The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
jMaoffifc ^ e w w D) J L L I N September/October 1995 Volume 80, Number 5 Ju •N WI N JAN 1 2 1996 mmm / FREJ RESEARCH LIBRAR i Federal Reserve Bank of Atlanta In This Issue: Is a Weak Dollar Inflationary? .Financial Crises and the Payments System: Lessons from the National Banking Era Policy Essay—Risk-Based Bank Capital: Issues and Solutions ftcQhOTÜiC jHewmw September/October 1995, Volume 80, Number 5 1 /bonomìe ^^eview Federal Reserve Bank of Atlanta President R o b e r t P. Forrestal S e n i o r Vice P r e s i d e n t a n d Director of R e s e a r c h Sheila L. T s c h i n k e l Research Department B. Frank King, Vice President and Associate Director of Research Mary Susan Rosenbaum, Vice President, Macropolicy Thomas J. Cunningham, Research Officer, Regional William Roberds, Research Officer, Macropolicy Larry D. Wall, Research Officer, Financial Public A f f a i r s Bobbie H. McCrackin. Vice President Joycelyn Trigg Woolfolk, Editor Lynn H. Foley, Managing Editor Carole L. Starkey, Graphics Ellen Arth, Circulation The Economic Review of the Federal Reserve Bank of Atlanta presents analysis of economic and financial topics relevant to Federal Reserve policy. In a format accessible to the nonspecialist, the publication reflects the work of the Research Department. It is edited, designed, produced, and distributed through the Public Affairs Department. Views expressed in the Economic Review are not necessarily those of this Bank or of the Federal Reserve System. Material may be reprinted or abstracted if the Review and author are credited. Please provide the Bank's Public Affairs Department with a copy of any publication containing reprinted material. Free subscriptions and limited additional copies are available from the Public Affairs Department, Federal Reserve Bank of Atlanta, 104 Marietta Street, N.W., Atlanta, Georgia 30303-2713 (404/521-8020). Change-of-address notices and subscription cancellations should be sent directly to the Public Affairs Department. Please include the current mailing label as well as any new information. ISSN 0732-1813 (Contents Federal Reserve Bank of Atlanta Economic Review September/October 1995, Volume 80, Number 5 Is a Weak Dollar Inflationary? Roberto Chang \ 5 financial Crises and the Payments System: Lessons From the National Banking Era William Roberds Since the early seventies, the U.S. dollar has been allowed to float freely, and its exchange rates have become extremely volatile and difficult to explain, let alone to predict. The dollar's erratic behavior has stimulated a lively debate in academic and policy circles over what the government's response should be. One of the major questions that must be answered before a response can be contemplated is whether dollar exchange rate changes influence U.S. inflation. This article examines the empirical relationship between dollar movements and inflation in the United States. Historical evidence suggests that a falling dollar causes inflation to increase but by a very small amount, and the author discusses why the inflationary effects of a weak dollar are so small. One possible answer is that when the dollar depreciates, sellers of traded goods may choose not to increase prices in response but to reduce their profit margins instead. Since the Great Depression, the Fed has historically intervened during potential financial crises to ensure that financial market participants were provided with the liquidity necessary to complete their transactions. In recent years, this part of the Fed's role in the payments system has come under increased scrutiny as advances in computer and communications technology have led to increases in the liquidity of many types of financial claims and to the creation of new financial markets and new forms of payment. Recent years have also seen passage of legislation designed to more precisely limit the scope and administration of the Fed's safety net as it applies to individual banks. This article examines the current debate over the Fed's payments safety net role in light of the historical experience of the National Banking Era (1864-1914). Though somewhat remote from modern experience, this period is highly relevant for the study of financial crises and payment system disruptions, offering a number of lessons for the effective provision of liquidity during crises. The most evident and least controversial lesson from the National Banking Era experience is that in crisis situations, timing is critical. Policy Essay—Risk-Based Bank Capital: Issues and Solutions Robert R. Bliss Market risk has become an integral consideration in bank business. Derivatives are increasingly used as a means of risk management, and bank involvement in derivatives trading represents a new, different, and very important line of business. Existing regulations for the determination of bank capital, based on the quality of assets held, are not appropriate for trading portfolio assets where exposure to market risk factors is of primary importance. This essay discusses three major proposals for dealing with market risk in determining banks' risk-based capital. The standard and internal models approaches are concerned with regulating the models used internally by banks for risk assessment and management. The third alternative, called the precommitment approach, emphasizes incentives and goals while leaving modeling issues entirely to banks. The author argues that, properly implemented, the precommitment approach is best suited to attaining regulatory goals. Is a Weak Dollar Inflationary? Roberto Chang J ^ ^ T ince the early seventies, the U.S. dollar has been allowed to float freely. What this means, in practiee, is that the U.S. government has abstained, for the most part, from intervening in the markets | W for foreign exchange, thus letting market forces determine the val^ « X lie of the dollar relative to other currencies. An important and well-known fact associated with floating is that exchange rates have become extremely volatile and difficult to explain, let alone to predict. 