions. For example, Kaufman (1991) cites credit crunches that occurred in 1959, 1969-70, the mid-1970s, 1981-82 and 1990-91. Except for the first, these periods correspond to each o f the recessions that have occurred since the late 1950s. The first instance, in 1959, preceded the only other recession since then, the recession from 11/1960 to 1/1961. Figure 2 shows two measures o f the flow of credit, namely the quarter-to-quarter change in impulse accounting for the decline in the quantity of credit would remain the shift in demand. Proponents of the credit crunch view, however, emphasize the supply channel as the principle source of the reduction in credit. SEPTEMBER/OCTOBER 1992 22 Figure 3 Short- and Long-Term Interest Rates Percent Quarterly Data Percent Periods of business recession are indicated by the shaded areas. private domestic nonfinancial debt (excluding federal debt) and the change in total domestic nonfinancial debt—both expressed as a percent o f gross domestic product (GDP). Typically, in recessions, the rate o f debt accumulation declines relative to GDP.9 This reflects the fact that credit growth tends to be cyclical, especially the growth o f private credit; typically, credit growth slows relative to GDP during recessions and rises during expansions.10 Thus, it is not possible to ascertain from figure 2 whether the decline in credit in each instance caused the recession or 9 The cyclical behavior of public sector deficits and overall credit demand and its implication for interest rates is discussed in more detail in Tatom (1984 and 1985). ' “Another factor influencing the private demand for credit, particularly business credit, is business’ use of internally generated funds to finance investment. The significance of the cyclical behavior of these funds is developed by Gilbert and Ott (1985). Internal funds include retained earnings and depreciation allowances. When a firm’s investment exceeds its internally supplied funds, it must turn to external financing to bridge this gap. Conversely, when internal funds exceed investment flows, a firm does not necessarily FEDERAL RESERVE BANK OF ST. LOUIS was merely a reflection o f the recession. To determine that, one must look at interest rate movements over the business cycle. Evidence: Interest Rate Movements The two scenarios shown in figure 1 provide us with a stark contrast: In figure la, the decline in credit results in a rise in the interest rate, while in figure lb, the decline in credit is associated with a decline in the interest rate. A rise in interest rates, however, is not typical in reces sions. Figure 3 shows a long-term interest rate require external financing. For nonfinancial corporations, the ratio of fixed investment (plant and equipment) and inventory investment to internal funds is greater than one during cyclical expansions, but usually falls below one during recessions, exhibiting the same cyclical nature of credit demand. During the seven recessionary periods from 1953 to 1982, the average ratio fell from 1.24 at the business cycle peak to 1.01 at its trough. A similar pattern developed in the most recent recession. In the third quarter of 1990, this ratio stood at 1.15; by the second quarter of 1991, it had fallen to 0.89. Thus, internal funds were relatively more abundant for financing investment, so less external financing, including bank credit, was demanded. 23 Table 1 The Federal Funds Rate and the Business Cycle____________ __________ Federal Funds Rates (percent)__________ A t b u s in e s s cycle p e a k/ tro u g h A t c lo s e s t p e a k 1 A t c lo s e s t tro u g h 1 111/1957-11/1958 3.24 / 0.94 SAME SAME 11/1960-1/1961 3.70 / 2.00 3.99 (-2 ) 1.68 (+2) IV/1969-IV/1970 8.94 / 5.57 8.98 (-1) 3.86 (+1) IV/1973-1/1975 10.00 / 6.30 10.57 (-1) 5.41 (+1) 1/1980-111/1980 15.07 / 9.83 SAME 111/1981-IV/1982 17.59 / 9.28 17.79 (-1) 8.66 (+1) 111/1990-11/1991 8.16 / 5.86 9.73 (-5 ) 3.77 (+4)2 B u sin e ss cycle p e a k /tro u g h SAME 'The number o< quarters earlier ( - ) or later (+) in which the closest federal funds rate peak or trough occurs is indicated in parentheses. 2Latest data available. (the yield on long-term government bonds) and a short-term interest rate (three-month Treasury bills) since 1947. Typically, both long- and short term interest rates decline during recessions, although the long-term rate is much less cyclical than the short-term rate. Indeed, short-term interest rates sometimes reach a peak before the business cycle peak. In the latest instance, in particular, the three-month Treasury bill rate peaked at 8.54 percent in the first quarter o f 1989, six quarters before the business cycle peak. since credit crunches are allegedly the result of Federal Reserve policy actions to reduce credit availability, the interest rate in the federal funds market may be a more useful indicator. Table 1 shows the average federal funds rate at the business cycle peak and trough and the nearest peak and trough o f the rate itself. The fed funds rate typically peaks before or at the business cycle peak, suggesting that the type of supply shift shown in figure la does not occur during recessions. Like the three-month Treas ury bill rate, the fed funds rate typically falls before and during recessions. An alternative measure of short-term interest rate movements is the interest rate in the federal funds market. Briefly, the fed funds market is the market in which banks compete daily for excess reserves to meet their level of required reserves. Since the fed funds market is heavily influenced by Federal Reserve policy actions (through open market operations), some analysts believe that the behavior o f this key short-term rate can provide direction about the relative tightness o f bank credit.11 Accordingly, There are periods, o f course, when short-term interest rates rise and the flow o f credit declines. Figures 2 and 3 support the earlier discussion which indicated that such developments occurred in 1966 and, to a certain extent, over stretches during the period 1986-88. Interest rates and the demand for credit, however, tend to be pro cyclical, both when the econom y is in recession "Since the Fed and banks create money largely by acquiring debt instruments (loans and investments in debt issued by governments, firms and individuals), analysts typically use the terms money and credit interchangeably. Some analysts, like Kaufman, regard the principal channel of influence of monetary policy to be its effect on credit, not on the money stock or other measures of monetary assets. Kahn (1991) suggests that the linkage between bank loans and monetary aggregates is weak, so that a slowing in monetary growth could not have caused bank loan growth to fall in 1990 nor could an increase in money growth have raised bank lending. Walsh (1991) argues that the recent credit crunch reflects a decline in the demand for credit, not in its supply, and that traditional monetary policy actions can change the supply of credit independently of bankers’ willingness to make loans. Some analysts look at the excess reserve holdings of commercial banks as an indicator of a credit crunch, but Haubrich (1991) finds this measure an unreliable indicator of credit conditions. SEPTEMBER/OCTOBER 1992 24 Bank C redit and E co n o m ic A ctivity Many analysts attribute a central, causal role in business cycle developments to bank credit movements, regardless o f the reasons for such movements. For example, some re searchers suggest that it is the interplay of credit and real econom ic activity that pro vides banks and monetary institutions with a potentially direct role in business cycle develop ments.1 In particular, they view the growth o f the m oney stock as purely passive, responding to the general movements o f the economy, while disruptions to credit markets have real consequences for output and employment decisions. Thus, in this view, a decline in credit supply is critically important in causing and maintaining recessionary conditions.2 A related view assigns a special role to banks in extending credit and promoting econom ic activity. This view emphasizes that many firms are relatively small and have limited access to organized financial markets (for example, the commercial paper market). Accordingly, these firms’ ability to expand is constrained by their access to bank credit.3 The special role o f banks, then, is to provide objective evaluations and monitoring services for the continuing viability and credit worthiness o f this relatively large sector o f the economy. Thus, one dollar o f bank credit is not a perfect substitute for one dollar o f other credit, like a direct personal loan or the proceeds from selling commercial paper. ’ This is known as the “ credit view” of the transmission mechanism. Gertler (1988) reviews the literature on the effects of credit market shocks. See also Gertler and Hubbard (1988) and Bernanke (1986). Bernanke (1983) and Hamilton (1987) argue for an independent role of credit in explaining the Great Depression. The credit view is an extension of the approach taken by real business cycle (RBC) researchers. See, for example, Plosser (1991). 2Analyses of the unique role of bank credit sometimes focus on changes arising from changes in reserve requirements. Since reserve requirement changes affect bank credit, given deposits, it is easy to infer that bank credit can change independently of movements in money. Such analyses ignore the role of the Federal Reserve. A change in bank credit because of a change in reserve require ments, however, results in an equal and offsetting change in credit supplied by Federal Reserve Banks, as the Fed accommodates the change in demand for required reserves available by buying or selling securities or other assets. Accordingly, there is no change in the net total of credit associated with a given money stock and adjusted FEDERAL RESERVE BANK OF ST. LOUIS Bank credit carries with it the banker’s certification o f credit-worthiness and the banker’s implicit contract for future moni toring services, the provision o f financial advising and other financial services. Furthermore, extensions o f bank credit to firms provide information to potential customers, financiers and other suppliers about the firm’s econom ic prospects. In this view, not only does the supply o f cred it play a unique role, independent of monetary policy developments, but bank credit is also the principal linchpin for the influence o f financial market developments on real econom ic activity Whether a slowing in the growth o f business loans is considered a source o f recessionary pressures on econom ic activity or simply a reflection o f the recession is important to both business cycle analysts and policymakers. If the supply o f credit plays a central role and is not simply a reflection o f monetary aggregate movements, then policymakers may need tools to operate specifically on the credit supply. If credit movements do not change independently or exert an indepen dent influence, however, traditional monetary policies—namely, open market operations— can reliably address cyclical problems; credit market conditions will provide no more than useful supporting information. monetary base. Thus, movements in credit or money during such episodes are not especially unique compared with those in periods when reserve requirements do not change. 3A related argument stresses segmented markets, so that relatively small firms have no access to organized capital markets. Blinder and Stiglitz (1983) discuss the independent function of bank loans. James (1987) provides evidence supporting the view that bank loan extensions raise the value of firms. Judd and Scadding (1981) were the first analysts to model an independent role for bank loans to cause changes in the money stock. See Anderson and Rasche (1982), however, for a critical discussion of their model. More recently, analysts have focused on the structure of loan contracts and on loan commitments as a mechanism for avoiding credit rationing. See Duca and VanHoose (1990) for a model of the effects of loan commitments on optimal monetary policy. 25 / and when it is not.12 In the latest instance, in fact, interest rates and credit flows were declining well before the econom y entered the recession. Thus, the evidence presented in this section is consistent with the hypothesis that, during episodes that have been characterized as credit crunches, the factors that affect the demand for credit tend to outweigh any possible effects from factors influencing the supply o f credit.13 The appropriate interpretation o f credit market developments during the latest "credit crunch” is that illustrated in figure lb. should rise relative to the interest rate at which they borrow (deposits). This is a poor test, how ever, because periods when credit crunches are believed to occur also tend to be periods of recessions, w hen the bankruptcy rate and default rates on business loans rise. Thus, the spread between a bank’s lending and borrowing rates should rise to compensate for these risks. Therefore, while interest rate spreads cannot provide definitive support for the existence o f a credit crunch, they can provide some evidence about the comparability o f the most recent episode by comparing the recent spreads with earlier ones. CO R R O B O R ATIN G EVIDENCE: INTEREST RATE SPREADS Figure 4 shows the quarterly interest rate spreads for some relatively risky short- and long-term securities from 1947 to the present. The short term spread measures the prime rate relative to the interest rate on three-month Treasury bills, while the long-term spread is the excess o f riskier BAA bond yields over AAA bond yields.15 Thus, a rise in the spread indicates an increase in risk. Economists have long pointed to the informational content o f interest rate spreads for econom ic activity.14 This observation stems from the fact that certain credit market instruments have similar characteristics and are more or less substitutable. Thus, if the interest rate on one instrument changes relative to its substitute, this may provide information about the behavior o f credit market participants. For example, evidence o f unusual bank lending behavior might be obtained by looking at the difference between bank lending rates and the rates banks pay for deposits or between rates available at banks and those avail able elsewhere. This "pricing” behavior o f banks has a direct application in examining credit market conditions. If banks becom e more reluctant to lend, then the interest rate at which they lend (bank loans) 12As pointed out by Gilbert (1976), one must distinguish between the demand for long-term credit and the demand for short-term credit when discussing the cyclical nature of credit demand. The demand for certain types of long-term credit tends to increase toward the end of a recession. (For example, many firms undertake bond and equity financing to lengthen the maturity of their capital structure). This rise reflects a shift away from short-term credit and not a rise in total credit demand. In effect, firms not only reduce credit demand, they also restructure their financing, shifting toward longer-term financing like bonds and especially new equity. 13Schreft (1990) discusses the role of credit policy in the 1980 recession. In this case, however, a rise in money demand, especially in its currency component, contributed to cyclical developments; for example, see Tatom (1981). Another recent example of a credit crunch that differs from the stereotype discussed here is the effect on world credit markets of the economic reformation of the Eastern and Central European economies. In this instance, a rise in their demand for capital ™ expected to raise the total demand and the real rate of interest in world credit markets. While an excess demand for credit will initially occur, and the interest rate is expected to rise, the total quantity of credit does not fall. See the Bank for International Settlements (1991) for a discussion of this credit crunch, although they do not refer to the situation as During recession periods, the spread on risky assets rises, especially the short-term spread, because—as figure 4 shows—that is where most o f the risk is. The rise in rates banks charge on loans that are inherently more risky in recessions— relative to their costs on insured deposits—need not reflect a new reluctance to lend beyond that arising from recession risk. Compared with the rise in spreads (on risky assets) in previous recessions, however, the recent spreads did not rise unusually. In fact, the long-term spread actually fell in 1991 before the recession ended. a credit crunch and they emphasize the effects of an earlier decline in world saving in their analysis. Also see Sesit (1991) for some doubts concerning this analysis. The Council of Economic Advisers (1991) argues that real interest rates were boosted by European developments in late 1990. Such a shift in the overall demand for credit and rise in the real interest rate would be expected to create the usual credit crunch conditions in other economies as credit is diverted to Eastern Europe. This argument has not been raised, however, in the context of the 1990-91 U.S. credit crunch. 14For example, Brunner and Meltzer (1968) emphasized the role of credit and asset prices, or interest rate spreads, in the transmission of monetary policy over the business cycle. See Bernanke (1990) for a recent study of interest rate spreads. 15Bernanke (1990) points out that the spread between the BAA bond rate and the AAA bond rate is a measure of default risk. SEPTEMBER/OCTOBER 1992 26 Figure 4 Selected Interest Rate Spreads Percent Q uarterly Data Percent Long-term spread is the excess of the yield on BAA-rated bonds over AAA-rated bonds. S hort-term spread is the prim e rate less the yield on 3-m onth Treasury bills. Periods o f business recession are indicated by the shaded areas. Moreover, the short-term spread was lower than in the three previous recessions. While figure 4 shows that the price o f risk rises in recessions, it provides insufficient evidence o f bank un willingness to lend. o f the risk arising from a recession.16 This spread rose from 196 basis points in III/1990—w hen the recession started—to about 265 basis points at the end o f the recession; it rose slightly more in the fourth quarter o f 1991. Although bank willingness to lend is certainly a function o f default risk, a bank’s demand for funds to intermediate is also a factor. Figure 5 shows the spread between banks’ lending and borrow ing rates, where the prime rate is the lending rate and the interest rate paid on large negotiable certificates o f deposit (CD) is used for the bank borrowing rate. A rise in this spread could reflect an increased reluctance to lend like that envisioned by credit crunch analysts, but this reluctance may simply reflect an assessment The rise in the spread and its recent levels, however, are both smaller than the peak spreads observed in the 1980 and 1981-82 recessions.17 While the spread in 1990-91 was higher than in the 1969-70 and 1973-75 recessions, the rise in the spread during each o f the four previous recessions was larger than the rise during the most recent recession. Therefore, if an increase in the spread between the prime rate and the CD rate is associated with an increased unwill- 16There is a distinct negative relationship between the prime rate and CD rate spread in figure 5 and the growth rate of C&l loans over the (available) sample period 1/1970 to 11/1992. The correlation between these two measures is -0.43, which is statistically significant at the 1 percent level. 