1 The fluctuations of the dollar-yen rate since the start of the current Mexico crisis provide an excellent example of this volatility. On December 20, 1994, the dollar was selling for 100 yen. Five months later, its value had fallen to 81 yen. And by October 1995 the dollar had bounced back to sell at above 100 yen. The author is a senior economist in the macropolicy section of the Atlanta Fed's research department. He thanks Mary Rosenbaum, Will Roherds, Tao Zha, Mick Devereux, and Harris Delias for helpful discussions. Reserve Bank of Atlanta DigitizedFederal for FRASER The dollar's erratic behavior has stimulated a lively debate in academic and policy circles over what the g o v e r n m e n t ' s response should be. S o m e people argue that the government should intervene systematically in foreign exchange markets to stabilize the dollar. These people tend to emphasize that exchange rate fluctuations are costly. In particular, they claim that a dollar's devaluation, by pushing up the dollar prices of internationally traded goods, is an important cause of inflation. 2 Others, while not necessarily approving of foreign exchange intervention, accept the premise that the dollar is a major force behind inflation. They argue, however, that controlling inflation requires a monetary policy that reacts to movements of the dollar, becoming more contractionary (other things being equal) when the dollar falls. 3 Economic Review 9 It is clear that, whatever the view on intervention, the q u e s t i o n of the e x t e n t to w h i c h e x c h a n g e rate movements are inflationary is crucial to the formulation of public policy. 4 There is a theoretical assumption that a r e d u c t i o n in the e x c h a n g e v a l u e of the dollar c a u s e s prices of i m p o r t e d g o o d s to increase, thus fostering inflation. But whether this reaction in fact occurs and whether dollar changes have a quantitatively important effect on inflation are, in the end, subtle empirical issues. This article will examine the empirical relationship between dollar movements and inflation in the United States. 5 Historical evidence suggests that a falling dollar c a u s e s inflation to i n c r e a s e but by a very small amount. Hence, while a weak dollar is indeed inflationary, the inflationary effects of the dollar, except in rare instances, are too small to warrant an activist policy response. 6 The article also discusses why the inflationary effects of a weak dollar are so small. One possible ans w e r is that w h e n the dollar d e p r e c i a t e s , sellers of traded goods may choose not to increase prices in response but to reduce their profit margins instead. 7 This possibility, called pricing to market, has been the subject of intense research recently and is r e v i e w e d in some detail. A First Look at Exchange Rates and Inflation Data The discussion begins by examining some facts related to the behavior of inflation and exchange rates, an exercise that will help dispose of some preconceptions and set the stage for m o r e detailed discussion later. Chart 1 tracks the U.S. price level and the value of the U.S. dollar f r o m January 1973, when the policy of allowing the dollar to float was initiated, to May 1995. The measure of the price level used is the Consumer Price Index (CPI). The most salient feature of the CPI over the period depicted is that it increased steadily. This reflects the well-known fact that inflation during this period was positive. The value of the dollar is measured by the Federal Reserve Bank of Atlanta's dollar exchange rate index, which is essentially a weighted average of the exchange rate of the U.S. dollar vis-àvis the currencies of its trading partners. The index is defined so that an increase indicates that the dollar is appreciating on a multilateral basis. As Chart 1 shows, the most striking aspect of the dollar's behavior during the 1973-95 period is the spike centered at 1985, which illustrates the e n o r m o u s appreciation of the dollar in the early eighties followed by the equally e n o r m o u s Chart 1 U.S. Price Level and the Dollar Value, 1973-95 January 1983=100 Source: Bureau of Labor Statistics and the Federal Reserve Bank of Atlanta. Economic Revieiv 2 September/October 1995 dollar depreciation after the so-called Plaza Agreement in September 1985. 8 Chart 1 contains the important lesson that the price level and the e x c h a n g e rate b e h a v e very differently and, in particular, that even huge dollar m o v e m e n t s seem to have little effect on inflation. If it is true that a fall in the dollar results in price increases for foreign goods and hence an increase in inflation, then the dram a t i c d o l l a r a p p r e c i a t i o n b e t w e e n 1979 and 1985 would have been accompanied by a noticeable fall in inflation. H o w e v e r , the CPT series s h o w s only o n e small change in the direction of inflation, after 1982. Conversely, a large and quick dollar devaluation such as that between 1985 and 1988 would be expected to make prices of foreign goods more expensive, thus fostering inflation; however, the rate of CPI inflation for those years shows essentially no response. Chart 1 focuses on the behavior of the levels of the dollar value and the CPI. However, there may be a relationship not between the levels of the CPI and the dollar but rather between their rates of change—that is, between the CPI inflation rate and the rates of dollar appreciation or devaluation. To examine this possibility, Chart 2 displays monthly inflation and appreciation or devaluation of the dollar, beginning in 1985. Inflation is measured by the percentage change in the CPI while the percentage change in the Atlanta Fed index gives exchange rate appreciation. As the chart shows, inflation rates move fairly smoothly and are positive in every month except two. During this period, c o n s u m e r prices increased no m o r e than two-thirds of 1 percent in any one month. In contrast, dollar fluctuations are erratic, there being no apparent correlation of CPI inflation and dollar changes. 9 Moreover, the exchange rate movements are comparatively large, often more than 2 or 3 percent. A clear conclusion f r o m the chart is that changes in exchange rates exhibit much more variability than do monthly inflation rates. Taken together, Charts 1 and 2 strongly suggest that there is no obvious systematic relation between CPI inflation and dollar movements. But whether inflation and the dollar are c o n n e c t e d can only be confirmed or refuted by systematic econometric evidence, which will now be considered. Do Exchange Rate Movements Help Predict Inflation? If dollar movements cause inflation, today's dollar movements should be useful in predicting tomorrow's Chart 2 Rates of Inflation and Dollar Appreciation/Devaluation, 1985-95 Percent C h a n g e .04 -.04 1985 1987 1989 1991 1993 1995 Source: Bureau of Labor Statistics and the Federal Reserve Bank of Atlanta. Federal Reserve Bank of Atlanta Economic Review 3 inflation. H e n c e it m a y be instructive to investigate whether the dollar and the CPI help forecast each other. The series for inflation and the value of the dollar displayed in Charts 1 and 2 can be used to conduct an econometric test of the hypothesis that dollar m o v e ments help predict inflation. The test is performed by first computing a regression of inflation on its past values and past values of dollar changes and then testing w h e t h e r the r e g r e s s i o n c o e f f i c i e n t s of p a s t d o l l a r changes are statistically different from zero. 1 0 If they are not, then one concludes that past dollar movements provide no information for predicting inflation that adds to the information already contained in past values of inflation itself. This kind of test, first proposed by Clive W. Granger (1969), is known as a Granger test. In contrast, Table 1 shows that the probability that past values of CPI inflation are not useful in forecasting exchange rates is more than 20 percent. This fairly large percentage is consistent with the well-known fact that exchange rates are highly unpredictable. 11 Table 1 displays the results of applying the Granger test to data underlying Charts 1 and 2. The table displays the probability that past values of the variable given by the column entry are not helpful in predicting the current values of the variable given by the row entry. For example, the upper left entry shows that the probability that past inflation is not useful in predicting inflation is essentially zero. The interpretation, therefore, is that past and current inflation rates do contain information that can help in forecasting future inflation. In addition, saying that dollar movements are useful in forecasting inflation does not imply that the dollar is causing inflation. Looking more closely at what a G r a n g e r test is supposed to do m a y help clarify the distinction. Essentially, a G r a n g e r test asks whether m o v e m e n t s in the dollar, say, precede m o v e m e n t s in inflation—but precedence does not imply causation. In particular, asserting that dollar movements cause inf l a t i o n m o v e m e n t s i m p l i e s that e l i m i n a t i n g d o l l a r movements would reduce the variability of inflation, a conclusion not warranted by a Granger test, which only establishes time precedence. To use C h r i s t o p h e r A. Sims's (1992) colorful analogy, the fact that the cock's crow precedes the sunrise does not imply that killing the bird would leave us in the dark. Do dollar movements help predict inflation, or vice versa? Table 1 indicates that the probability that past dollar changes are not helpful in predicting inflation is less than 1 percent. In other words, one can be highly confident that past dollar movements contain useful information about current and future inflation. Table 1 Granger Tests Applied to CPI Inflation and Dollar Changes Probability t h a t . . . d o e s not help p r e d i c t . . CPI Inflation Dollar C h a n g e s CPI Inflation Dollar Changes 0.000 