17The difference between movements in inflationary expectations in the most recent recession and in the previous recessions may explain part of the difference. Changes in inflationary expectations, however, should affect both interest rates in the same direction. FEDERAL RESERVE BANK OF ST. LOUIS 27 Figure 5 Spread Between the Prime Rate and the Rate on 3-Month Certificates of Deposit Percent Quarterly Data Percent Periods of business recession are indicated by the shaded areas. ingness to lend, then in the most recent case such unwillingness was smaller than usual.18 Another related measure that could indicate a rise in bankers’ reluctance to lend is the spread on banks' borrow ing rates com pared to rates available to lenders elsewhere. Figure 6 shows how bank borrow ing rates change relative to other safe, short-term rates during recessions. Because the intermediation process occurs through the initial issuance o f bank liabilities (that is, a bank attracts deposits that it will then relend), a bank’s demand for funds to intermediate should show up in the CD rate. >8lt is possible that the spread for some borrowers, or in some sectors or regions, rose unusually in this recession, perhaps reflecting some risk beyond those typically associated with recessions. 19See Gilbert (1976) for a discussion of the earlier episode and the typical weakness of loan demand early in a recovery. 20Figure 6 also shows that the rate on CDs generally increases at some point early in a recession. This rise probably occurs Figure 6 shows that the spread between the three-month CD rate and the three-month Treasury bill, although somewhat volatile, always declines from peak to trough. Furthermore, this spread typically falls sharply late in recessions and for a while thereafter. For example, this spread fell through most o f 1970, at the beginning and end o f the 1973-75 recession, at the trough in 1980 and during much o f the 1981-82 recession.19 Generally, however, this spread appears to be relatively higher at some point in each recession than it was at the business cycle peak.20 Following a brief surge in the fourth quarter o f 1990, the CD rate declined at the end o f the recession and because of the perceived riskiness of CDs due to bank failure; however, since the “ too-big-to-fail” doctrine, and especially since the savings and loan bailout, CDs are about as safe as Treasury bills. Thus, this spread— as shown in figure 6— had moved to relatively low levels even before the latest recession. The decline in the spread immediately before the recession is consistent with an earlier decline in bank loan growth. SEPTEMBER/OCTOBER 1992 28 Figure 6 Spread Between the Rate on 3-Month Certificates of Deposit and the Rate on 3-Month Treasury Bills Percent Quarterly Data Percent “ I— I— I— T T — I— I— I— I— T T - T T — I— I— I— I— 1— I— I— T— 1970 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 1992 Periods o f business recession are indicated by the shaded areas. early in the recovery—just as it did previously. By the third quarter o f 1991, the spread between the CD rate and the Treasury bill rate had fallen to its lowest level since 1976-77. This narrowing o f the spread (that is, a relative decline in the CD rate) and its relatively low level are consistent with reductions in banks' demand for funds—either because loan demand was weaker or because bankers were reluctant to lend. In either case, however, the recent decline in this spread is much smaller than the peak-to-trough decline in the four previous recessions. The evidence on bank pricing indicates that the spread between the lending (prime) and borrowing (CD) rate widened during the recent recession, as it typically does, although not by as much. It also indicates that the prime rate rose and the CD rate fell relative to the Treasury bill rate. The decline in the latter was not unusual, nor was it unusually large. Since it is not unusual for bank margins to rise in recessions or for the CD rate to fall at FEDERAL RESERVE BANK OF ST. LOUIS the end o f recessions, these arguments, while consistent with the credit crunch hypothesis, do not support the view that bankers have been less willing to make business loans. The critical issue in reaching this conclusion is whether interest rate spreads reacted unusually in the most recent recession. As shown in figures 4 through 6, they did not. So far we have focused on the specifics o f a credit crunch—as hypothesized by a simple model for the demand and supply o f credit— and on the evidence from interest rate spreads. This evidence questions the supply-side argument behind the credit crunch hypothesis and shows the recent episode was not unusual relative to earlier instances when, others have argued, cred it crunches occurred. This leaves open the issue of the source of movement in the demand for business credit at banks during these periods. The nexus between the demand for business loans and business inventories is explored below. 29 COMMERCIAL AND INDUSTRIAL LOANS: CAUSE FOR CONCERN? W H Y DOES THE FLOW OF CREDIT DECLINE IN A RECESSION? The U.S. econom y officially entered a recession in the third quarter o f 1990, when real GDP contracted at a 1.6 percent rate. Output continued to contract in the fourth quarter o f 1990 and the first quarter o f 1991, declining at a 2.9 percent rate over the threequarter period. One o f the factors blamed for the recession was the unusual weakness o f busi ness loans.21 Even after signs o f recovery began to emerge in the spring and summer o f 1991, commercial and industrial loans at banks (hereafter, business loans) remained weak. For example, nominal commercial bank loans to business fell from $642.5 billion in the fourth quarter o f 1990 to $620.7 billion in the fourth quarter o f 1991, after rising at only a 0.1 percent rate over the previous year. By the first quarter o f 1992, business loans fell to $612.8 billion; they dropped further in the second quarter to $602.8 billion. As a result, concern about wheth er the recession had ended or whether the econom y would have a double-dip, with real GDP resuming its earlier decline, continued well into the winter o f 1991-92. W hen sales slow, firms have an incentive to reduce production and employment to avoid an accumulation o f undesired inventory. Such a reduction in output and employment constitutes a typical recession. But firms also alter their other investment decisions during recessions. For instance, firms also reduce their demand for new plant and equipment based on lower desired output levels and growing excess capacity. As a result, overall investment and its financing tend to decline during recessions. There are at least two reasons why analysts are concerned about business loan growth. The first is the concern raised by proponents o f the credit crunch view o f the recent recession: slow growth of bank lending could reflect unusual structural problems in banking.22 Second, slow growth in bus iness loans is an indication that business activity is not expanding. If businesses are reluctant to expand, the potential for econom ic recovery is jeopardized. What is absent from the discussion, however, is the fact that business loans typically grow more slowly in recessions. This is examined in greater detail in the next section. 21See the references in footnote 1. For differing analyses, see Brenner and Schmidt (1991), Corcoran (1992), Furlong (1991) and Bacon and Wessel (1991), Heinemann (1991), Jordan (1992), Meltzer (1991), Passell (1991) and Prowse (1991). Syron (1991) attributes the weakness in bank loans and economic activity to a shortage of capital induced by higher capital requirements and bank losses. Parry (1992) argues that policy-related changes in the real cost of intermediation have been appropriate, even if they have permanently changed the extent of bank intermediation. 22The Council of Economic Advisers (1991) discusses several reasons for the decline in bank credit growth. It notes, how ever, that the substitution of other debt should have offset the decline in bank lending. Strongin (1991) also stresses that credit reductions have been offset by increased equity financing. Bernanke and Lown (1991) point to an absence of business loan growth. They indicate that other sources of business loans show a decline that is unusually large for The Decline in Inventory Investment in Recessions The role o f inventory investment in recessions is especially important. Indeed, one principal type o f recession is called an inventory recession because o f the central role o f changes in inventories.23 In an inventory recession, an unanticipated decline in sales growth leads to an undesired build-up o f inventory followed by adjustments to produc tion and employment. As firms reduce inventory to eliminate the initial excess, inventory investment becom es negative; however, such investment must eventually be restored to continue meeting the slower pace o f expected sales. This eventual rise in inventory investment implies that some firms’ production and sales have risen, thereby setting in motion an overall cyclical expansion. Thus, an inventory recession is characterized by, first, an initial rise in inventories relative to sales (before, or in, the initial stage o f a recession), second, a subsequent decline in inventory invest ment to a negative pace, and finally, a rebound in inventory investment before or at the recession’s end. them during recessions, reinforcing the view that the overall demand for business loans fell, not the supply. They also provide evidence that New England banks’ efforts to raise capital had a relatively small impact on bank lending, although not necessarily on business loans. Feldstein (1992) argues that the imposition of risk-based capital standards has re stricted banks’ ability to intermediate. 23The cyclical behavior of inventory investment and inventory recessions are described in more detail in Tatom (1977). The first effort to formally model the inventory cycle is Metzler (1941). Blinder and Maccini (1991) provide a recent review of the state of research on inventories. SEPTEMBER/OCTOBER 1992 30 Figure 7 Change in Business Inventories B illio n s o f 1987 d o lla rs 1959 61 63 65 Q u arterly Data 67 69 71 73 75 77 B illio n s o f 1987 d o llars 79 81 83 85 87 89 1991 P eriods o f bu sine ss recession are in dicated by th e shaded areas. Figure 7 shows the change in inflation-adjusted business inventories since 1959. Inventory invest ment does not always rise unusually at the busi ness cycle peaks. Indeed, in half the instances shown, including the latest, inventory investment falls at the business cycle peak. Also, it is un com m on for inventory investment to register increases at the end o f a recession or in the trough quarter. It is not unusual for it to rise in the first quarter o f the recovery. Such a rise occurred in seven o f the past eight recessions, although, in four o f these cases, inventory investment remained negative in the quarter following the business cycle trough. Therefore, while all recessions do not conform to the stereotypical inventory recession pattern, there is no question that movements in inventory investment play a central role in recessions. The decline in overall investment and real GDP FEDERAL RESERVE BANK OF ST. LOUIS in recessions is concentrated most heavily in their inventory investment component. Table 2 shows that the decline in the constantdollar change in business inventories accounts for much o f the business cycle peak-to-trough decline in real GDP. Excluding the relatively large swings in inventory investment (as a share o f the decline in GDP) in the 1960-61 and 1969-70 recessions, the decline in inventory investment in the most recent recession 44#.2 percent) was fairly typical. It exceeded that in three o f the previous eight recessions, although inventory investment already had declined rather substantially from early in 1989 to the business cycle peak in III/1990. For the recent period o f decline in real GDP (11/1990 to 1/1991), the decline in inventory investment o f 54.6 percent o f the production decline exceeded that in four o f the previous eight recessions. 31 Table 2 The Decline in Inventory Investment in Recessions Recession peak-trough Change in real inventory investment1 Change in real GDP' Column 1 as a percent of column 2 IV/1948-IV/1949 $-28.3 billion $-20.9 billion 111/1953-11/1954 -11.1 (-18.4) -37.4 (-43.3) 29.7 (42.5) 111/1957-11/1958 -20.1 (-22.5) -44.9 (-53.1) 44.8 (42.4) 11/1960-1/1961 -15.7 (-45.5) 5.7 (-15.8) -275.4 (288.0) IV/1969-IV/1970 -22.5 (-19.8) -1.8 (-25.0) 1250.0 (79.2) 135.4% IV/1973-1/1975 -84.7 1/1980-111/1980 -44.3 (-10.7) -97.3 (-98.2) 45.5 (10.9) 111/1981-IV/1982 -80.6 (-35.0) -104.9 (-110.1) 76.8 (31.8) 111/1990-11/1991 -31.6 (-57.9) -65.5 (-106.0) 48.2 (54.6) -135.1 62.7 'Prior to 1960, data are expressed in 1982 dollars. From 1960 to the present, data are in 1987 dollars. Numbers in parentheses correspond to the data for the respective peak-to-trough periods for real GDP: II/53-II/54, III/57-I/58, I/60-IV/60, III/69-II/70, I/80-II/80, 111/81-111/82 and 11/90-1/91. The Linkage Between Business Loans & Business Inventories Inventory decisions are also central to business loan behavior during recessions.24 Since banks tend to hold short-term liabilities, w hich in large part are payable on demand, they have a strong incentive to hold relatively short-term loans. Thus, bank loans and lines o f credit to business are principally related to business financing o f short term assets, such as inventories. Inventory assets are crucial because they are superior collateral to receivables. Moreover, the value o f receivables can disappear more easily than that o f inventory in the event o f default; inventory also can be taken over and liquidated more easily. Figure 8 shows the stock o f business inventories and business loans since 1959, both measured in nominal terms.25 Business loans and business inventories move together over time. For example, both rise slowly until 1973, then accelerate sharply until late 1974. At the end o f the 1973-75 recession and during the early quarters o f the recovery, business loans declined along with business 24For a textbook treatment of the interplay between business loans and business inventories, see Goldfeld (1966). 25The business loan series begins in 1959, but is only available for the last Wednesday of each month until 1973. After 1973, the data are quarterly averages of Wednesday data. The stock of business inventory is end-of-quarter data, based on estimates of the change in business inventory. The value inventories. The growth rates o f each series appear to slow in the 1980 recession and at the end o f the 1981-82 recession. Inventory growth is unusually slow from early 1982 to the end o f 1986 compared with loan growth, however. Over this period, business loans rose from about 44 percent o f inventories to about 59 percent, reflecting the slowing in inventory growth. Since the end o f 1986, both business loans and inventories have grown at about the same rate, keeping business loans at about 60 percent of inventories. Both measures declined in the recent recession, following a slowing in growth in 1989 and 1990. For example, from the first quarter o f 1987 to the first quarter o f 1989, business loans rose at a 5.9 percent annual rate, while business inventories rose at a 7.9 percent rate. During the same interval, overall nominal final sales in the U.S. econom y grew at a 7.9 percent rate.26 Over the next six quarters, final sales growth slowed to a 5.7 percent rate, while inventory growth slowed to a 4.3 percent rate and business loans slowed to a 3.0 percent rate of business inventory typically is much larger than that of business loans. 26Final sales is the sum of gross private domestic fixed investment, personal consumption expenditures, net exports and government purchases. GDP is the sum of final sales and the change in business inventories. SEPTEMBER/OCTOBER 1992 32 Figure 8 Business Loans and Business Inventories B illions of dollars Quarterly Data B illions of dollars Periods o f business recession are indicated by the shaded areas. o f advance. Finally, during the recent recession, from the third quarter o f 1990 to the second quarter o f 1991, final sales growth slowed to a 3.1 percent rate and inventory fell $27.6 billion, or at a 3.3 percent rate. The decline in business loans over the recession totaled $11.9 billion— a 2.5 percent rate o f decline. The link between nominal business loans and nominal business inventories is more systematic than the simple upward trends in figure 8 might suggest. Quarter-to-quarter changes in business loans are statistically related to quarter-toquarter changes in the stock o f inventories in a significant and positive fashion. Figure 9 shows the change in business inventories and the change in business loans (both in current dollars); the two series are expressed as percentages o f GDP to scale the data, but this has no effect on the close visual relationship between the two series. The correlation coefficient for these changes over FEDERAL RESERVE BANK OF ST. LOUIS the period II/1959-II/1992 is 0.52, w hich is statistically significant at the 5 percent level. Stronger evidence for the relationship between business loans and business inventories can be obtained from causality analysis. In simple terms, causality analysis examines the statistical direction o f influence from one variable to another; in particular, it assesses whether there is a statis tically significant, temporal sequence between changes in one measure and another. This issue is addressed in the appendix. The results there support the hypothesis that changes in business loans are significantly influenced by changes in business inventories. W hen inventory investment falls, as it does in every recession, it is not surprising, therefore, to see an accompanied weakness in commercial bank business loans. CONCLUSION The recent decline in the growth o f business loans at commercial banks reflects normal 33 Figure 9 Quarterly Changes in Business Inventories and Business Loans (Percent of GDP) Percent Percent Periods o f business recession are indicated by the shaded areas. cyclical phenomena. While this decline has been referred to as a credit crunch, it is unlikely to have occurred because bankers were unusually reluctant to make business loans, as is sometimes suggested. Instead, as in earlier credit crunches/ recessions, the decline most likely originated on the credit demand side. No doubt there are individual cases in which supply factors have been important in reducing credit availability. Indeed, some researchers have alleged that such occurrences explain, to a small extent, business loan slowings in some parts of the country owing to changes in bank capital requirements or other regulatory changes. These analyses generally do not control for the normal cyclical phenomena addressed here, however. The decline in business loan growth in recessions is due, in large part, to the cyclical nature of business loan demand. Bank loans are typically short-term collateralized loans, so that the prime commercial asset that is financed by bank credit is inventories. The evidence presented here sug gests that business loans and business inventory holdings are very closely related statistically, so that business loans and inventory move up or down in tandem. Since businesses typically reduce their desired inventory holdings during recessions, business loans at banks tend to decline as well. An additional consideration is the recent movement in interest rates and interest rate spreads. As in earlier periods o f so-called credit crunches, the recent decline in business loans has been accompanied by reductions in interest rates, particularly short-term rates. This behavior is inconsistent with a shortage o f credit from a simple supply and demand perspective. While interest rate spreads for "risky” credit have risen recently, including the difference between bank lending and borrow ing rates, this is also a normal cyclical phenomenon. This spread, as well as that between the prime rate and the three-month Treasury bill rate, have not been unusually large SEPTEMBER/OCTOBER 1992 34 com pared with previous recessions, nor has their increase been unusually large. Although the spread between the large CD rate and three-month Treasury bill rate has fallen, this too is not un usual in a recession nor in the initial stages o f a recovery. To summarize, the theory and evidence pre sented here suggests that recessions cause in ventory demand and the grow th o f business borrow ing to slow. To the extent that this argu ment and evidence characterize recent develop ments, the solution to the recent decline in credit grow th is likely to be found, as usual, in a restoration o f business inventory accumul ation and its financing. REFERENCES Anderson, Richard G., and Robert H. Rasche. “ What Do Money Market Models Tell Us about How to Implement Monetary Policy?” Journal of Money, Credit and Banking (November 1982), pp. 796-828. Duca, John V., and David D. VanHoose. “ Loan Commitments and Optimal Monetary Policy,” Journal of Money, Credit and Banking (May 1990), pp. 178-94. Feldstein, Martin. “ Revise Bank Capital Standards Now,” Wall Street Journal, March 6, 1992. Forrestal, Robert P. “ Policy Implications of a Credit Crunch,” speech delivered at the Conference on “ Credit Crunches— Causes and Cures,” Wellington, New Zealand, August 16, 1991. Furlong, Fred. “ Financial Constraints and Bank Credit,” Fed eral Reserve Bank of San Francisco Weekly Letter (May 24, 1991). Gertler, Mark. “ Financial Structure and Aggregate Economic Activity: An Overview," Journal o f Money, Credit and Banking (August 1988), pp. 559-88. Gertler, Mark, and R. Glenn Hubbard. “ Financial Factors in Business Fluctuations,” National Bureau of Economic Research, Working Paper No. 2758 (November 1988). Gilbert, R. Alton. “ Requiem for Regulation Q: What It Did and Why It Passed Away,” this Review (February 1986), pp. 22-37. _______ “ Bank Financing of the Recovery," this Review (July 1976), pp. 2-9. Bacon, Kenneth H., and David Wessel. “ Wary Lenders,” Wall Street Journal, September 30, 1991. Gilbert, R. Alton, and Mack Ott. “ Why the Big Rise in Business Loans at Banks Last Year?” this Review (March 1985), pp. 5-13. Bank for International Settlements. 