0.207 0.008 0.000 Each entry is an estimate of the probability that the column variable does not help predict the row variable once past values of the row variable are taken into account. Source: Based on the data underlying Charts 1 and 2. Computed by the author. Economic Revieiv 4 In short, Table 1 indicates that exchange rate movements help in predicting inflation; there is a sense in which a weak dollar may be related to inflation. At the same time, inflation is not particularly helpful in forecasting e x c h a n g e rates. It is important, however, to note that the fact that exchange rates help predict inflation does not m e a n that they are good predictors; further, these results do not say anything about h o w much future inflation is associated with, say, a 10 percent dollar depreciation. In fact, there are scenarios in which the dollar has no effect on inflation, yet dollar movements help predict inflation. Imagine, for instance, that both the dollar and inflation respond to a third variable, say, the federal funds rate. Assume that an increase in the funds rate tends to be deflationary, albeit with some lags. Finally, assume that the dollar has no effect on inflation but that the dollar responds to expected future inflation. Under these circumstances, a reduction in the funds rate today would make the dollar devalue today and inflation increase in the future. Examining only the dollar and the CPI, one would find that dollar movements precede inflation, as the Granger test finds, but such precedence would not imply that dollar movements cause inflation. On the contrary, the dollar would be responding to anticipated inflation, and in this sense inflation would be causing the dollar movements. To summarize, while Granger tests tell us whether dollar movements and inflation help predict each other, they are silent on whether such predictive power is economically significant and on whether such predictive power implies causation. In order to shed light on September/October 1995 these issues, the results of other, more powerful tools will be examined. inflation Shocks and Exchange Rate Shocks As shown by the preceding discussion, a main difficulty in examining how dollar movements and inflation are related is that both exchange rates and prices may be r e s p o n d i n g to s o m e other variable, such as monetary policy. It may be, in fact, that such an omitted variable is the ultimate cause of most dollar and CPI fluctuations. The most satisfactory way to look at the relation between dollar changes and inflation may therefore be to build a complete empirical model of the U.S. economy and trace its implications for exchange rates and prices. Such a model would include a description of which variables are exogenous, that is, determined from outside the model, and which ones are endogenous, or determined by the model solution. Presumably, both the inflation rate and the v a l u e of the dollar w o u l d b e treated as endogenous variables whose movements are determined by e x o g e n o u s ones such as productivity shocks and unexpected changes in fiscal and monetary policy. Inflation and dollar m o v e m e n t s would in no s e n s e be t h o u g h t to c a u s e e a c h other; rather, both would be responding to exogenous forces. Some time ago, the relation between inflation and exchange rates was in fact analyzed by tracing the implic a t i o n s of such e m p i r i c a l m a c r o e c o n o m i c m o d e l s . However, this approach has fallen out of favor for two reasons. The first is that traditional empirical models were found to be based on "incredible" theoretical assumptions about economic behavior.' 2 For instance, these models would often assume expected inflation to be a fixed function of current and past inflation and not of other variables, such as the exchange rate. In contrast, modern m a c r o e c o n o m i c theory e m p h a s i z e s that rational agents base their inflation forecasts on all available information, not only current and past inflation. The second, and perhaps more important, reason to consider an alternative approach is that estimated models of exchange rate movements were found to perform poorly. In particular, in an influential 1983 article, Richard Meese and Kenneth Rogoff showed that the empirical models of exchange rate determination that prevailed at the time could not forecast better than a simple random walk model— that is, a model in which the best prediction of tomorrow's exchange rate is today's exchange rate. 13 Federal Reserve Bank of Atlanta An alternative, m o r e a g n o s t i c route to a n a l y s i s , most prominently advocated by Sims (1980), is k n o w n as a vector autoregression (VAR) approach. In a VAR model, a number of variables—for example, inflation and changes in the value of the dollar—are chosen for analysis. Each selected variable is then assumed to be " h i t " every period by an e x o g e n o u s " s h o c k , " which represents random, unexpected events that influence the v a r i a b l e in q u e s t i o n . A c c o r d i n g l y , an i n f l a t i o n shock is assumed to capture random forces hitting the market for aggregate output: examples are productivity improvements that increase the supply of output and put d o w n w a r d pressure on prices. Similarly, an exchange rate shock is intended to capture disturbances The question of the extent to