61st Annual Report (June 10, 1991). Goldfeld, Stephen M. Commercial Bank Behavior and Economic Activity (North Holland: Amsterdam, 1966). Bernanke, Ben S. “ On the Predictive Power of Interest Rates and Interest Rate Spreads,” Federal Reserve Bank of Boston New England Economic Review (November/December 1990), pp. 51-68. Greenspan, Alan. “ Statements to Congress,” Federal Reserve Bulletin (May 1991), pp. 300-310. _______ “ Alternative Explanations of the Money-lncome Correlation,” Real Business Cycles, Real Exchange Rates and Actual Policies, Carnegie-Rochester Conference Series on Public Policy (Autumn 1986), pp. 49-100. Hamilton, James D. “ Monetary Factors in the Great Depres sion,” Journal of Monetary Economics (March 1987), pp. 145-69. Haubrich, Joseph G. “ Do Excess Reserves Reveal Credit Crunches?” Federal Reserve Bank of Cleveland Economic Commentary (July 15, 1991). _______ “ Nonmonetary Effects of the Financial Crisis in the Propagation of the Great Depression,” American Economic Review (June 1983), pp. 257-76. Heinemann, H. Eric. “ The ‘Credit Crunch’ Is a Red Herring,” Christian Science Monitor, October 1, 1991. Bernanke, Ben S., and Cara S. Lown. “ The Credit Crunch,” Brookings Papers on Economic Activity (2:1991), pp. 205-47. James, Christopher. "Some Evidence on the Uniqueness of Bank Loans,” Journal of Financial Economics (December 1987), pp. 217-35. Blinder, Alan S., and Louis J. Maccini. “ Taking Stock: A Crit ical Assessment of Recent Research on Inventories,” Jour nal of Economic Perspectives (Winter 1991), pp. 73-96. Blinder, Alan S., and Joseph E. Stiglitz. “ Money, Credit Con straints, and Economic Activity,” American Economic Re view (May 1983), pp. 297-302. Brenner, Joel Glenn, and Susan Schmidt. “ Bankers Say There’s No ‘Credit Crunch’,” Washington Post, October 12, 1991. Jordan, Jerry L. “ The Credit Crunch: A Monetarist’s Per spective,” paper presented at the 28th Annual Conference on Bank Structure and Competition, Federal Reserve Bank of Chicago, May 7, 1992. Judd, John P., and John L. Scadding. “ Liability Manage ment, Bank Loans, And Deposit ‘Market’ Disequilibrium," Federal Reserve Bank of San Francisco Economic Review (Summer 1981), pp. 21-44. Brunner, Karl, and Allan H. Meltzer. “ Liquidity Traps for Money, Bank Credit, and Interest Rates,” Journal of Politi cal Economy (January/February 1968), pp. 1-37. Kahn, George A. “ Does More Money Mean More Bank Loans?” Federal Reserve Bank of Kansas City Economic Review (July/August 1991), pp. 21-31. Burger, Albert E. “ A Historical Analysis of the Credit Crunch of 1966,” this Review (September 1969), pp. 13-30. Kaufman, Henry. “ Credit Crunches: The Deregulators Were Wrong,” Wall Street Journal, October 9, 1991. Corcoran, Patrick J. “ The Credit Slowdown of 1989-91: The Role of Demand,” paper presented at the 28th Annual Conference on Bank Structure and Competition, Federal Reserve Bank of Chicago, May 6-8, 1992. LaWare, John P. “ Setting the Global Scene: A Global Credit Crunch?” speech delivered at the Conference on “ Credit Crunches— Causes and Cures,” Wellington, New Zealand, August 15, 1991. Council of Economic Advisers. Economic Report of the Presi dent (Government Printing Office, 1992). Meltzer, Allan H. “ There Is No Credit Crunch,” Wall Street Journal, February 8, 1991. _______ Economic Report of the President (Government Printing Office, 1991). Metzler, Lloyd A. “ The Nature and Stability of Inventory Cycles,” Review of Economic Statistics (August 1941), pp. 113-29. http://fraser.stlouisfed.org/ FEDERAL RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis 35 O’Brien, Paul Francis, and Frank Browne. “ A Credit Crunch? The Recent Slowdown in Bank Lending and Its Implications for Monetary Policy,” Organization for Economic Cooperation and Development Working Paper No. 107, 1992. Parry, Robert T. ‘‘The Problem of Weak Credit Markets: A Monetary Policymaker’s View,” Federal Reserve Bank of San Francisco Weekly Letter (January 3, 1992). Strongin, Steven. “ Credit Flows and the Credit Crunch,” Federal Reserve Bank of Chicago, Chicago Fed Letter (November 1991). Syron, Richard F. “Are We Experiencing a Credit Crunch?” Federal Reserve Bank of Boston New England Economic Review (July/August 1991), pp. 3-10. Passell, Peter. ‘‘Capital Crunch: Capitol Quandary,” New York Times, November 6, 1991. Tatom, John A. “ Two Views of the Effects of Government Budget Deficits in the 1980s,” this Review (October 1985), pp. 5-16. Plosser, Charles I. “ Money And Business Cycles: A Real Business Cycle Interpretation,” in Michael T. Belongia, ed., Monetary Policy on the 75th Anniversary of the Federal Re serve System (Kluwer, 1991), pp. 245-74. . “A Perspective on the Federal Deficit Problem,” this Review (December 1984), pp. 5-17. Prowse, Michael. “ The Credit Crunch as Scapegoat,” Financial Times, November 18, 1991. Reuters News Service. “Japan Predicts Capital Crunch,” International Herald Tribune, December 4, 1991a. _______ “ France Sees No ‘Crunch’ in Lending Slowdown,” International Herald Tribune, December 12, 1991b. Schreft, Stacey L. “ Credit Controls: 1980,” Federal Reserve Bank of Richmond Economic Review (November/December 1990), pp. 25-55. Sesit, Michael R. “ Fears of a Global Credit Crunch Are Overdone, Some Analysts Say,” Wall Street Journal Europe, November 5, 1991. _______ “ Energy Prices and Short-Run Economic Perform ance,” this Review (January 1981), pp. 3-17. _______. “ Inventory Investment in the Recent Recession and Recovery,” this Review (April 1977), pp. 2-9. Walsh, Carl E. “ The Credit Crunch and The Real Bills Doctrine,” Federal Reserve Bank of San Francisco Weekly Letter (May 3, 1991). Wojnilower, Albert M. “ The Central Role of Credit Crunches in Recent Financial History,” Brookings Papers on Economic Activity (2:1980), pp. 277-339. Appendix Business Loans and Business Inventories: Some Statistical Evidence The relationship between inventory decisions and the growth o f business loans can be exam ined using a straightforward causality test, which tests the statistical significance o f the temporal sequence between measures that are hypothesized to be related. The growth rate of business loans (L = 400 AlnL) and business inventories (I = 400 Alnl) can be examined to see (1) whether one measure "causes” the other; (2) whether they each cause the other (bi-directional causality) or (3) whether they are statistically independent.1 This is done by examining the statistical significance o f past values o f one measure in explaining the other, controlling for the time series properties o f the other. Considered alone, the growth in business loans (L) and in business inventories (I) are firstand third-order autoregressive series, AR1 and AR3, respectively, w hich means that current val ues o f each are highly related to their own past value one quarter earlier and three quarters earlier, respectively, but not to earlier changes. 1The statistical analysis presented here uses variables mea sured in nominal terms because the credit crunch hypothesis concerns the effects of nominal bank credit and the latter finances, in part, nominal inventory. When the procedures are performed on the same variables measured in constant-dollar terms, the results are essentially the same and the conclusions are not altered. SEPTEMBER/OCTOBER 1992 36 To conduct the causality tests, up to eight past values o f each variable were added to the autoregressive model o f the other to see if one variable is statistically significant in explaining future values o f the other. For loan growth during the period from II/1959 to 11/1992, the only statistically significant relation ship between the two measures is: (1) L, = 2.124 + 0.236 i(1 + 0.565 Lt_, (2.77) (2.47) (7.13) R2 = 0.46 S.E. = 5.586 D.W. = 1.85 This equation indicates that inventory growth causes business loan growth, because the coefficient on the change in inventory (0.236) is significantly different from zero at a 5 percent level according to the relatively high value of the t-statistic given in parentheses. The adjusted R2 and Durbin-Watson (D.W.) statistics suggest, respectively, that there is a strong relationship and that the first lag value o f the dependent variable (Lt l) is sufficient for removing any problematical serial correlation in the errors o f the estimated equation.2 No other lagged value of i or L or combination o f lagged values is statistically significant. While the evidence presented here indicates that changes in inventories result in changes in business loans, there is nonetheless some evidence o f reverse causality—although the effect is ephemeral in nature. In particular, the best time series representation for inventory growth is an AR(3) model. The inventory equation similar to equation 1, for the period 1/1960 to 11/1992, is: 2Both business loans and inventory are integrated of order one and are cointegrated according to tests that are not reported here. Equations 1 and 2, or 1 and 3, were also estimated using the seemingly-unrelated-regression method to allow for contemporaneous correlations of the error terms. This does not affect the causality conclusions. The inclusion of the lagged residual from an estimated cointegrating vector for the levels also does not alter these results. http://fraser.stlouisfed.org/ FEDERAL RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis (2) it = 1.157 + 0.560 It_, - 0.079 i(_2 + 0.274 i,_3 (1.88) (6.32) (-0 .8 0 ) (3.31) + 0.264 Lt_, - 0.216 Lt_2 (3.76) (-3 .0 2 ) R2 = 0.55 S.E. = 4.200 D.W. = 1.92 Since the lagged terms on the change in business loans are approximately equal in magnitude and opposite in sign, it is possible that only a transi tory effect is present, so that a rise in loan growth has no effect on inventory growth after two quarters. To test this hypothesis, the coefficients on the two lags o f business loan growth were constrained to sum to zero. The resulting estimate is: (3) i, = 1.318 + 0.580 it_t (2.33) (6.98) 0.067 It_2 (-0 .6 9 ) + 0.281 It_3 + 0.241 ALt_, (3.42) (3.96) R2 = 0.55 S.E. = 4.189 D.W. = 1.94 A test o f the restriction (that is, regressions 3 and 2, respectively) yields an F, 124 = 0.42, which is not statistically significant. Thus, one cannot reject the hypothesis that there is a transitory causal link from increased loan growth to increased inventory growth. After two quarters, however, a change in business loan growth has no statistically significant effect on business inventory growth.3 These results suggest that policies designed to increase bank lending, taken alone, are unlikely to raise inventory investment, which is one o f the principal effects expected by proponents o f the credit crunch/recession linkage. Moreover, these results also reaffirm the behavior o f business loans during and immediately following a recession.4 “According to Gilbert and Ott (1985), business loans at large banks typically remain at their trough level during the first year of the recovery, before giving way to moderate growth in the second year. 37 John W. Keating John W. Keating, assistant professor, Department o f Economics, Washington University in St. Louis, was a visiting scholar at the Federal Reserve Bank of St. Louis while this paper was written. Richard I. Jako provided research assistance. Structural Approaches to Vector Autoregressions 1 ' HE VECTOR AUTOREGRESSION (VAR) model o f Sims (1980) has becom e a popular tool in empirical m acroeconom ics and finance. The VAR is a reduced-form time series model o f the econom y that is estimated by ordinary least squares.1 Initial interest in VARs arose because o f the inability o f economists to agree on the econom y’s true structure. VAR users thought that important dynamic characteristics o f the econom y could be revealed by these models without imposing structural restrictions from a particular econom ic theory. Impulse response functions and variance decompositions, the hallmark o f VAR analysis, illustrate the dynamic characteristics o f empirical models. These dynamic indicators w ere initially obtained by a mechanical technique that some believed was unrelated to econom ic theory.2 Cooley and LeRoy (1985), however, argued that this method, w hich is often described as atheoretical, actually implies a particular econom ic structure that is difficult to reconcile with eco nomic theory. This criticism led to the development o f a "structural” VAR approach by Bernanke (1986), Blanchard and Watson (1986) and Sims (1986). This technique allows the researcher to use eco nomic theory to transform the reduced-form 1A VAR can be derived for a subset of the variables from a linear structural model. Furthermore, it is a linear approxi mation to any nonlinear structural model. The accuracy of the VAR approximation will depend on the features of the nonlinear structure. VAR model into a system o f structural equations. The parameters are estimated by imposing con temporaneous structural restrictions. The crucial difference between atheoretical and structural VARs is that the latter yield impulse responses and variance decompositions that can be given structural interpretations. An alternative structural VAR method, developed by Shapiro and Watson (1988) and Blanchard and Quah (1989), utilizes long-run restrictions to identify the econom ic structure from the reduced form. Such models have long-run characteristics that are consistent with the theoretical restric tions used to identify parameters. M oreover, they often exhibit sensible short-run properties as well. For these reasons, many economists believe that structural VARs may unlock econom ic information embedded in the reduced-form time series model. This paper serves as an introduction to this developing literature. The VAR model is shown to be a reduced-form for a linear simultaneous equations model. The contemporaneous and long-run approaches to identifying structural parameters are developed. Finally, estimates o f contemporaneous and long-run structural VAR models using a com m on set o f macroeconom ic variables are presented. These models 2A Choleski decomposition of the covariance matrix for the VAR residuals. SEPTEMBER/OCTOBER 1992 38 are intended to provide a comparison between contem poraneous and long-run structural VAR modeling strategies. The implications o f the empirical results are also discussed. and attractive assumptions are that shocks have either temporary or permanent effects. If shocks have tem porary effects, zt equals et, a serially uncorrelated vector (vector white noise).5 That is, THE V A R REPRESENTATION OF A SIMULTANEOUS SYSTEM OF EQUATIONS (3) z, = The standard, linear, simultaneous equations model is a useful starting point for understanding the structural VAR approach. A simultaneous equations system models the dynamic relation ship between endogenous and exogenous vari ables. A vector representation o f this system is (1) Axt = C(L)xt_1 + Dzt, w here xt is a vector o f endogenous variables and zt is a vector o f exogenous variables. The elements o f the square matrix, A, are the struc tural parameters on the contem poraneous en dogenous variables and C(L) is a kth degree matrix polynomial in the lag operator L, that is, C(L) = C0+ CjL + C2L2+... + CkLT, w here all o f the C matrices are square. The matrix D measures the contem poraneous response o f endogenous varia bles to the exogenous variables.3 In theory, some exogenous variables are observable while others are not. Observable exogenous variables typically do not appear in VARs because Sims (1980) argued forcefully against exogeneity. Hence, the vector z is assumed to consist o f un observable variables, which are interpreted as disturbances to the structural equations, and x, and zt are vectors with length equal to the num ber o f structural equations in the model.4 A reduced-form for this system is (2) xt = A ‘c(L) x,_j + A 'D z (. Alternatively, z can be m odeled as a unit root process, that is, (4) z, - zt_, = Equation 4 implies that z equals the sum o f all past and present realizations o f e. Hence, shocks to z are permanent. The assumptions in equations 3 and 4 are not as restrictive as they might ap pear. If these shock processes w ere specified as general autoregressions, the VARs would have additional lags. The procedures to identify struc tural parameters, however, would be unaffected. Under the assumption that exogenous shocks have only temporary effects, equation 2 can be rewritten as, (5) xt = |5(L)xt_1 + et, where /?(L) = A^CIL) and et = A 'De,. The equation system in 5 is a VAR representation o f the structural model because the last term in this expression is serially uncorrelated and each variable is a function o f lagged values o f all the variables.6 The VAR coefficient matrix, f}(L), is a nonlinear function o f the contem poraneous and the dynamic structural parameters. If the shocks have permanent effects, the VAR model is obtained by applying the first difference operator (A = 1 —L) to equation 2 and inserting equation 4 into the resulting expression, to obtain (6) Axt = /?(L)Axt j + et, with /3(L) and et previously defined. A particular structural specification for the "er ror term ” z is required to obtain a VAR representation. T w o alternative, com monly used This is a com m on VAR specification because many m acroeconom ic time series appear to have a unit root.7 Because o f the low pow er o f tests 3This model can accommodate lags of z; this feature is omitted, however, to simplify the discussion. H'his model can also be written in levels form: 4lf observable exogenous variables exist, they are included as explanatory variables in the VAR. sThe individual elements in a vector white noise process, in theory, may be contemporaneously correlated. In structural VAR practice, they are typically assumed to be independent. 6The last term represents linear combinations of serially uncorrelated shocks, and these are serially uncorrelated as well. See any textbook covering the basics of time series analysis for a proof of this result. FEDERAL RESERVE BANK OF ST. LOUIS x, = [ A - ’ C(L) + I]xt _, - A - 1C (L)x,_2 + A - 1Dct. Sims, Stock and Watson (1990) show that this reduced form is consistently estimated by OLS, but hypothesis tests may have non-standard distributions because the series have unit roots. 