which exchange rate movements are inflationary is crucial to the formulation of public policy. to foreign exchange markets such as a speculative sellout of the dollar. O n e objective of VAR analysis is to determine the relative importance of different shocks in d e t e r m i n i n g the b e h a v i o r of the variables u n d e r study. Sims's VAR approach avoids the criticism that it imposes incredible assumptions. In fact, no restrictions are imposed on the current response of inflation and the dollar to past disturbances, and the assumptions that are necessary for separating inflation shocks from exchange rate shocks are mild. This approach would be consistent with a model in which inflation expectations depend on all the variables included in the VAR. In addition, a VAR approach will typically allow the exchange rate to behave as a random walk if such behavior is consistent with the data. A main concern is that a VAR model only partially explains the ultimate sources of inflation and exchange rate m o v e m e n t s , a limitation resulting f r o m the assumption that the number of shocks hitting the model is equal to the number of variables under consideration. If there are more shocks than variables, a VAR analysis Economic Review 5 will erroneously attribute the effects of the additional shocks to the shocks associated with the variables in the VAR. This point will be discussed further after examining a VAR that includes monthly CPI inflation and monthly changes in the Atlanta Fed dollar index— the same variables underlying the discussion so far. In o r d e r f o r the i n f l a t i o n s h o c k s and the d o l l a r shocks to be well defined, some additional assumptions on the model are necessary. A natural assumption in interpreting shocks as exogenous is that current inflation shocks are uncorrelated with all dollar shocks and with all past and future inflation shocks. The VAR allows current changes in the dollar to be determined by current and past dollar shocks and current and past values of inflation. Current inflation is assumed to be determined by the current inflation shock, past values of inflation, and possibly past values of dollar changes; current inflation is not allowed to depend on current changes in the dollar, an assumption that is justifiable given that it is well known that the aggregate price level is a smooth (or "sticky") variable, at least in the short run. This assumption is crucial but mild in that it restricts the response of prices to the exchange rate within only the current month. T h e d o l l a r - i n f l a t i o n VAR w a s e s t i m a t e d f o r the 1973-95 period. 1 4 B e c a u s e displaying the estimated VAR coefficients provides little information, focus was instead placed on the implications of the estimated VAR f o r d e t e r m i n i n g the i n f l a t i o n a r y e f f e c t s of a changing dollar. O n e answer is found by plotting the expected effect of a typical inflation or exchange rate shock on the future paths of inflation and the dollar. These plots, known as impulse response functions, are shown in Charts 3 and 4. Chart 3 shows the expected response of CPI inflation to a typical current inflation shock and a typical dollar shock. The typical size of a shock is taken to be one standard deviation of the shock, calculated to be 1.22 percent for dollar shocks and 0.22 percent for inflation shocks. 1 5 As shown on the chart, a typical inflation shock today increases the current m o n t h ' s inflation by m o r e than 0.2 percent. This response is large relative to the average change in inflation during this period, which was about 0.3 percent. The inflation effect over the next two months decreases to about 0.08 percent and from then on declines smoothly toward zero. A typical inflation shock therefore seems to have an immediate, sizable effect on the first month's inflation and a slowly decaying effect on future inflation. In contrast, an average exchange rate appreciation has essentially no impact on future inflation, except perhaps for the sixth Chart 3 Responses of CPI Inflation to an Inflation Shock and a Dollar Shock Percentage Months Source: Author's calculations. Economic Revieiv 6 September/October 1995 month after the shock when the effect is to reduce inflation by 0.04 percent. 16 Hence the VAR shows that a typical dollar devaluation makes inflation increase but by a negligible amount. The information summarized above can be easily used to answer other questions about the effect of a fall in the dollar on inflation. For instance, one can deduce that a 10 percent dollar devaluation—a rare event—increases the annual inflation rate by 0.76 percent in the first year following the shock and by 0.96 percent, 0.45 percent, and 0.21 p e r c e n t in the second, third, and fourth year, respectively. The effect of a dollar shock on annual inflation is relatively small even at its peak and is only transitory. 17 Chart 4 displays the impulse response functions of the exchange rate. A dollar shock that occurs today is translated into a 1.25 percent change in the exchange rate in the current month, a smaller change next month, and no effect in subsequent months. An inflation shock today has essentially no impact on either current or future exchange rate changes. 