39 for unit roots, their existence is controversial. VARs can accommodate either side o f the de bate, how ever.8 a contemporaneous structural VAR, however, generally does not impose restrictions on the reduced form. The VAR is a general dynamic specification because each variable is a function o f lagged values o f all the variables. This generality, h ow ever, com es at a cost. Because each equation has many lags o f each variable, the set o f varia bles must not be too large. Otherwise, the m od el would exhaust the available data.9 If all shocks have unit roots, equation 6 is estimated. If all shocks are stationary, equation 5 is used.10 If some shocks have temporary effects while others have permanent effects, the empirical model must account for this. An alternative approach o f Doan, Litterman and Sims (1984) estimates the VAR in levels with a Bayesian prior placed on the hypothesis that each time series has a unit root. The Bayesian VAR model permits m ore lags by imposing restrictions on the VAR coefficients, reducing the number o f estimated parameters (called hyper-parameters). The reduction in parameters contributes to the Bayesian m odel’s propensity to yield superior out-of-sample forecasts compared with unrestricted VARs with symmetric lag structure. Recently, Blanchard and Quah (1989) have es timated a VAR model w here some variables w ere assumed to be stationary while others had unit roots. Alternatively, King, Plosser, Stock and Watson (1991) use a cointegrated model, where all the variables are difference stationary but some linear combinations o f the variables are stationary. The stationary linear combinations are constructed by cointegration regressions prior to VAR estimation. They impose the cointegration constraints using the vector error-correction model o f Engle and Granger (1987). Sometimes unit-root tests com bined with theory suggest the coefficients for stationary linear combinations. Shapiro and Watson (1988), for example, present evidence that the nominal interest rate and in flation each have a unit root while the differ ence between these tw o variables is stationary. They impose the cointegration constraint by selecting this noisy proxy for the real rate of interest as a variable for the model. CO N TEM PORAN EO U S STRU C T U R A L V A R MODELS It is clear from equations 5 and 6 that, if the contemporaneous parameters in A and D w ere known, the dynamic structure represented by the parameters in C(L) could be calculated from the estimated VAR coefficients, that is, C(L) = A/3(L). Furthermore, the structural shocks, t x, could be derived from the estimated residuals, that is, £t = D _1Aet. Because the coefficients in A and D are unknown, identification o f structural parameters is achieved by imposing theoretical restrictions to reduce the number o f unknown structural parameters to be less than or equal to the number o f estimated parameters o f the variance-covariance matrix o f the VAR residuals. Specifically, the covariance matrix for the residuals, Xe, from either equation 5 or equation 6 is Unrestricted versions o f the VAR model (and the error-correction model) are estimated by o r dinary least squares (OLS) because Zellner (1962) proved that OLS estimates o f such a sys tem are consistent and efficient if each equation has precisely the same set o f explanatory variables. If the underlying structural model provides a set o f over-identifying restrictions on the reduced form , however, OLS is no longer optimal.11 The simultaneous equations system in An OLS estimate o f the VAR provides an estimate o f Xe that can be used with equation 7 to obtain estimates o f A, D and S£. The contemporaneous structural approach imposes restrictions on these three matrices. There are n2 elements in A, 8 Alternatively, the unit root could result from parameters in the dynamic structural model. 10VAR lag length is often selected by statistical criteria such as the modified likelihood ratio test of Sims (1980). 9 The lag structure of a VAR can be shown to represent vari ous sources of economic dynamics. Structural models with rational expectations predict restrictions on the VAR model. Dynamics in these models are often motivated by the costs of adjustment to desired or equilibrium positions. The lag structure of the VAR can also be motivated by dynamic processes for structural disturbances. 11A two-step structural VAR estimator will generally not be efficient if there are structural restrictions for C(L) since this implies restrictions on /}(L). For example, Sargent (1979) derives restrictions on VAR coefficients from a particular model of the term structure of interest rates under rational expectations. The full structural system is estimated by maximum likelihood. (7) I e =E[ete'| = a ' d e Ie/J d 'a ' 1 = a ’ d I d 'a '"1, where E is the unconditional expectations operator, and S£ is the covariance matrix for the shocks. SEPTEMBER/OCTOBER 1992 40 n2 elements in D, and n (n + l)/2 unique elements in Z£, but only n (n + l)/2 unique elements in Xe. The maximum num ber o f structural parameters is equal to the num ber o f unique elements in 2 . Thus, a structural model will not be identified un less at least 2n2 restrictions are imposed on A, D and XI . Often these restrictions are exclusion restrictions; o f course, that need not be the case. Typically, Z£ is specified as a diagonal matrix because the primitive structural disturbances are assumed to originate from independent sources. The remaining parameters are identified by imposing additional restrictions on A and D. The main diagonal ele ments o f A are set to unity because each structural equation is normalized on a particular endogenous variable. The main diagonal for D has this same specification since each equation has a structural shock. These normalizations provide 2n restric tions. Identification requires at least 3 n (n -1 )/2 ad ditional restrictions based on econom ic theory. Alternatively, the restrictions may be based on the contemporaneous information assumed available to particular econom ic agents following Sims (1986). Keating (1990) and West (1990) ex tend this approach by showing how rational ex pectations restrictions can be imposed in the contem poraneous structural VAR framework. Except for Bernanke (1986) and Blanchard (1989), existing models typically have not at tempted to identify the structural parameters in D. Hence, D is usually taken to be the identity matrix, leaving at least n ( n - l) /2 additional iden tifying restrictions to be imposed on A. A two-step procedure is used to estimate structural VAR models. First, the reduced-form VAR, with enough lags o f each variable to eliminate serial correlation from the residuals, is estimated with OLS. Next, a sufficient number o f restrictions is imposed on A, D and X£ to identify these parameters. This paper obtains the parameters in equation 7 with an algorithm for solving a nonlinear system o f equations. Blanchard and Quah (1989) use this approach to estimate a structural VAR m odel.12 Standard errors for the parameters, the impulse responses and the variance decompositions are calculated 12Their model is identified by long-run restrictions. 13The actual residuals are randomly sampled, and the sampled residuals are used as shocks to the estimated VAR. After the artificial series are generated, they are used to perform the same structural VAR analysis. After 200 replications of the model, standard errors were calculated for the parameter estimates, the impulse responses and the variance decom positions. FEDERAL RESERVE BANK OF ST. LOUIS using the Monte Carlo approach o f Runkle (1987), w hich simulates the VAR model to generate distributions for these results.13 The identification technique used in this paper is adequate for a model in which the number o f parameters is equal to the num ber o f unique elements in Xe. Alternative methods are needed to estimate a model with few er parameters. Bernanke (1986) uses the method-of-moments approach o f Hansen (1982) to estimate the parameters in equation 7 and obtain standard errors. Sims (1986) estimates the system o f simultaneous equations for the residuals in equation 5 using maximum likelihood.14 Blanchard and Watson (1986) also estimate the system o f equations for residuals; however, they employ a sequential instrumental variables tech nique in w hich estimated structural shocks are used as instruments in all subsequent equa tions.15 The following four-equation contem poraneous structural VAR model is used to illustrate a par ticular set o f such identifying restrictions. The residuals from a VAR consisting o f the price lev el (p), output (y), the interest rate (r) and m oney (m) are used in the model. This model is used in the empirical w ork w hich follows. Equation 8 provides three restrictions by assuming that the price level is predetermined, except that producers can respond immediately to aggre gate supply shocks. Equation 9 is a reducedform IS equation that models output as a function o f all the variables in the model. This approach was taken instead o f explicitly m odel ing expected future inflation to calculate the real interest rate and explicitly modeling the term structure o f interest rates to tie the short term rate in the model with the long-term rate predicted by theory. The IS disturbance is also a factor in the output equation. The m oney sup ply function in equation 10 allows the Fed to adjust short-term interest rates to changes in the m oney stock. Tw o restrictions are obtained from assuming that the Fed does not immediately observe aggregate measures o f output and price. The last equation is a short-run money demand function specifying nominal money holdings as ,4ln contrast to the typical simultaneous equations model, this approach has no observable exogenous variables. 15This technique requires a structural model for which there are no estimated parameters in the first equation, the second equation has one parameter, the third has two parameters, etc. While the recursive model fits this description, this technique can estimate a much broader set of models. 41 a function o f nominal GNP and the interest rate. This specification is motivated by a bu ffer stock theory w here short-run m oney holdings rise in proportion to nominal income, yielding the final restriction for a just-identified model. Each equation includes a structural disturbance. (8) eP = £js (9) e[ = A,eP + A2e; + A3etm+ eJs (10) etr = A4e^ + £tms dynamic response o f the variables to the shocks. If the variables in x are stationary, then the im pulse responses must approach zero as i b e comes large. Variance decompositions allocate each variable’s forecast error variance to the individual shocks. These statistics measure the quantitative effect that the shocks have on the variables. If Et (xt is the expected value o f xt based on all information available at time t - j , the forecast error is: i-i x. - e m x. = X > , £.-i' i«0 (11) etm= As(e[ + e(p) + AGe[ + £” d Standard VAR tools are em ployed after the structural parameters are estimated. Impulse re sponse functions and variance decomposition functions conveniently summarize the dynamic response o f the variables to the shocks, which is known as the moving average representation (MAR). The MAR for the VAR is obtained by applying simple algebra to a function o f the lag operator. Take the VAR model for x: xt = /3(L)xt_1 + et, and subtract /J(L)xt t from both sides o f this equation: since the information at time t - j includes all £ occurring at or before time t - j and the condition al expectation o f future e is zero because the shocks are serially uncorrelated. The forecast error variances for the individual series are the diagonal elements in the following matrix: i-i E(xt - Em x ,) ( x , - i-0 If 9ivs is the (v,s) element in 9( and as is the standard deviation for disturbance s (s = l,...,n), the j-steps-ahead forecast variance o f the v-th variable is easy to calculate: E <x v, - E,-jx v / = xt - /}(L)xt_, = et. Then factor terms in xt using the lag operator, [I - p(L) L]x( = et. Multiply both sides o f this equation by the in verse o f [l-/J(L)L]: Et_.xt) = ^ 9 ,^ 9 '. Yj i-0 I X s-1 °s v = !< 2 ,..., n The variance decomposition function (VDF) writes the j-steps-ahead percentage o f forecast error variance for variable v atributable to the k-th shock: E (13) VDF(v,k,j) = x, =[I - /}(L) L] *et. ------------- x 100. E E c °: i=o s-1 Insert the expression from equation 5 for et into this last equation: (12) xt = [I - /?(L)L J ' a 'Dc, = e(L)£t, CO w here 9(L) = i =0 and each 9 i is an n x n matrix o f parameters from the structural model. Equation 12 implies that the response o f xt+i to £t is 9 r Hence, the sequence o f 9 i from i = 0 ,1 ,2 ,..., illustrates the The same analysis can be used to derive the MAR for the VAR model in equation 6. (14) Axt = 9(L) et, where 9(L) = [I—/3(L) L] *A ' d . The response of x, rather than the change in x, is frequently of greater interest to economists. These impulse responses can be generated recursively by as suming that all the elements o f £ at time zero and earlier are equal to zero.16 For example, 16lf the pre-sample e is nonzero, its effects are lumped to gether with x0 which represents the initial conditions. SEPTEMBER/OCTOBER 1992 42 LONG-RUN STR U CTU RAL V A R MODELS x , = xo + V , and X2 = X! + 90£2 + 0 1£1' Inserting the expression for x, into the x2 equa tion yields: x 2 = x o + % £2 + <0 o + 0 , ) £r Repeating this operation for all x up to xt, yields the following: t-1 x, = xo + V , + < 9o + e .K - . + - + <E e>)£ij-o This result is equivalently written as t-i (15) x, = x0 +r(L)£t = x0 + E i-0 i where H = 0.. j-0 The response o f xt+ito £t is F. Since the differenced specification assumes that Ax is stationary, the 0! matrix goes to zero as j gets large. This implies that l~ converges to the sum o f coefficients in 0(L). Restrictions on this sum o f coefficients are used to identify long-run structural VAR models. The variance decompositions for this model are identical to equation 13 except that 0 is replaced by r. Shapiro and Watson (1988) and Blanchard and Quah (1989) developed the alternative approach o f imposing identifying restrictions on long-run multipliers for structural shocks. An advantage is that these models do not impose contem poraneous restrictions, but they allow the data to determine short-run dynamics based condi tionally on a particular long-run m odel.18 If each shock has a permanent effect on at least one o f the variables and if cointegration does not exist fo r the variables in x, the VAR in equation 6 can be estimated.19 The impulse response function fo r x in equation 15 shows that the long-run effect o f t converges to the sum o f coefficients in 0(L). It is obvious from the definition o f 0(L) that replacing L by one yields the sum o f coefficients. Hence, this sum is conveniently written as 0(1), and this matrix is used to parameterize long-run restrictions. The relationship between parameters o f the structural MAR, contem poraneous structural parameters and VAR lag coefficients is given by (16) 0(L) = [I-/?(L)L]~IA "1D. The long-run multipliers are obtained by replacing L in equation 16 with unity. Setting L equal to unity, solving equation 16 for A ’ d and inserting the result into equation 7 yields (17) [ l - / 3 ( l ) ] ’ 1Ee [I -P d )]" 1' = 6(1 )E £0 (1 )', In contrast to the atheoretical VAR models developed by Sims (1980), the structural approach yields impulse responses and variance decom positions that are derived using parameters from an explicit econom ic m odel.17 Finding dynamic patterns consistent with the structural model used for identification would provide evidence in support o f the theoretical model. Otherwise, the theory is invalid or the empirical model is som ehow misspecified. ,7The relationship between structural and atheoretical VARs is addressed in the shaded insert at right. 18For example, agents may temporarily be away from longrun equilibrium positions or monetary policy may be non neutral in the short run. 19Unit-root tests and cointegration tests support this assump tion for the time series used in this paper. See Keating (1992) for this evidence. FEDERAL RESERVE BANK OF ST. LOUIS where the matrix /J(l) is the sum o f VAR coeffi cients. This equation can be used to identify the parameters in 0(1) and E£. A minimal set o f restrictions on the long-run response o f macroeconom ic variables to structural disturbances is used to identify long-run structural VAR models. Estimates o f the matrices on the left side of 43 The Relationship between Atheoretical and Structural VAR Approaches Atheoretical VAR practitioners separate the residuals into orthogonal shocks by calculating a Choleski decom position o f the covariance matrix for the residuals. This decomposition is obtained by finding the unique lower triangular matrix A that solves the following equation: S e = AA'. This statistical decomposition depends on the sequence in which variables are ordered in x. The residuals’ covariance matrix from a VAR ordered by output, the interest rate, money and the price level yields a Choleski decom position that is algebraically equivalent to estimating the following four equations by ordinary least squares: e-v = v? e,r = R t f + K en; = R2e> + R3 e; + v™ ep = R4e[ + Rsetr + R6e™+ vp Hence, each v shock is uncorrelated with the other shocks by construction. This system implies that the first variable responds to its ow n exogenous shock, the second variable responds to the first variable plus an exogenous shock to the second variable, and so on. In practice, atheoretical VAR studies report results from various orderings. The total num ber of possible orderings o f the system is n!, a number that increases rapidly with n .1 Investigators sometimes note that certain properties o f the model are insensitive to alternative orderings. Results sensitive to VAR orderings are difficult to interpret, especially if a recursive econom ic structure is implausible. This atheoretical approach has been criticized by Cooley and LeRoy (1985). First, if the Choleski technique is in fact atheoretical, then the estimated shocks are not structural and will generally be linear combinations o f the structural disturbances.2 In this case, standard VAR analysis is difficult to interpret because the impulse responses and variance decompositions for the Choleski shocks will be complicated functions o f the dynamic effects o f all the structural disturbances. The second point attacks the claim that Choleski decompositions are atheoretical. The Choleski ordering can be interpreted as a recursive contemporaneous structural model. Unfortunately, most econom ic theories do not imply recursive contemporaneous systems. Such criticisms o f the atheoretical approach inspired structural approaches to VAR modeling. If therory predicts a con temporaneous recursive econom ic structure, a particular Choleski factorization o f the covariance matrix for the residuals is appro priate. But a researcher using the structural approach would not experiement with various orderings, unless these specifications w ere predicted by alternative theories. ’ For example, 3! = 6 but 6! = 720. 2This result is easy to prove. The Choleski decomposition yields a system in which e. = Rv. but the true structural model is e, = A “ 1Dct, implying tnat the shocks from the Choleski decomposition are linear combinations of the structural distrubances; v, = R -1 A “ 1D tt. SEPTEMBER/OCTOBER 1992 44 equation 17 are obtained directly from the unconstrained VAR.20 8(1) has n2 elements and Ee has n(n + l)/2 unique elements. The n(n + l)/2 unique elements in the symmetric matrix on the left side o f equation 17 is the number o f param eters in a just-identified model.21 Thus, at least n identifying restrictions must be applied to 8(1) and Z E. The elements o f the main diagonal for 8(1) can each be set equal to one, analogously to the normalization used in the contem po raneous model. If each element o f z is assumed to be independent, then £ . is diagonal. Hence, n ( n - l ) /2 additional restrictions are needed for 8(1) to identify the model. Several alternative approaches for obtaining the structural parameters have been developed. Shapiro and Watson (1988) impose the long-run zero restrictions on 8(1) by estimating the simultaneous equations model with particular explanatory variables differenced one additional time. King, Plosser, Stock and Watson (1991) impose long-run restrictions using the vector error-correction model with some o f the longrun features o f the model chosen by cointegration regressions. Gali (1992) combines contemporaneous restrictions with long-run restrictions to identify a structural model. In the empirical section, we use the approach developed by Blanchard and Quah (1989). Equations 18 through 21 present the long-run identifying restrictions used in the empirical example.22 The time subscripts are omitted because the restrictions pertain to long-run behavior. Three restrictions com e from equation 18, which specifies that aggregate supply shocks are the sole source o f permanent movements in output.23 Tw o m ore restrictions are obtained from the long-run IS or spending balance equation, 19, w hich specifies the interest rate as a function o f output and the IS shock.24 Note the coefficient Sj should be negative. The final restriction com es from the money demand 20The Bayesian approach is not employed since a unit root is required to be certain that shocks have permanent effects. 21An over-identified long-run model will imply restrictions on the reduced-form coefficients. 22This is a simplified version of the model in Keating (1992). 23lf interest rates affect capital accumulation, then IS shocks may permanently affect output and the restrictions in equation 19 may not be appropriate. If this is the only misspecification of the model, actual money supply and money demand shocks will be identified but the empirical aggregate supply and IS shocks will be mixtures of these real structural disturbances. FEDERAL RESERVE BANK OF ST. LOUIS function, 20, w hich sets real m oney equal to an increasing function o f output, a decreasing function o f the interest rate, and a money demand shock. Equation 21 allows the supply o f m oney to respond to all variables in the model and a m oney supply shock.25 (18) y = £as (19) r = S,y + £,s (20) m - p = S2y + S3r + £md (21) m = S ^ + S5r + S6(m -p ) + £ms EM PIRICAL EXAM PLES AND RESULTS The examples from the previous tw o sections are estimated to illustrate the long-run and contemporaneous identification methods. Both models use real GNP to measure output, the GNP deflator for the price level, M l as a measure o f the stock o f money, and the three month Treasury bill rate determined in the secondary market as the interest rate. The data are first-differenced. Statistical tests suggest that this transformation makes the data stationary.26 The first step is to estimate the reduced-form VAR model. The estimated variance-covariance matrix from the reduced form is used to obtain the second-stage structural estimates. Four lags are used for the VAR model and the sampleperiod is from the first quarter o f 1959 to the third quarter o f 1991. A Long-Run Structural Model Table 1 reports the parameter estimates for the long-run model in equations 18 through 21. The first four parameters are standard deviations for the structural shocks, and each o f these 24Technically, the IS equation should use the real interest rate, an unobservable variable. However, if permanent movements in the nominal rate and the real rate are identical, this specification is legitimate. This would be true, for example, if the Fisher equation held and if inflation followed a stationary time series process. 25Thus, 9(1) is: 1 0 -S . 1 0 0 0 0 -S 2 -S 3 1 0 -S 4 -S 5 -S 6 1 26For empirical evidence, see the unit-root tests in Keating (1992). 45 Table 1 Estimates for the Long-Run Model °as °is °md °ms si S2 S3 s4 S5 s6 Parameter Standard error .0144* .0092* .0149* .0172* -.117 1 .8722* -2 .2 7 6 * -1 .4 1 1 * 1.569 .9048* .0041 .0028 .0031 .0042 .3068 .4219 .7083 .5778 1.020 .3842 NOTE: An asterisk (*) indicates significance at the 5 percent level. estimates is significantly different from zero. The coefficient in the IS equation, S,, is negative as hypothesized, but insignificantly different from zero. Each parameter in the m oney demand function is statistically significant and has the sign predicted by econom ic theory. The coefficient on real GNP, S2, is not statistically different from one. Parameters for the m oney supply equation can be interpreted as a policy reaction function in which the Fed reduces m oney if output rises but increases m oney if interest rates rise. This last effect is not statistically significant. The increase in money in response to an increase in real money may reflect the fact that the Fed has typically smoothed interest rate fluctuations in the post war period, so that a money demand shock that raises real money will be accommodated by a comparable increase in nominal money. The impulse responses fo r the long-run model are shown in figures 1 through 4. Point estimates and 90 percent confidence intervals are graphed for the variables. If the long-run parameter estimates are consistent with the theoretical model, the asymptotic properties o f the impulse responses must also be consistent with the theory. Economic restrictions are not imposed on the dynamic properties o f the model. The empirical aggregate supply shock raises output and real m oney and lowers the interest rate, the price level and nominal money. The real spending shock raises output only temporarily because o f the restriction that aggregate supply shocks are the only factor in long-run output movements. The interest rate and the price level rise, while the nominal and real measures o f money decline after each variable initially rises by a small amount. The m oney demand shock has a strong positive effect on nominal and real money. The other effects are relatively weak, with prices falling, output temporarily falling and the interest rate temporarily rising. Money and the price level both rise in response to an increase in the m oney supply, w hich also causes a temporary decline in the interest rate and a temporary increase in output and real money. The impulse response functions provide evidence that the shocks affect each variable as theory predicts. The variance decompositions in table 2 show the average amount o f the variance o f each variable attributable to each shock. Standard errors for these estimates are in parentheses. The output variance due to the supply shock one quarter in the future is only 17 percent. Eight quarters in the future, however, the estimate becom es nearly half o f output’s variance and, at 48 quarters, 90 percent o f the variance o f output is attributed to supply shocks. Variability in the price level is dominated by aggregate supply shocks, particularly in the short run. The other shocks have temporary output effects. This long-run feature is obtained because the model forces aggregate supply shocks to explain all permanent output movements. Shortrun output movements are dominated by real spending shocks. This shock explains most of the interest rate variance and the variance of real m oney in the long run. The money supply shock has a gradual effect on output that peaks at 13 percent o f the variance tw o years in the future. This shock accounts for a large portion o f the short-run variance o f the interest rate and the long-run movement in nominal money. The money demand shock has virtually no effect on output, in terest rates o r prices. M an y th eories predict that m oney demand shocks will not have much effect on these variables if the Fed uses the interest rate as its operating target. Money demand shocks have strong effects on nominal m oney and real money. In general, the results for this model are consistent with most economists’ views about econom ic behavior, although some might be surprised by the relatively small effect on output o f money supply disturbances. Contemporaneous Structural Model The parameter estimates for the contem porane ous model in equations 8 through 11 are reported in table 3. The coefficients in the reducedform IS equation are all negative. The negative SEPTEMBER/OCTOBER 1992 46 Figure 1 Responses to an Aggregate Supply Shock in the Long-Run Model Output Price Interest Rate 0.03 Real Money http://fraser.stlouisfed.org/ FEDERAL RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis NOTE: The dashed lines enclose 90 percent confidence intervals which were calculated using Runkle's (1987) Monte Carlo simulation method. 47 Figure 2 Responses to a Real Spending Shock in the Long-Run Model Output Price Interest Rate Real Money NOTE: The dashed lines enclose 90 percent confidence intervals which were calculated using Runkle's (1987) Monte Carlo simulation method. SEPTEMBER/OCTOBER 1992 48 Figure 3 Responses to a Money Demand Shock in the Long-Run Model Price 0.006 0.004 0.002 0 - 0.002 -0.004 -0.006 -0.008 - 0.01 1 7 13 Interest Rate 19 25 31 37 43 49 37 43 49 Money 0.0225 0.02 0.0175 0.015 0.0125 0.01 0.0075 0.005 0.0025 0 1 7 13 19 25 31 Real Money http://fraser.stlouisfed.org/ FEDERAL RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis NOTE: The dashed lines enclose 90 percent confidence intervals which were calculated using Runkle's (1987) Monte Carlo simulation method. 49 Figure 4 Responses to a Money Supply Shock in the Long-Run Model Output Interest Rate Price Money 0.0016 Real Money NOTE: The dashed lines enclose 90 percent confidence intervals which were calculated using Runkle's (1987) Monte Carlo simulation method. SEPTEMBER/OCTOBER 1992 50 Table 2 Variance Decompositions for the Long-Run Model Variable Quarter(s) ahead Aggregate supply shock Real spending shock Money demand shock Money supply shock 1 2 4 8 16 32 48 17(24) 16 (23) 24 (25) 45 (23) 67 (15) 84 ( 8) 90 ( 5) 76 78 65 42 24 11 7 (24) (24) (24) (19) (13) ( 7) ( 5) 1( 1( 0( 0( 0( 0( 0( 8) 8) 6) 4) 2) 1) 1) 7(16) 5(14) 11 (14) 13(11) 9 ( 7) 5 ( 3) 3 ( 2) 1 2 4 8 16 32 48 13 (13) 14(13) 13(11) 8(10) 5(11) 4(13) 4(14) 49 57 71 84 91 94 95 (25) (23) (15) (12) (12) (13) (14) 1( 1( 3( 2( 1( 1( 0( 8) 7) 5) 3) 2) 1) 1) 37 (23) 27 (19) 13 ( 9) 6 ( 6) 3 ( 4) 1 ( 2) 1 ( 1) 1 2 4 8 16 32 48 7(11) 12 (14) 18 (17) 26 (19) 25 (19) 26 (21) 27 (22) 2(10) 3 ( 8) 8(12) 24 (17) 38 (20) 44 (21) 45 (21) 90 73 59 41 33 29 28 (17) (18) (19) (18) (18) (19) (19) 2(11) 12(14) 14 (14) 9(10) 4 ( 5) 2 ( 3) 1 ( 1) 1 2 4 8 16 32 48 6(10) 3 ( 9) 1 ( 9) 0(10) 1 (11) 4(13) 6(15) 5(14) 2(10) 3(11) 10 (15) 14(17) 11 (17) 9(17) 73 63 59 50 46 41 39 (22) (22) (23) (21) (20) (19) (19) 15 32 38 40 39 44 46 (20) (21) (22) (21) (20) (20) (20) 9(13) 5(10) 2 ( 8) 1 ( 7) 0 ( 7) 0 ( 7) 0 ( 7) 16 12 13 16 20 23 24 (23) (21) (21) (20) (19) (19) (19) Output Interest Rate Real Money Money Price 1 2 4 8 16 32 48 73 78 77 69 60 55 54 (27) (27) (28) (28) (27) (27) (27) NOTE: Standard errors are in parentheses. FEDERAL RESERVE BANK OF ST. LOUIS 2(12) 4(14) 7(16) 15 (19) 20 (21) 22 (21) 22 (22) 51 Table 3 Estimates for the Contemporaneous Model °as °is °ms °md Ai a2 A3 A4 A5 A6 Parameter .0038* .0086* .0087 .0087 -.1164 - .0469 - .3327 1.030 .5632 - .9397 Standard error .0003 .0027 .0419 .0324 .3255 .4122 .6539 8.302 2.098 5.737 NOTE: An asterisk (*) indicates significance at the 5 percent level. coefficient on money would be unexpected in a structural IS equation; however, these estimates are reduced-form coefficients, not structural pa rameters. The coefficient on m oney in the interest rate equation is positive. This supports the view that the central bank attempts to stabilize money growth by raising interest rates. In the money demand equation, the coefficient on nominal spending is roughly one-half, and the interest rate coefficient is almost -1 .0 . Hence, the parameter estimates in this structural model are consistent with econom ic theory. Unfortunately, each o f these structural para meters is statistically insignificant. Figures 5 through 8 plot the impulse responses. In contrast to the long-run model, the aggregate supply equation is normalized on the price level. Hence, an aggregate supply shock raises the price level and reduces output. The aggre gate supply shock has this expected effect on prices and output. Real money also decreases. The adverse supply shock has a weak positive effect on money and the interest rate. The IS shock raises prices, output and the interest rate. Real and nominal money initially increase, although both subsequently fall. The money supply equation is normalized on the interest rate so a reduction in the supply o f m oney raise interest rates. This shock raises the interest rate and causes a decline in nominal money, real money and the price level. Surprisingly, output rises briefly before it begins to decline. The money demand shock causes the interest rate, nominal and real money to increase while output falls. The rising price level is inconsistent with theory, although this effect is not statistically significant. In contrast with the longrun model, there are a few unusual features in the impulse responses for the contemporaneous specification. Most o f the dynamic patterns, however, are consistent with the structural model. The variance decompositions for the con temporaneous model are shown in table 4. Many features o f this table are comparable to the long-run model's results. For example, the aggregate supply shock gradually explains most o f output’s variability, is the most important shock for the price level and is never an important source o f interest rate movements. The IS shock is the most important source of short-run output movement, and it explains most o f the long-run variance o f the interest rate. Some results, however, are considerably different com pared with the results from the long-run model. The m oney demand shock has its greatest effect on output in the long run. This shock explains a large amount o f the shortrun variance o f the interest rate but virtually none o f the long-run variance o f real or nominal money balances. The money supply shock has essentially no effect on output, while accounting for a large amount o f the variance in real money, even in the long run, and none o f the variance o f prices. These results are inconsistent with most m acroeconom ic theories. CONCLUDING REM ARK S This paper outlines the basic theory behind structural VAR models and estimates tw o models using a standard set o f m acroeconom ic data. The results for the tw o specifications are often similar. The long-run structural VAR model in this paper generally provides empirical results that are consistent with the structural model. Some o f the variance decompositions and the impulse responses for the contemporaneous model w ere inconsistent with standard macroeconom ic theory. These inconsistencies pertain to the effects o f m oney supply and money demand disturbances. Another result is that structural parameters in the long-run model are more precisely estimated than parameters in the contemporaneous model. W herever a significant discrepancy exists between the two models, the model with long-run restrictions yields sensible results, while the results from the contemporaneous model are inconsistent with standard econom ic theories. SEPTEMBER/OCTOBER 1992 52 Figure 5 Responses to an Aggregate Supply Shock in the Contemporaneous Model Price Interest Rate FEDERAL RESERVE BANK OF ST. LOUIS Money NOTE: The dashed lines enclose 90 percent confidence intervals which were calculated using Runkle's (1987) Monte Carlo simulation method. 53 Figure 6 Responses to a Real Spending Shock in the Contemporaneous Model Output 0.015 0.0075 1 0.015J 1 > / 0.01250.01 - / 0.005 0.0025 A Price 0.02 0.010.005- \ n / 0 -0.0025 ----- 1-------1------- 1-------1------ 1-------1-------1-----1 7 13 19 25 31 37 43 49 Interest Rate 0 .006-1 -0.005- ' I i 13 ”! 19 ! .......! 25 31 ! 37 r 43 49 Money Real Money NOTE: The dashed lines enclose 90 percent confidence intervals which were calculated using Runkle's (1987) Monte Carlo simulation method. SEPTEMBER/OCTOBER 1992 54 Figure 7 Responses to a Money Supply Shock in the Contemporaneous Model Price Interest Rate FEDERAL RESERVE BANK OF ST. LOUIS NOTE: The dashed lines enclose 90 percent confidence intervals which were calculated using Runkle's (1987) Monte Carlo simulation method. 55 Figure 8 Responses to a Money Demand Shock in the Contemporaneous Model Output Price Interest Rate 0.01 0.0075- i \ , \ 0.0050.0025 V\ 0 -0.0025- \ _ _______ -0.005 NOTE: The dashed lines enclose 90 percent confidence intervals which were calculated using Runkle's (1987) Monte Carlo simulation method. SEPTEMBER/OCTOBER 1992 56 Table 4 Variance Decompositions for the Contempo raneous Model Variable Quarter(s) ahead Aggregate supply shock Real spending shock Money demand shock Money supply shock 2 ( 4) 3 ( 5) 1 ( 3) 2 ( 5) 2 ( 7) 2 ( 9) 2(10) Output 1 2 4 8 16 32 48 1 ( 2) 1 ( 3) 3 ( 5) 12 (10) 28 (16) 47 (21) 55 (23) 94 93 91 73 51 31 23 (16) (16) (15) (19) (20) (21) (22) 4(14) 3(13) 5(13) 13(14) 19 (14) 20 (14) 20 (14) 1 2 4 8 16 32 48 2 ( 2) 4 ( 4) 10 ( 7) 8 ( 6) 8 ( 7) 8(10) 8(11) 12 (12) 17(13) 38 (15) 56 (18) 63 (20) 67 (22) 68 (22) 37 (26) 34 (24) 27 (18) 19 (16) 12(15) 9(14) 8(14) 49 44 25 17 16 16 15 (31) (30) (23) (20) (20) (20) (20) 1 2 4 8 16 32 48 19 18 24 34 38 42 45 ( 7) ( 8) (11) (13) (16) (20) (21) 11 (14) 5(11) 2 ( 8) 4 ( 7) 10 (10) 14(12) 15(12) 36 (21) 16 (18) 6(16) 2(14) 1 (15) 0(15) 0(16) 34 60 68 61 51 44 40 (22) (20) (19) (19) (21) (23) (24) 2 ( 2) 1 ( 2) 0 ( 2) 0 ( 3) 1 ( 6) 4(10) 7(13) 14 (15) 7(12) 3(10) 1 ( 9) 1 (11) 1 (11) 0(11) 44 (24) 23 (22) 13 (21) 7(20) 6 (20) 7(20) 7(20) 41 68 83 92 92 89 86 (31) (27) (26) (25) (25) (25) (25) Interest Rate Real Money Money 1 2 4 8 16 32 48 Price 1 2 4 8 16 32 48 100 98 95 89 84 82 81 ( 0) ( 2) ( 4) ( 7) (10) (11) (12) NOTE: Standard errors are in parentheses. FEDERAL RESERVE BANK OF ST. LOUIS 0 ( 0) 0( 1) 2 6 12 14 15 ( 2) ( 6) ( 8) ( 9) (10) 0( 2( 3( 4( 4( 3( 3( 0) 1) 3) 5) 6) 7) 7) 0( 0( 0( 0( 0( 1( 1( 0) 1) 2) 5) 7) 9) 9) 57 These comparisons betw een contem poraneous and long-run specifications may not generalize to all structural VAR applications, but they sug gest that long-run structural VARs may yield theoretically predicted results m ore frequently than VARs identified with short-run restrictions. This result is not surprising. One reason is that econom ic theories may often have similar longrun properties but different short-run features. For example, while output movements are driven solely by aggregate supply shocks in a typical real business cycle model, supply shocks will ac count for permanent output movements in Keynesian models, but every shock may have some cyclical effect. In addition, long-run struc tural VAR models may provide superior results because they typically do not impose contem poraneous exclusion restrictions. Keating (1990) shows that contem poraneous "zero” restrictions may be inappropriate in an environ ment with forward-looking agents w ho have ra tional expectations. The intuition behind this result is that any observable contemporaneous variable may provide information about future events. One implication from that paper is that different short-run restrictions can be obtained from alternative assumptions about available in formation. Further research should investigate other examples o f contem poraneous and longrun structural VAR models to determine whether the superior perform ance o f this paper’s long-run model is a special case or a m ore general result. REFERENCES Bernanke, Ben S. “ Alternative Explanations of the MoneyIncome Correlation,” Carnegie-Rochester Conference