1 8 An alternative and complementary way to look at the VAR results is to estimate h o w much of the inflation and exchange rate variability can be attributed to the different shocks. Such estimates are called decompositions of variance (see Table 2). The upper panel of the table shows how the uncertainty in the forecast of inflation at different horizons in the future can be attributed to either inflation or dollar shocks during the forecasting period. For example, the first entry in the table shows that 98.6 percent of the uncertainty in forecasting inflation six m o n t h s f r o m n o w is attributable to inflation shocks hitting the system for the next six months. An interesting aspect of the information in the upper panel of Table 2 is that dollar shocks explain between 5 and 6 percent of the variability of inflation at the one-, two-, three-, and four-year horizons. This result confirms the view that the effect of dollar fluctuations on inflation uncertainty seems to be fairly small although not zero. In contrast, the lower panel of the table confirms the view that CPI changes have essentially no effect on dollar uncertainty. CPI shocks never explain as much as 3 percent of exchange rate variability. Summarizing, both the impulse responses and the variance decompositions of a VAR including CPI inflation and dollar changes suggest that dollar shocks are inflationary, but mildly so. In addition, they tend to c o n f i r m the M e e s e and R o g o f f result that e x c h a n g e rate changes are essentially unpredictable. Now the question is whether the shocks isolated by the inflation-dollar VAR accurately portray the effect Chart 4 Responses of the Exchange Rate to an Inflation or Dollar Shock Percentage Months Source: Author's calculations. Federal Reserve Bank of Atlanta Economic Review 7 of exogenous shocks to exchange rate markets on inflation. A s previously discussed, the dollar m a y be responding to information about anticipated future inflation, which is caused by a third kind of shock, different from inflation or dollar shocks. To explore this possibility the set of variables included in the VAR needs to be expanded to include a variable that is sensitive to anticipated inflation. Such a variable should pick up the effect of a shock that is causing anticipated inflation—and, correspondingly, the dollar—to change. The price of gold is one variable thought to reflect expected inflation; therefore, a VAR that included CPI inflation, changes in the price of gold, and changes in the Atlanta Fed dollar index was estimated. As in the previous specification, it was assumed that prices are sticky in the sense that this m o n t h ' s inflation cannot r e s p o n d to current " g o l d s h o c k s " or dollar s h o c k s . Similarly, it was assumed that the price of gold cannot respond this month to current dollar shocks. 1 9 For brevity, the focus will be on the decomposition of variance for the inflation-gold-dollar VAR presented Table 2 V a r i a n c e D e c o m p o s i t i o n for the InflationDollar Change VAR Inflation Variance Forecast variance attributable to . . . at horizon . . . 6 months 1 year 2 years 3 years 4 years Inflation Shocks Dollar Shocks 98.6 94.8 94.2 94.1 94.1 1.4 5.2 5.8 5.9 5.9 Variance of Dollar Changes Forecast variance attributable to . . . Inflation Shocks at horizon . . . 6 months 1 year 2 years 3 years 4 years 2.1 2.5 2.8 2.9 2.9 Economic Revieiv 8 Dollar Shocks 97.9 97.5 97.2 97.1 97.1 in Table 3. (Plotting impulse response functions tells essentially the same story.) As shown by the table, gold shocks explain between 4.7 and 5.2 percent of the uncertainty in forecasting inflation at all horizons. It seems unlikely that these data reflect a cause-and-effect relationship, given that the price of gold is a negligible c o m p o n e n t of the CPI. It seems m o r e plausible that gold shocks are picking up the effect of some unobserved information that is useful in forecasting inflation but that was incorrectly attributed to either inflation shocks or dollar shocks in the CPI-dollar VAR. As Table 3 shows, the contribution of dollar shocks to inflation forecast variability is greatly reduced, to less than 3 percent at all horizons. Hence, when the effect of other forces affecting anticipated inflation is taken into account, the effect of a shock to the dollar's value on inflation is even smaller than concluded earlier. A legitimate question, though, is whether this procedure tends to underestimate the inflationary impact of dollar shocks in some other way. In particular, suppose that dollar shocks do significantly affect inflation but that monetary policy reacts to dollar shocks. If, say, in response to an unexpected dollar depreciation the Federal Reserve adopts a contractionary policy designed to prevent inflation, the full effects of a dollar shock may not show up in the VARs examined so far. A four-variable VAR, adding the federal funds rate to the previous three-variable VAR, was used to examine this possibility. Here, as in Ben Bernanke and Alan Blinder's (1992) work, the federal funds rate is intended to measure the stance of monetary policy. Inflation was again assumed to be sluggish in the sense that it did not respond within a given month to that month's shocks to gold, the funds rate, or the dollar. Similarly, gold prices were assumed not to respond within the month to shocks to the funds rate or to the dollar. The f u n d s rate w a s a s s u m e d not to r e s p o n d w i t h i n the month