Series on Public Policy (Autumn 1986), pp. 49-100. Blanchard, Olivier Jean. “ A Traditional Interpretation of Mac roeconomic Fluctuations,” American Economic Review (De cember 1989), pp. 1146-64. Blanchard, Olivier Jean, and Danny Quah. “ The Dynamic Ef fects of Aggregate Demand and Supply Disturbances,” American Economic Review (September 1989), pp. 655-73. Blanchard, Olivier Jean, and Mark W. Watson. “ Are Business Cycles All Alike?” in Robert J. Gordon, ed., The American Business Cycle (University of Chicago Press, 1986). Cooley, Thomas F., and Stephen F. LeRoy. “ Atheoretical Macroeconomics: A Critique,” Journal of Monetary Eco nomics (November 1985), pp. 283-308. Dickey, David A., and Wayne A. Fuller. “ Distribution of the Estimators for Autoregressive Time Series With a Unit Root,” Journal of the American Statistical Association (June 1979), pp. 427-31. Doan, Thomas, Robert B. Litterman, and Christopher Sims. “ Forecasting and Conditional Projection Using Realistic Prior Distributions,” Econometric Review (Vol. 3, No. 1,1984), pp. 1-100. Engle, Robert F., and C. W. J. Granger. “ Co-Integration and Error Correction: Representation, Estimation, and Testing,” Econometrica (March 1987), pp. 251-76. Fuller, Wayne A. Introduction to Statistical Time Series (Wiley Press, 1976). Gali, Jordi. “ How Well Does the IS-LM Model Fit Postwar U.S. Data,” Quarterly Journal of Economics (May 1992), pp. 709-38. Hansen, Lars Peter. “ Large Sample Properties of Generalized Method of Moments Estimators,” Econometrica (July 1982), pp. 1029-54. Keating, John W. “ Postwar Business Cycles in a Natural Rate Model,” revised version of Washington University Working Paper #159 (St. Louis, MO, July 1992). _______ “ Identifying VAR Models under Rational Expecta tions,” Journal of Monetary Economics (June 1990), pp. 453-76. Keating, John W., and John V. Nye. “ Permanent and Transitory Shocks in Real Output: Estimates from Nineteenth Century and Postwar Economies,” Washington University Working Paper #160 (St. Louis, MO, July 12, 1991). King, Robert G., Charles I. Plosser, James H. Stock, and Mark W. Watson. “ Stochastic Trends and Economic Fluc tuations,” American Economic Review (September 1991), pp. 819-40. Nelson, Charles R., and Charles I. Plosser. “ Trends and Random Walks in Macroeconomic Time Series: Some Evi dence and Implications,” Journal of Monetary Economics (September 1982), pp. 139-62. Runkle, David E. “ Vector Autoregressions and Reality,” Journal of Business and Economic Statistics (October 1987), pp. 437-42. Sargent, Thomas J. “ A Note on Maximum Likelihood Estimation of the Rational Expectations Model of the Term Structure,” Journal of Monetary Economics (January 1979), pp. 133-43. Schwert, G. William. “ Effects of Model Specification on Tests for Unit Roots in Macroeconomic Data,” Journal of Monetary Economics (July 1987), pp. 73-104. Shapiro, Matthew D., and Mark W. Watson. “ Sources of Business Cycle Fluctuations,” in Stanley Fischer, ed., NBER Macroeconomics Annual 1988 (MIT Press, 1988), pp. 111-48. Sims, Christopher A. “ Are Forecasting Models Usable for Policy Analysis?” Federal Reserve Bank of Minneapolis Quarterly Review (Winter 1986), pp. 2-16. _______ “ Macroeconomics and Reality,” Econometrica (January 1980), pp. 1-48. Sims, Christopher A., James H. Stock, and Mark W. Watson. “ Inference in Linear Time Series Models with Some Unit Roots,” Econometrica (January 1990), pp. 113-44. Stock, James H., and Mark W. Watson. “ A Simple Estimator of Cointegrating Vectors in Higher Order Integrated Systems,” Federal Reserve Bank of Chicago Working Paper WP-91-4, (February 1991). _______ “ Testing for Common Trends,” Journal of the American Statistical Association (December 1988), pp. 1097107. _______ “ Variable Trends in Economic Time Series,” Journal of Economic Perspectives (Summer 1988), pp. 147-74. West, Kenneth D. “ The Sources of Fluctuations in Aggregate Inventories and GNP,” Quarterly Journal of Economics (November 1990), pp. 939-71. Zellner, Arnold. “ An Efficient Method of Estimating Seeming ly Unrelated Regressions and Tests for Aggregation Bias,” Journal of the American Statistical Association (June 1962), pp. 348-68. SEPTEMBER/OCTOBER 1992 58 Anna J. Schwartz, Anna J. Schwartz is a Senior Research Fellow at The National Bureau of Economic Research. This paper was the sixth annu al Homer Jones Memorial Lecture, which was presented at St. Louis University on April 9, 1992. The views expressed are those of the author and do not reflect official positions of the Board of Governors of the Federal Reserve System or the Fed eral Reserve Bank of St. Louis. The Homer Jones Memorial Lecture is an annual event sponsored by the St. Louis Gateway Chapter of the National Association of Business Economists in conjunction with Washington University, St. Louis University, the University of Missouri-St. Louis and the Federal Reserve Bank of St. Louis. Author’s note: For sharing his knowledge of Federal Reserve practices, Walker F. Todd has been immensely helpful to me, but he is not responsible for the views I express in this paper. I have also benefited from comments by Allan H. Meltzer. The Misuse of the Fed's Discount Window I AM HONORED to have the opportunity to give the Homer Jones Memorial Lecture. In remembering him today, I pay tribute to his contributions to monetary policy making and to quantitative monetary research in his capacity as director o f research at the St. Louis Fed. I got to know Homer during the period o f his membership in the Shadow Open Market Com mittee. At our meetings he was diffident about his knowledge and insistent on scrupulous at tention to statistical evidence as backing for the policy conclusions w e reached. I cannot think of tw o more admirable qualities in an economist. It was a privilege for me to have had this associa tion with Homer. In 1925 the Federal Reserve Board collected data on the number o f m em ber banks continu ously indebted to their Reserve Banks for at least a year. As o f August 31, 1925, 593 m em ber banks had been borrow ing for a year or more. Of this number, 239 had been borrow ing since 1920 and 122 had begun borrow ing before that. The Fed guessed that at least 80 percent o f FEDERAL RESERVE BANK OF ST. LOUIS the 259 national member banks that had failed since 1920 had been "habitual borrow ers” prior to their failure. Of 457 continuous borrow ers in 1926, 41 banks suspended operations in 1927, while 24 liquidated voluntarily or merged (Shull 1971, 34-35). The reason for citing these statistics for the 1920s is to call attention to an early episode in Federal Reserve history that contravened the ancient injunction to central banks to lend only to illiquid banks, not to insolvent ones, and that is eerily similar to a current episode. The Fed apparently learned little from the earlier epi sode, since there is even less justification for the use made o f the discount window in the current than in the earlier episode. The current episode came to light after the House Banking Committee requested data on all insured depository institutions that borrow ed funds from the discount w indow from January 1, 1985, through May 10, 1991. Regulators grade banks on their perform ance, according to 59 a scale o f 1 to 5. The grades are based on five measures known by the acronym o f CAMEL, for Capital adequacy, Asset quality, Manage ment, Earnings, Liquidity.1 The Federal Reserve reported that o f 530 borrow ers from 1985 on that failed within three years o f the onset of their borrowings, 437 w ere classified as most problem-ridden with a CAMEL rating o f 5, the poorest rating; 51 borrow ers had the next lowest rating, CAMEL 4. One b orrow er with a CAMEL rating o f 5 remained open for as long as 56 months. The whole class o f CAMEL 5-rated institutions w ere allowed to continue operations for a mean period o f about one year. At the time o f failure, 60 percent o f the b o r row ers had outstanding discount w indow loans. These loans w ere granted almost daily to insti tutions with a high probability o f insolvency in the near term, new borrow ings rolling over balances due. In aggregate, the loans o f this group at the time o f failure amounted to $8.3 billion, o f which $7.9 billion was extended when the institutions w ere operating with a CAMEL 5 rating. Three months prior to failure, b orrow ings o f all 530 institutions peaked at $18.1 bil lion. Rather than encouraging banks to pursue strategies to preserve their size, regulators often encourage institutions that are about to fail to shrink drastically first, so as to diminish the pool o f assets that have to be liquidated after closing. Some observers o f bank perform ance have as serted that it is impossible to know whether an institution that applies for discount w indow as sistance faces a liquidity or solvency problem. That assertion may be defensible for discount w indow lending in the 1920s even though an estimated 80 percent o f long-time borrow ers 1Brief official descriptions of composite CAMEL 4 and 5 rat ings follow: CAMEL 4 “ Institutions in this group have an immoder ate volume of serious financial weaknesses or a combi nation of other conditions that are unsatisfactory. Major and serious problems or unsafe and unsound conditions may exist which are not being satisfactorily addressed or resolved." CAMEL 5 “ This category is reserved for institutions with an extremely high immediate or near-term probability of failure.” CAMEL ratings of banks are not uniform from one district to another. Some New York CAMEL 4 banks may be rated 5 elsewhere. failed, since CAMEL ratings did not then exist. Currently, CAMEL ratings 4 and 5 are known promptly. W hy should it be impossible or even difficult to distinguish between an illiquid and an insolvent bank? Support by the Fed for banks with a high probability o f insolvency in the near term is not the only recurrent problem with the discount window. Equally troublesome is the history of actual or proposed Fed capital loans to non banks. Such use o f the discount w indow dis tracts the Fed’s attention from its monetary control function. Recent experience reinforces earlier doubts about the need for the discount window. The time has com e for a truly basic change: eliminate the discount w indow and retrict the Fed to open market operations.2 This change would have the added value o f obliterat ing the symbolic role o f the discount rate and weakening the tendency to regard the Fed as determining interest rates. In the rest o f this paper, I document the ero sion o f the historic restriction, at least since the 1930s, o f Federal Reserve discount w indow assis tance to liquidity-strained banks on the security o f sound assets. Section 1 deals with lending operations from the founding until the postW orld War II period, during which loans to nonbanks first occur. I then discuss Federal Reserve actions in recent decades that have fu r ther blurred the distinction between liquidity and solvency, and also the emergence o f various nonbanks as candidates for discount w indow as sistance. I ask why these developments have o c curred w hen there has been no change in official declarations o f commitment to supply only li quidity, not solvency or capital, to individual banks, not nonbanks (section 2). In the next sec- changes would eliminate the problem of political pressures on the Fed to lend to nonbanks. As for the problem of loans to insolvent banks, access to the window as a right at a penalty rate might only result in worsening adverse selection. See also Kaufman (1991), who argues that the discount window is not needed to protect the money supply — the basic justification for a lender of last resort — and that li quidity strains can be mitigated by open market operations without involving the Fed in discount window assistance. Credit-worthy banks can borrow at market rates, large ones in the Fed Funds market, small ones from their correspon dent banks. 2This recommendation has been disputed on the ground that establishing access to the discount window as a right — not a privilege — administered at a penalty rate would solve the problems that face the current administration of the window. It is not clear to me, however, that these SEPTEMBER/OCTOBER 1992 60 tion I examine the costs o f Federal Reserve sup port for problem institutions that regulatory authorities eventually close and for nonbanks that would otherwise have to meet a market test (section 3). Finally, I consider whether re form s o f discount w indow practices that have been proposed could rem edy the inherent problems o f the mechanism. I comment on p ro visions in the FDIC Improvement Act o f 1991 that may be w orthy reform proposals but do not address these problem s (section 4). I offer my conclusions in section 5. appears in an internal Federal Reserve history o f the discount mechanism: "extended b o rro w ings by a m em ber bank from its Reserve Bank would in effect constitute a use o f Federal Reserve credit as a substitute for the m em ber’s capital” (Hackley 1973, 194). The 1973 version o f Regulation A states, as a general principle, that "Federal Reserve credit is not a substitute for capital and ordinarily is not available for ex tended periods.” Both the 1980 and 1990 ver sions o f Regulation A state, as a general require ment, that "Federal Reserve credit is not a sub stitute for capital.” 1. H ISTORIC ROLE OF DISCOUNT W IN D O W LENDING A broader statement o f the foregoing princi ple, covering banks and nonbanks, appeared in 1932 in a conference report by representatives o f the Federal Reserve Bank o f New York, w ho had met with South American central banks: "Central banks must not in any way supply cap ital on a permanent basis either to m ember banks or to the public, which may lack it for the conduct o f their business” (Federal Reserve Bulletin 1932, 43). A. B efore the New Deal Regulation A—the first one adopted by the Federal Reserve Board at its creation, in recogni tion o f the expectation that the discount w indow at Federal Reserve Banks would serve as the main purveyor o f m em ber bank reserves— establishes the rules under w hich the Banks may extend credit. Periodically revised, the regu lation has consistently set out the procedures that banks had to follow to gain access to the discount w indow . The initial regulation provid ed for rediscounting short-term agricultural, in dustrial, and commercial paper, defined as eligible paper. Later, to accommodate Treasury financing needs in W orld W ar I, the Reserve Banks w ere em pow ered to extend direct col lateral loans to m em ber banks, occasionally se cured by pledge o f eligible paper but usually by obligations o f the U.S. government. Until the 1930s, even though the Federal Reserve had be com e familiar with open market purchases as a source o f member bank reserves, bills discount ed usually exceeded U.S. government securities in Reserve Bank portfolios. The discount w indow provided accom m oda tion for periods up to 90 days for a nonagricultural discount or advance collateralized by eligible paper or governm ent obligations, but o f up to nine months for agricultural paper. As noted earlier, continuous borrow ing year in and year out in the 1920s was not uncom m on. A later (1954) Federal Reserve document, deploring continuous borrow ing, noted that "it was possi ble by the mid-Thirties to speak o f an estab lished tradition against member bank reliance on the discount facility as a supplement to its resources” (Shull 1971, 41). A similar statement http://fraser.stlouisfed.org/ FEDERAL RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis Legislative changes in the administration of the discount window, made in response to bank failures in the Great Depression, sometimes o b served this advice. The Glass-Steagall Act o f February 27, 1932, authorized Federal Reserve Banks (in a new section 10(b) added to the Fed eral Reserve Act) to make advances to member banks on their promissory notes secured by otherwise ineligible collateral, if acceptable to the Reserve Banks, for periods up to four months at rates not less than one-half percent per an num above the prevailing highest discount rate. No provision was made for solvency loans of capital or loans to receivers o f closed banks. Although no provision was made for solvency loans o f capital, the Emergency Relief and Con struction Act o f July 21, 1932 (in a new section 13(3) added to the Federal Reserve Act), opened the discount w indow to nonbanks. It permitted the Reserve Banks to discount for individuals, partnerships and corporations, with no other sources o f funds, notes, drafts and bills o f ex change eligible for discount for member banks. The New Deal confirm ed this type o f authority (in section 403 o f Title III o f the Emergency Banking Act o f March 9, 1933, that added section 13(13) to the Federal Reserve Act). It allowed 90-day advances to individuals, partnerships and corporations on the security o f direct obliga tions o f the U.S. government, at interest rates fixed by the Reserve Banks. 61 B. From the New Deal to PostWorld II Before turning to the New Deal change that involved Reserve Banks in extending capital loans to nonbanks, let me review the other changes in the Emergency Banking Act o f 1933. Title II created conservatorships for national banks. Section 304 of Title III authorized the Reconstruction Finance Corporation, not the Federal Reserve, to subscribe to preferred stock "o f any national banking association or any State bank or trust com pany in need o f funds for capital purposes.” It is significant that Title IV, referring to the Federal Reserve, conferred on it no authority to make solvency loans of capital. Section 402 authorized Federal Reserve Banks, until March 3, 1934, to make advances in exceptional and exigent circumstances to mem ber banks on their ow n notes on the security of any acceptable assets, superseding the provision regarding advances to m em ber banks in the February 1932 Glass-Steagall Act. By Presidential proclamation, this provision, the forerunner of the present section 10(b), was extended until March 3, 1935, w hen it expired (Board Annual Report 1933, 261-265). The new form o f section 10(b) becam e permanent as part o f the Banking Act o f 1935. I now turn to the New Deal change that authorized the Federal Reserve to make solvency/ capital loans to nonbanks. The Act o f June 19, 1934, added a new section 13(b) to the Federal Reserve Act, which authorized Reserve Banks directly or in participation with a member or nonmem ber bank to make advances to estab lished commercial or industrial enterprises for the purpose o f supplying working capital if the borrow er w ere unable to obtain assistance from usual sources. A participating m em ber or nonmember bank had to obligate itself for at least 20 percent o f any loss. The loans w ere for peri ods up to five years (Board A nnual Report 1934, 50-51). Through 1939, the Reserve Banks had approved 2,800 applications and commitments amounting to $188 million at rates from 2.5 to 6 percent (Smead 1941, 259). Although the Recon struction Finance Corporation then assumed a more important role in providing working capi tal for established businesses, the Reserve Banks continued to participate, approving an additional 742 applications amounting to $382 million through 1946 (Board Annual Report 1946, 10). The aggregate amount of such loans was limit ed by law to the surplus o f the Reserve Banks as o f July 1, 1934, plus $139 million that the Treasury was to repay the Banks for their re quired subscription to the Federal Deposit In surance Corporation in an amount equal to one-half o f their surplus on January 1, 1933. The required subscription was stipulated in the Banking Act o f 1933 that created the FDIC (Board Annual Report 1933, 276-277). A com m en tator has noted that this "is the only instance in United States history in which Congress re quired the central bank to expend its ow n funds to subscribe for more than a de m inim is amount o f the capital o f another unrelated en terprise, other than obligations o f the Treasury itself” (Todd 1988, 60). In any event, the Reserve Banks charged o ff the value o f FDIC stock on their books in the same calendar year in which they paid fo r the subscription. Capital loans to nonbanks under the authority o f sec tion 13(b), unlike the FDIC subscription, w ere voluntary decisions o f the Reserve Banks. In section 2 below, I note a more recent attempt by the Treasury that was foiled to require the Fed to expend its ow n funds in support o f the FDIC. Congressional authorization and Federal Reserve implementation o f loans to nonbanks for use as capital was, in my judgment, a sorry reflection on both Congress's and the Fed’s un derstanding o f the System’s essential monetary control function. In 1946 the Federal Reserve Board sought to eliminate its authority under section 13(b) to make loans directly to business enterprises and replace it with a mandate re stricted to guaranteeing such loans. A bill in troduced in Congress early in 1947 on the Board's behalf would have authorized Reserve Banks to guarantee, up to a maximum o f 90 percent, loans, extended by banks for as long as 10 years to small- and medium-size business en terprises, subject to a fee charge that increased with the guarantee percentage. The bill also provided for the return o f the amounts ad vanced by the Treasury (up to $139 million) for the Banks’ industrial loan operations, and pledged that no further Treasury appropriations for this purpose would be required (Board A n nual Report 1946, 8-10; 1947, 11-12). Since the bill was not enacted, Reserve Banks for the next decade continued to make and co finance working capital industrial loans. Section 13(b) was finally repealed by the Small Business Investment Act o f 1958, under the terms of which the Reserve Banks restored to the Trea sury the amounts it had advanced under section SEPTEMBER/OCTOBER 1992 62 13(b) (Board A nnual Report 1958, 95). Chairman William McChesney Martin, in testimony before the Subcommittee on Small Business o f the Senate Banking and Currency Committee, when it was considering this bill, stated well the Fed eral Reserve’s considered judgment on its ven ture into capital industrial loans: Its primary objective was “guiding monetary and credit poli cy,” and "it is undesirable for the Federal Reserve to provide the capital and participate in management functions" o f the proposed small business financing institutions (Federal Reserve Bulletin 1957, 769). I conclude this survey o f Federal Reserve lending activities since its founding by noting its support for solvency/capital lending programs under wartime V-loan authority that it adopted on April 6, 1942, and revised on September 26, 1950 (Board Annual Report 1942, 89-91; 1950, 72-74). Federal Reserve Banks w ere authorized to act on behalf o f the guaranteeing agencies (War Department, Navy Department, etc.) as fis cal agents o f the United States, meaning, the Treasury was required to reimburse them for their outlays. In acting as a guarantor o f defense production loans, the Federal Reserve provided a model o f guaranteeing authority that was later invoked w hen bailouts o f peacetime enter prises w ere proposed in the 1970s. Those de velopments and Federal Reserve lending since the 1980s to institutions with a high probability o f insolvency in the near term represent a major departure from its historic mandate to provide loans to illiquid but not insolvent depository institutions. In the section that fol lows I discuss the apparent change in how the Federal Reserve regards its mandate. 2 . ASSISTANCE T O INSOLVEMENT NONBANKS AN D BANKS SINCE THE 1 9 7 0 s A. Assistance to Insolvent Nonbanhs An appropriate starting point is the official response to financial problems o f the Penn Cen tral Railroad that surfaced in 1970. Though it did not lead to discount w indow assistance, nonetheless it reveals clearly the pressures that w ere to lead to such a practice. The Nixon Ad ministration proposed to give the com pany a Vloan as a bailout. However, for the loan to be of any use, legislative approval was required, since guarantees o f loans for defense production were http://fraser.stlouisfed.org/ FEDERAL RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis to expire on June 30. W hen legislation stalled in Congress, the Administration requested the Fed eral Reserve Board to authorize the Federal Reserve Bank o f New York to give the company discount w indow assistance. The request was rejected and, as a result, the com pany filed for bankruptcy on June 21, 1970. On June 30, Con gress approved a Joint Resolution, extending the Defense Production Act but limiting the maxi mum obligation o f any guaranteeing agency (for example, the Federal Reserve) to $20 million, and stipulating that "The authority conferred by this section shall not be used primarily to pre vent the financial insolvency or bankruptcy o f any person, unless" the President certifies “a direct and substantially adverse effect upon defense production" (Federal Reserve Bulletin 1970, 720). If the incident had ended at this point, it would have been rem em bered primarily as a political dispute between the Administration and the Congress. Penn Central’s bankruptcy would have been regarded simply as the key to the restructuring o f its operations. Instead, the Fed eral Reserve reacted as though the com pany’s default on $82 million o f commercial paper it had outstanding w ould generate a financial cri sis. The Federal Reserve assumed that lenders w ould not discriminate between a troubled issuer and other perfectly sound issuers o f commercial paper, so that the latter would be unable to roll over their issues. "It was made clear that the Federal Reserve discount w indow w ould be available to assist banks in meeting the needs of businesses unable to roll over maturing com mercial paper” (Board Annual Report 1970, 18). However, commercial paper issuers that faced difficulties did so not because o f the condition o f the market as such, but because o f condi tions peculiar to themselves (for example, Chrys ler Financial and Commercial Credit) (Carron 1982, 398). The fully justified verdict o f the 1971 Economic Report o f the President (69) was that no "genuine liquidity crisis existed in mid-1970." The Penn Central episode fostered the view that bankruptcy proceedings by a large firm created a financial crisis, and that, if possible, bankruptcy should be prevented by loans and loan guarantees; the "too big to fail” doctrine in embryo. The financial difficulties faced by New York City in 1975 led the Federal Reserve to inform 63 Congress o f its response to suggestions that it might serve as a source o f em ergency credit. Governor George W. Mitchell cautioned “against any proposals that would provide direct access to central bank credit by hard-pressed govern mental units" (Federal Reserve Bulletin 1975, 410). Chairman Arthur F. Burns reiterated that caution and reported on contingency plans to increase temporary discount w indow lending to com m er cial banks in the event o f a major municipal default. However, he added that if a default ulti mately required "write-downs that could seri ously impair the capital o f some banks,” the Federal Deposit Insurance Corporation, not the Federal Reserve, had statutory pow ers to assist federally insured banks that found their capital impaired (Federal Reserve Bulletin 1975, 635-636). In the end, Congress passed a law guaranteeing New York City loans o f up to $2.3 billion in 1975, reduced to $1.65 billion in 1978, but the Federal Reserve’s involvement in the rescue ar rangement was only limited fiscal agency serv ices.3 The Fed also acted as fiscal agent for Treasury loan guarantees issued during the bailouts o f the Lockheed (1971) and Chrysler (1979) corporations.4 I have been unable to find any reference to the Fed's involvement in these three loan guaran tees in any o f its publications. By consulting the U.S. Code—a source for lawyers, not econ o mists—however, I have been able to ferret out the details o f that involvement. My guess is that the Fed was unwilling to publicize its role be cause it was not voluntary. The most recent attempt to use the discount w indow to assist a nonbank involved the FDIC. In March 1991, the FDIC chairman, William Seidman, requested Congressional authorization for a direct loan o f $25 billion by the Federal Reserve to the Bank Insurance Fund. Chairman Alan Greenspan, testifying before the Senate Banking Committee the next month, opposed such a loan. That did not deter Treasury Under Secretary for Domestic Finance Robert Glauber from renewing the request in testimony on May 29, 1991, before a subcommittee o f the House Ways and Means Committee. The Federal Reserve’s required subscription to the FDIC on its establishment in his view served as a prece dent. The FDIC recapitalization and banking re form bill that the Treasury prepared included 3See U.S. Code (1975, 1978). provisions authorizing the FDIC to borrow $25 billion from the Federal Reserve Banks and amending section 13 o f the Federal Reserve Act, authorizing any Federal Reserve Bank "to make advances to the Federal Deposit Insurance Cor poration, upon its request” (Treasury bill 1991, sec. 402, 245). The July 23, 1991, bill prepared by the House Committee on Banking, Finance and Urban Affairs did not include those provi sions (H.R.6, 102nd Cong., 1st sess.). The final legislation, the FDIC Improvement Act o f 1991, follows the House bill in increasing from $5 bil lion to $30 billion the FDIC’s authority to b o r row directly from the Treasury, not from the Fed. Despite the restraint in the Act with respect to recapitalizing the FDIC, restraint is absent from another provision. The Act amended Sec tion 13 o f the Federal Reserve Act that deals with the Federal Reserve’s authority to discount for nonbanks. The amendment eliminated the requirement that the notes, drafts or bills ten dered by nonbanks be eligible for discount by member banks. As interpreted by Sullivan & Cromwell, a New York law firm, for its clients in a memorandum o f Decem ber 2, 1991, this provision enables the Fed to lend directly to security firms in em ergency situations. Tradi tionally, commercial banks, knowing they had access to the discount w indow, have lent to brokerage firms and others short o f cash in a stock market crash. It is not clear w hy the traditional practice was deemed unsatisfactory. In my view, the provision in the FDIC Improve ment Act o f 1991 portends expanded misuse o f the discount window. To this date, the Fed has apparently been a reluctant participant in loans and loan guaran tees to nonbanks. The question must be asked whether it will be firm in the future in resisting pressures to fund insolvent firms that are politi cally well-connected. B. Assistance to CAMEL 4- and 5-Rated Banks In 1974 the Federal Reserve behaved contrary to traditional principles when uninsured deposi tors started a run on the Franklin National Bank after news surfaced that it had large foreign ex change losses.The Comptroller o f the Currency did not close it promptly. The decision o f the “See U.S. Code (1971) for Lockheed and U.S. Code (1980) for Chrysler. SEPTEMBER/OCTOBER 1992 64 regulators was that the Federal Reserve discount window, starting in May, would provide loans until the FDIC found a purchaser o f the failed institution. Over the next five months, the Fed eral Reserve Bank o f New York lent continuous ly to Franklin; the maximum amount lent, on October 7, 1974, was $1.75 billion, representing nearly one-half o f Franklin’s assets. On October 8, the bank was declared insolvent and taken over by a foreign consortium. Among the precedents established by discount w indow lending to Franklin National was that its London branch assets w ere accepted as col lateral, and that, for the first time, the b orrow ings covered withdrawals from the London branch as well as Franklin's domestic branches. Although, on one hand, Franklin National simply borrow ed at the discount rate, what was un usual in this episode was that in September 1974 the Federal Reserve assumed responsibility to execute Franklin’s existing foreign exchange contracts, since bidders for the bank w ere un willing to honor them. It also agreed to extend discount w indow assistance, if needed, to the purchasing bank. The FDIC, m oreover, did not immediately repay the discount w indow loan— its normal practice—but signed a three-year note obligating itself to do so as the collateral Franklin supplied was liquidated. In effect, the Federal Reserve lent capital funds that the in surance agency contributed to the purchasing bank. The interest cost to the Fed o f subsidizing loans to Franklin has been estimated at $20 mil lion (Garcia and Plautz 1988, 228). In executing Franklin ’s foreign exchange contracts, the Fed also incurred opportunity costs o f staff time and lost interest on part o f its portfolio. The rescue o f Franklin National Bank shifted discount w indow use from short-term liquidity assistance to long-term support o f an insolvent institution pending final resolution o f its prob lems. The bank was insolvent when its b orrow ing began and insolvent when its borrow ing ended. The loans merely replaced funds that depositors withdrew, the inflow from the Reserve Bank matching withdrawals. The undeclared insolvency o f Continental Il linois in 1984 was also papered over by exten sive discount w indow lending from May 1984 to February 1985, albeit with smaller subsidies than in the case o f Franklin National—an amend ment to Regulation A as o f September 25, 1974, permitted application o f a special rate on em er gency credit after eight weeks that was closer FEDERAL RESERVE BANK OF ST. LOUIS to a market rate. The borrow ing covering Con tinental’s holding com pany as well as the bank at some dates amounted to as m uch as $8 billion. Again the FDIC assumed the bank’s loan, which it eventually repaid from the proceeds o f liq uidating the bank’s assets, concluding with one large $2.1 billion payment in September 1989— an enormous cash drain. The discount w indow has been valued by the Federal Reserve as a mechanism for directing funds to an individual bank with liquidity problems. It regarded its loans to Franklin Na tional and Continental Illinois as exceptional o c currences. However, when hundreds o f CAMEL 4- and 5-rated banks, as noted at the beginning o f this paper, w ere receiving extended accom modation even though they faced a high proba bility o f near-term failure, discounting can no longer be regarded simply as a means o f provid ing temporary liquidity. What explains this transformation in practice? C. Why the Departure From the Historic Norm o f Discount Window Use? In the United States, with federal spending budgeted at an all-time high, policy makers see the discount w indow as a mechanism for providing funds o ff budget. Legislation is neces sary to authorize the Fed to provide assistance to favored nonbanks, so the use o f the window isn't kept secret, but it may seem a cost-free way o f funding them because repayment can be rolled forw ard indefinitely. This may explain the recent spate o f efforts to use discount win dow assistance fo r nonbanks. With respect to loans to banks with a high probability o f insolvency in the near term, it is noteworthy that in 1974 only four banks failed. By the late 1980s, failures w ere num bered in triple digits, and the FDIC’s problem banks, in four digits. Having once set the precedent of lending to such problem institutions, at least tw o explanations may account for this practice by the Federal Reserve: (1) As in the case of Franklin and Continental Illinois Banks, its ac tions may have been taken at the bidding o f the FDIC, or perhaps on its ow n initiative, in order to mitigate the FDIC’s plight. (2) The Federal Reserve may regard failure o f a large institution as potentially triggering a financial collapse or a run on the dollar. Fear o f contagion may have becom e the Federal Reserve’s overriding concern. 65 Even if these explanations account for the change in Federal Reserve discount window practice, neither may justify the change. Before dealing with this issue, it is important to assess the costs o f wholesale discount w indow lending to insolvent institutions, to which I now turn. 3. COSTS OF LENDING T O INSOL VENT INSTITUTIONS For the Fed to lend directly to the Treasury or to government agencies that the Treasury would otherwise fund through regular ap propriations is a slippery slope. The costs are politicization o f the m oney supply process. The Fed’s charter wisely prohibits such lending. Discount w indow accommodation to insolvent institutions, whether banks or nonbanks, misallocates resources. Political decisions substitute for market decisions. Institutions that have failed the market test o f viability should not be supported by the Fed's m oney issues. A depository institution traditionally was said to be eligible for discount w indow assistance when it was illiquid but solvent. In recent years, it has been given assistance when it was liquid but its insolvency was undeclared. On a market value basis, an institution is insolvent w hen assets are less than liabilities. Since book values are the usual measure o f assets and lia bilities, the divergence between assets and liabil ities may only be revealed long after the market value o f its assets has fallen below the market value o f its liabilities. In addition, an institution may be liquid but insolvent, so long as its cash flow is positive. A decision to declare an institution insolvent is the prerogative o f the chartering agency: the Comptroller o f the Currency for national banks, state authorities for state-chartered banks su pervised by the Fed and the FDIC. The FDIC, since 1989, however, can both rem ove deposit insurance and substitute itself as receiver o f state banks. It can force a state to close banks and, as the thrift insurer, close insolvent in sured thrifts. (It does not have authority to close credit unions.) If the chartering agency has delayed closure, the Federal Reserve has acted as if a troubled institution is entitled to discount window assistance provided it can fur 5lt was only extended credit borrowers that tailed in the data the House Banking Committee obtained from the Fed in 1991. Banks that obtained seasonal credit at the win nish acceptable collateral. The question I raise is whether the Federal Reserve's position is defen sible when inferior CAMEL ratings provide in dependent evidence on the likelihood o f insol vency in the near or immediate future. Since the Federal Reserve routinely sterilizes discount w indow reserve infusions, does it make any difference whether the reserves are provided to solvent or insolvent banks or whether the period for which the reserves are provided is limited or extended? After all, if the reserves w ere not made available through the discount window, open market purchases would add an equivalent reserve contribution to the banking system. I believe that it does make a difference whether reserves are injected by open market purchases or by discount window lending, espe cially if insolvent banks are permitted to b o r row for extended periods. Discount w indow lending may not affect proposed monetary growth, but it has other effects that make it an undesirable source o f reserves. Open market operations are anonymous. The market allocates reserve injections or withdraw als among participants according to their rela tive size and current opportunities. Much greater discretion is exercised by the Federal Reserve in the allocation o f reserves through discounting, since the Fed knows the institu tions that request discount accommodation. The public learns about the magnitude o f both open market operations and discount window credit from the data the Federal Reserve publishes, but it does not learn the names o f the banks that received loans. The data made available to the