to shocks to the dollar. Hence, the dollar is considered the most flexible of the four variables. These a s s u m p t i o n s are mild, and the main results did not change much when different restrictions were imposed. Table 4 summarizes the variance decomposition for the inflation-gold-fed funds-dollar VAR. The fed funds rate explains a significant fraction (around 10 percent) of the variability of inflation at all horizons. In contrast, dollar shocks, although showing greater effect than in the inflation-gold-dollar VAR, explain less than 4 percent of inflation variance. The implication is that the previous results do not seem to reflect the effects of a monetary policy that reacts to the exchange rate. 20 In c o n c l u s i o n , the VAR e v i d e n c e p r e s e n t e d here supports the view that dollar movements have a statis- September/October 1995 tically nonzero but economically small effect on inflation. In particular, even a dollar fall as large as 10 percent raises annual inflation by less than 1 percent and has a short-lived effect. Given that the typical monthly dollar shock is not 10 percent but only a little m o r e than 1 percent, the findings imply that the behavior of the dollar has essentially no impact on inflation. Why Is the Dollar Not an Important Cause of Inflation? T h e m a c r o e c o n o m i c evidence m a y b e surprising. What is wrong with the straightforward presumption that w h e n the dollar depreciates prices of imported goods correspondingly increase, causing inflation? O n e important part of the answer is that prices of imported goods are not in fact as sensitive to dollar changes as commonly thought. See, for example, Table 5, which shows the change of the prices of several categories of imports between May 1994 and May 1995. During the one-year period covered the Atlanta Fed dollar index fell by 8.9 percent. Consequently, as the table shows, the dollar prices of foods and industrial supplies and materials increased somewhat more than the amount of dollar depreciation; the difference is presumably attributable to worldwide inflation. But the prices of capital goods, automotives, and consumer goods rose by much less than the dollar devalued. This evidence is typical. Analysts have noted that the dollar's unprecedented appreciation during the first half of the past decade did not produce a corresponding decrease in the prices of many U.S. imports. And many import prices increased only mildly in response to the subsequent fall of the dollar. 21 Table 3 V a r i a n c e D e c o m p o s i t i o n for the Inflation-Gold-Dollar V A R Inflation Forecast Variance Attributable to Inflation Shocks at horizon . . 6 months 1 year 2 years 3 years 4 years Gold Shocks 93.4 92.4 92.0 91.9 91.9 Dollar Shocks 5.2 4.7 5.1 5.2 5.2 1.4 2.9 2.9 2.9 2.9 Table 4 V a r i a n c e D e c o m p o s i t i o n for the Inflation-Gold-Fed Funds-Dollar V A R Inflation Forecast Variance Attributable to Shocks to . . . at horizon . . . 6 months 1 year 2 years 3 years 4 years Federal Reserve Bank of Atlanta Inflation Gold 81.0 78.1 76.9 76.8 76.9 8.7 8.1 8.7 8.8 8.8 Fed Funds 9.5 10.5 10.8 10.7 10.7 Economic Review Dollar 0.8 3.4 3.6 3.6 3.6 9 That import prices are rather insensitive to the value of the dollar is certainly consistent with the macroecon o m i c e v i d e n c e d i s c u s s e d in the p r e v i o u s sections. Why, though, are import prices so insensitive? A significant amount of research has been devoted to answering that question. T h e remainder of this article reviews selected relevant findings. The conventional belief that import prices increase when the dollar falls combines two assumptions. The first assumption is called the law of one price, which states that the dollar price of a traded good approximately equals its price abroad times the exchange rate. 22 For e x a m p l e , if a M e r c e d e s - B e n z car costs 9 0 , 0 0 0 marks in Frankfurt and the exchange rate is 1.5 marks Table 5 U . S . Import Price Indexes for Selected Categories (Percentage changes, May 1994-May Foods, Feeds, and Beverages Industrial Supplies and Materials, e x c l u d i n g Petroleum Capital G o o d s Automotive Vehicles, Parts, and Engines C o n s u m e r G o o d s , excluding Autos 1995) 10.3 13.5 3.0 3.8 2.1 Source: Bureau of Labor Statistics per dollar, the same M e r c e d e s - B e n z should cost app r o x i m a t e l y $ 6 0 , 0 0 0 ( 9 0 , 0 0 0 m a r k s divided by 1.5 marks/dollar) in Miami. The law of one price is assumed to be true because if Mercedes-Benz cars were selling for, say, $80,000 in Miami, much money could be m a d e by buying them in F r a n k f u r t and reselling them in Miami. In other words, the existence of spatial price arbitrage should prevent a tradable good f r o m selling at different prices in different locations. The second assumption behind the conventional belief is that the United States is a small country relative to world markets, and this position ensures that if the dollar devalues, prices of goods traded abroad do not change. According to this small-country assumption, if the exchange rate changed, the price of a MercedesBenz in Frankfurt would still be 90,000 marks. Under the law of one price and the small-country assumption, a fall in the dollar's value internationally must be reflected (passed through) by an exactly proportional increase in imported goods prices. In