House Banking Committee in 1991 revealed the names o f the institutions that had failed despite extended discount w indow loans, but not the names o f the few banks that had received such loans but had not failed.5 The secrecy may be good public policy, but it leaves open the question whether provision o f loans on a case-by-case basis assures equal treatment for all. This is an argument against discount w indow lending in general, not specifically to insolvent banks, an argument that has often been made in the past without reference to the specific problem o f insolvent banks (for exam ple, Friedman 1960, 38). dow did not fail. As footnote 2 on page 59 contends, the discount window is not essential for this use. SEPTEMBER/OCTOBER 1992 66 Since the 1970s, the Federal Reserve has ex tended long-term discount w indow assistance to depository institutions that by objective stan dards w ere likely to fail. It has done so in the belief that, in the absence o f such assistance, contagious effects w ould spread from the trou bled institutions to sound ones. The belief is particularly entrenched for large troubled inter mediaries, reflecting an apprehension that halt ing the operation o f such institutions would have dire unsettling effects on financial mar kets. Before 1985, the goal o f such discount w indow assistance was a restructuring o f the problem institution as a viable entity with both insured and uninsured deposits made whole. That was the situation in 1984 when Con tinental Illinois was rescued. The implication was that any other response w ould have brought on contagion. Yet if the bank had been closed before its net w orth turned negative, the institutional depositors, foreign bank depositors and creditors w ho ran on it might well have redeposited their withdrawals elsewhere or bought financial assets to replace the certificates o f deposit Continental issued and that they w ere no longer willing to buy. Even if some in terbank depositors that held as much as half their equity in uninsured deposits at Continental had obtained only a fraction o f their claims im mediately upon the closing, they would ulti mately have recovered the full nominal value of their claims after liquidation of the bank.6 The market would have known that the claimants on Continental w ere not in jeopardy. Even if closing Continental had led to runs on the fo r eign interbank depositors — ostensibly the rea son for keeping Continental in operation — the lenders o f last resort in the nations concerned could have provided adequate liquidity in their markets to tide the banks over if the Continen tal deposits w ere their only problem. Fear of contagion should not determine a regulator’s de cision to keep an insolvent bank open. It should lead the Fed to lend to the market to prevent the contagion. If fear o f contagion is a lesson the Federal Reserve has learned from the banking panics of 1930-33, it is the w rong lesson. Contagion then occurred in an environment in w hich the Fed 6This assumes that the bank's value would have been real ized in a forced sale of assets. H'he subsidized rate may have risen to take into account rates on market sources of funds but it is still a subsidy http://fraser.stlouisfed.org/ FEDERAL RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis permitted the money supply to decline drastical ly, rendering banks insolvent not because o f their ow n actions but simply because o f the col lapsing econom y. The right lesson is that con tagion need not arise if open market purchases are made adequate both to reassure the market and to prevent a collapse in the quantity o f money. Examples are the Fed’s provision o f li quidity to cushion the econom y from the effects o f the 1987 stock market crash and the collapse o f Drexel Burnham. Since 1985, prolonged discount w indow as sistance has generally terminated not with re structuring but with closure o f the insolvent banks. W hen banks are known to be insolvent, postponement o f recognition o f losses that have occurred might well have increased current losses. Uninsured depositors have m ore time to withdraw their funds. The insurance agency, which is to say the taxpayer, ultimately bears any added costs o f delayed closure. By lending to the banks in question, the Federal Reserve encourages this practice. Absent regulatory re straints or incentives to the contrary, the policy clearly encourages risk-taking and invites moral hazard problems. If a bank with the least desirable CAMEL rating can obtain subsidized discount window assistance, that institution ob tains a competitive advantage over solvent ones for as long as the Federal Reserve supports it.7 The Federal Reserve Banks decide whether they will extend a loan on the basis o f collateral that the would-be borrow ers offer. This deci sion is undoubtedly influenced by the condition o f the insurance agency: w hether it has the funds to pay o ff depositors and take over the failing bank’s assets w hen it cannot arrange a merger o f an insolvent bank with a solvent one or arrange removal o f existing management by placing an institution in receivership. The condi tion o f the insurance agency in turn is also af fected by a chartering agency’s forbearance or prompt action in declaring the insolvency of one o f its constituents. The Federal Reserve has cooperated by extending long-term support to an institution with a high probability o f failure until a resolution o f its problems was arrived at. Discount w indow lending to insolvent banks might cease if, under the terms o f the FDIC Imcompared to what deposit brokers and other non-Fed sources of funds would charge if the dying banks were left to fend for themselves in the market. 67 provement Act o f 1991, the Fed no longer ad vanced funds to keep critically undercapitalized institutions in operation, and the insurance agency took prom pt corrective action to appoint a receiver for those institutions.8 4. COULD REFORM REMEDY W H A T ’S W R O N G W IT H THE DIS COUNT W IN D O W ? W ould discontinuation o f Fed lending to insol vent banks establish the inviolability o f the dis count window? For many reasons that is not the case. Regardless o f one's attitude to discount win dow lending for seasonal and adjustment purposes—the private sector could accommodate those needs—many economists customarily as sign one indispensable function to the discount window, namely, as lender o f last resort: The Federal Reserve should use discounting in "ex ceptional circumstances" (in the w ords o f Regu lation A) to provide extended credit to solvent institutions with liquidity problems. However, the Federal Reserve does not need the window to serve as a lender o f last resort. The case against operation o f the discount win dow has rested on grounds that open market operations are sufficient for the execution of monetary policy in both ordinary and exception al circumstances, that individual banks that need and can justify special assistance can receive such assistance through the federal funds market, and that discounting invites dis cretionary subsidies to banks favored by the Fed (Goodfriend and King 1988, 216-53). There is still another reason that the discount w indow serves no useful purpose. A review of the use o f the discount mechanism by the Fed eral Reserve since its founding demonstrates a series o f misconceptions on its part about what it can achieve by affording banks the opportuni ty to acquire b orrow ed reserves. The miscon ceptions have varied over time, depending on the objectives the System pursued. Let me note some o f the misconceptions: (1) The Federal Reserve can determine whether borrow ed reserves serve “productive” rather than "speculative" use o f credit. (2) Banks borrow only for "need” and not for profit. (3) The absolute level o f free reserves or b o r rowings indicates whether banks choose to acquire or liquidate assets. (4) The spread between discount rates and mar ket rates does not affect bank borrowing. (5) The tradition against continuous borrow ing is a satisfactory substitute for a penalty dis count rate. (6) Borrowing by banks signals bank weakness. (7) Little bank borrow ing signifies money mar ket ease. (8) "Technical” adjustments may explain discount rate changes, but their announcement con veys useful information to financial markets. (9) Banks can be dissuaded from excessive use o f the discount w indow by Reserve Bank ap peals and exhortations instead o f discount rate increases. Since the 1970s, the Federal Reserve has acted on the belief that discount w indow assistance to banks with a high probability o f insolvency in the near term, especially large ones, will, by delaying closure, eliminate contagious effects on financial markets. In practice, the Federal Reserve’s discount w indow activities have created perverse incen tives, shifting risk from depositors to taxpayers. If a threat o f systemic bank failures did arise, the Fed should counter it by open market oper ations, rather than by assistance to individual institutions. However, if the Federal Reserve prevents serious declines in the money supply, it is highly unlikely that the failure o f an in dividual bank, how ever large, would trigger sys temic bank failures. Closing the discount w indow would free the Fed not only from likely future misconceptions but also from the recurrent pressures, to which it has been subject since the New Deal, to ac commodate nonbanks, not to mention pressures from the Treasury to fund government agencies o ff budget. 5. CONCLUDING COMMENTS I’ve surveyed the changes in Federal Reserve discount w indow practices since the System’s founding. Early on the System emphasized that 8For an alternative view of the possible consequences of a prompt corrective action strategy, see R. Alton Gilbert (1991). SEPTEMBER/OCTOBER 1992 68 borrow ing was supposed to be limited to short term reserve needs. By the 1980s, hundreds o f banks rated by regulators as having a high probability o f failure in the near term and which ultimately failed w ere receiving extended accommodation at the discount window. My reading o f this recent experience is that the change in discount w indow practices, by delaying closure o f failed institutions, increased the losses the FDIC and ultimately taxpayers bore.9 Recent legislation limits use o f the dis count w indow for long-term loans to troubled banks. M ore important, it also provides that a supervisory agency is to appoint a receiver for an institution that falls below a critical capital ratio, curtailing the regulator's discretion regarding when to intervene in the case o f an undercapitalized bank. These changes, even if implemented—it is not yet know n whether they will be—do not sanctify the continued operation o f the discount window. Individual banks that need assistance, if credit worthy, can obtain loans without subsidies from the federal funds market. The Fed can be an ef fective lender o f last resort if restricted to open market operations. In administering the discount window, the Fed has been prone to mistake the 9My conclusion from reading the 1991 Senate and House hearings on the long-delayed closings of the Bank of New England and Madison National Bank is that delay in creased resolution costs. The condition of the institutions deteriorated as the Fed continued to lend to them — they were already rated CAMEL 5 when the continuous loans began. With the value of liabilities greater than of earning assets in these institutions, the gap grew as interest in curred on liabilities exceeded earnings on assets. Delay, moreover, allowed outflows of uninsured deposits. Had the institutions been closed promptly, the earnings deficiency could have been offset at least somewhat by reducing the principal paid to uninsured depositors. Even so, at the closing, Bank of New England held $2.3 billion in uninsured deposits, of which $1.25 billion were brokered. At their peak, Fed advances amounted to $2.26 billion. The FDIC’s loss was $2.5 billion. As L. William Seidman testi fied, the FDIC decided to protect all depositors of the Bank of New England, at “ the additional cost [of] somewhere in the $200 to $300 million range up front” (see his statement in Senate Hearings on the failure of the Bank of New En gland, pp. 25-26). In the absence of Fed lending to a bankrupt institution, early closing would have prevented a flight of uninsured deposits. In effect, Fed lending merely replaced withdraw als of uninsured deposits. Before it was declared insolvent, the Fed had lent the Madison National Bank $125 million, which kept the bank open to permit withdrawals of $108 million in uninsured deposits. The FDIC’s final loss was $156 million. The FDIC suffers losses not only in cases in which it liq uidates failed banks but also in cases in which it arranges some form of purchase and assumption. When it liquidates a failed bank, the FDIC pays depositors the face value of their claims and suffers a reduction in its reserves meas http://fraser.stlouisfed.org/ FEDERAL RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis effects o f its actions. These mistakes have marred its execution o f monetary policy. Without a discount window, the Fed will avoid pressures to lend o ff budget to nonbanks and to government agencies, which should be funded through regular appropriations. Without the distraction o f monitoring collateral and deciding which bank applicants qualified for assistance, it can concentrate its energies on open market operations, the single instrument it needs to control the quantity o f money. Without a discount window, there will be no announce ment effects since the Fed will not have to set the discount rate. That should dispel the im pression that it controls market interest rates. A Federal Reserve System without the discount w indow would be a better functioning institution. REFERENCES Board of Governors of the Federal Reserve System. Annual Report (Washington, D.C., selected years). Economic Report of the President (Washington, D.C., 1971). Federal Deposit Insurance Corporation Improvement Act of 1991. Federal Reserve Bulletin (Washington, D.C., selected years and months). ured by the full difference between book and current values of the assets that support the deposits. When it ar ranges a purchase and assumption, it transfers the deposits to the buyer and, depending on the purchase price it has accepted, the FDIC is responsible for the re maining difference between book and current values of as sets that support the deposits transferred. In my opinion, delay in recognizing losses that already exist cannot be justified by the claim that the FDIC uses the time to improve the behavior of the bankrupt institu tion, even if it were true that supervisors would be suc cessful at this late date in the institution’s history in reforming it. Delay may, however, be helpful to the FDIC’s public image by postponing a publicly disclosed decline in the stated value of the reserves in the fund. That may be the real reason the FDIC welcomes delay. The FDIC has stated that the value of a bank to potential bidders goes down when it is already involved in resolving a failed bank case, as if that would validate delay. In fact, it is the difference between book and market value of a bank’s assets which a potential bidder learns that accounts for any decline in the bank’s value, not the fact that an ef fort at resolution is under way. What needs to be explained is why the Fed is the lender and not the FDIC, which has had the authority since 1982 to lend to open bankrupt banks and since 1989 to conser vators. Some buyers might argue that assets are worth more if FDIC can step in after it has fixed a price in a pur chase and assumption transaction and abrogates contracts the institution has with third parties, e.g., for rent, utilities, etc. That would be harder to do legally if the FDIC already controlled the bank or had an outstanding loan to it. That is not an argument that should encourage the Fed to lend instead of the FDIC. 69 Friedman, Milton. A Program for Monetary Stability. The Millar Lectures. Number Three (Fordham University Press, 1960). Garcia, Gillian and Elizabeth Plautz. The Federal Reserve: Lender of Last Resort (Ballinger, 1988). Gilbert, R. Alton. “ Supervision of Undercapitalized Banks: Is There a Case for Change?” this Review (May/June 1991), pp.16-30. Goodfriend, Marvin and Robert G. King. “ Financial Deregula tion, Monetary Policy and Central Banking,” in W.S. Haraf and R.M. Kushmeider (eds.) Restructuring Banking & Finan cial Services in America (American Enterprise Institute, 1988). _______ New York City Seasonal Financing Act of 1975. Formerly enacted (now omitted from Code) as 31 United States Code Sections 1501-1510. Public Law 94-143; 89 Stat. 797 (Dec. 9, 1975), pp. 797-99; Legislative History, p. 1509. _______ New York City Loan Guarantee Act of August 8, 1978. Formerly enacted as 31 United States Code Sections 1521-1531; see note under 31 U.S.C. Section 6702. Public Law No. 95-339. _______ Chrysler Corporation Loan Guarantee Act of 1979. 15 United States Code Sections 1861-1875. Public Law 96-185;85 Stat. 1324 (January 7, 1980), pp. 1324-37. Hackley, Howard. Lending Functions of the Federal Reserve Banks: A History (Board of Governors of the Federal Reserve System, 1973). U.S. House of Representatives, Committee on Banking, Finance and Urban Affairs, Staff. “An Analysis of Federal Reserve Discount Window Loans to Failed Institutions” (Processed, 1991). Kaufman, George. “ Lender of Last Resort: A Contemporary Perspective,” Journal of Financial Services Research, Vol. 5 (1991), pp. 95-110. U.S. House of Representatives, Committee on Banking, Finance and Urban Affairs. A Bill to Reform the Deposit In surance System ... and For Other Purposes (1991). Shull, Bernard. “ Report on Research Undertaken in Connec tion With a System Study,” in Reappraisal of the Federal Reserve Discount Mechanism, Vol. 1 (Board of Governors of the Federal Reserve System, 1971), pp. 27-76. _______ . Failure of Madison National Bank, Hearing, 102nd Cong. 1st sess., Serial No. 102-38 (Government Printing Office, May 31, 1991). Smead, Edward L. “ Operations of the Reserve Banks,” in Banking Studies (Board of Governors of the Federal Reserve System, 1941), pp. 249-70. Todd, Walker F. “ Lessons of the Past and Prospects for the Future in Lender of Last Resort Theory,” Federal Reserve Bank of Cleveland Working Paper 8805 (1988). U.S. Code. Emergency Loan Guarantee Act. 15 United States Code Sections 1841-1852. Public Law 92-70; 85 Stat. 178-182 (August 9, 1971), pp. 193-98; Legislative History, pp. 1270-78. _______ The Failure of the Bank of New England Corporation and Its Affiliate Banks, Hearing, 102nd Cong. 1st sess., Serial No. 102-49 (Government Printing Office, June 13, 1991). U.S. Senate, Committee on Banking, Housing, and Urban Af fairs. The Failure of the Bank of New England, Hearings, 102nd Cong. 1st sess., Serial No. 102-354 (Government Printing Office, January 9, May 6, and September 19, 1991). U.S. Treasury Department, Capital Markets Legislation Office. “A Bill to Reform the Federal Deposit Insurance System ... and for Other Purposes: (Processed, 1991). SEPTEMBER/OCTOBER 1992 Federal Reserve Bank of St. Louis Post Office Box 442 St. Louis, Missouri 63166 The Review is published six times per year b y the Research and Public Information Department o f the Federal R eserve Rank o f St. Louis. Single-copy subscriptions are available to the public f r e e o f charge. Mail requests f o r subscriptions, back issues, or address changes to: Research and Public Information Department, Federal Reserve Rank o f St. Louis, P.O. Rox 442, St. Louis, Missouri 63166. The views expressed are those o f the individual authors and do not necessarily reflect official positions o f the Federal Reserve Rank o f St. Louis or the Federal Reserve System. Articles herein may be reprinted provided the source is credited. Please provide the Rank's Research and Public Information Department with a cop y o f reprinted material.