reality, 10 Economic Revieiv however, the pass-through is much less than proportional, at least for some goods. Consequently, one or both of these assumptions must be false. In fact, both are false. It is obvious that the United States is not a small country in world markets. If the dollar price of Mercedes-Benz cars were to increase, A m e r i c a n s would buy f e w e r of t h e m . T h i s reaction would probably represent a significant contraction in the w o r l d d e m a n d f o r M e r c e d e s - B e n z c a r s , w h i c h would in turn tend to reduce their prices everywhere, including Frankfurt and Miami. This force acts against the inflationary impact of a dollar devaluation, which as a consequence is reflected less than proportionally on import prices. What is not so obvious is that the law of one price also fails. Prices of traded goods are not, in reality, the same in different locations. Mercedes-Benz car prices, expressed in dollars, are not the same in Frankfurt as in M i a m i . T h e s e c a s u a l o b s e r v a t i o n s h a v e been c o n firmed by the academic literature. 23 Why is the law of one price not a "law"? One reason is that transport costs may place significant limits on spatial arbitrage. But this seems not to be the whole story, and a satisfactory answer is the subject of current research. In particular, a recent study by Charles Engel and John H. Rogers (1994) shows that, even after taking into account transport costs, the prices of the same good in two cities, one in the United States and the other in Canada, can display substantial variation. 24 This variation implies that failures of the law of one price can be attributed, in part, to a crossing-the-border effect whose exact nature is yet to be understood. T h e crucial fact seems to be that, in many cases, markets for the same physical good are geographically separated. Finns that sell in different countries are able to charge different prices for the same good in different locations. M e r c e d e s - B e n z , to continue the e x a m p l e , m a y c h o o s e to charge different prices for the s a m e m o d e l sold in F r a n k f u r t a n d in M i a m i . T h i s p h e nomenon has been called pricing to market. Pricing to market implies that a foreign firm selling in the United States may choose not to increase its dollar prices if the dollar loses value. In that case, the foreign firm must be willing to reduce its profit margins, that is, its profits per unit sold in the United States. Conversely, if the dollar appreciates, the foreign firm may choose not to lower dollar prices, thereby increasing its profit margins instead. Profit margins may therefore act as a buffer that dampens dollar price responses to a dollar shock. T h e thesis that d o l l a r c h a n g e s h a v e a relatively small impact on import prices because of the adjust- September/October 1995 ment of profit margins has been advanced by Catherine L. Mann (1986) and Paul Krugman (1987). Both authors presented empirical evidence showing the existence of pricing to market in response to dollar fluctuations. Their original findings have been strongly confirmed by more recent work. In particular, a recent study by Michael M. Knetter (1993) shows that pricing to market implies that the profit margins of Japanese, German, and British exports absorb more than one-third of the i m p a c t of e x c h a n g e rate c h a n g e s . K n e t t e r also shows that pricing to market is more pervasive in some industries than in others, an observation consistent with Table 5 and the evidence presented by Krugman (1987) and elsewhere. One may ask why a foreign firm would choose to reduce its profit margins instead of raising prices when the dollar falls. In spite of much research, no complete answer to this question has yet been found. It is not difficult to fomiulate theories that are consistent in principle with the observed degree of pricing to market; however, it is typically difficult to test these theories empirically, and the f e w attempts to do so have found the e v i d e n c e for f r o m o v e r w h e l m i n g . For instance, one may conjecture that a foreign firm's profits from selling in the U n i t e d States m a y d e p e n d on its m a r k e t share, giving the firm a reason to try to increase market share even if its ultimate objective is to maximize profits. In particular, if the dollar falls the firm has to decide whether to keep its dollar prices the same, which erodes its profit margin, or to increase dollar prices, which decreases market share. The firm's optimal response may depend on a variety of factors, including the degree to which the dollar's fall is expected to be transitory, in which case the firm is more likely to decide to keep dollar prices unchanged and accept a temporary decrease in profit margins. The above hypothesis was formulated and tested by Kenneth Froot and Paul Klemperer (1989). In spite of the intuitive plausibility of the hypothesis, Froot and Klemperer found only mild support, partly, perhaps, because it is impossible to observe directly whether the private sector expects a change in the dollar to be transitory or p e r m a n e n t . A s a c o n s e q u e n c e , F r o o t and Klemperer use survey data as proxies for market expectations. But the usefulness of survey data has long been in question. 25 An alternative would have been to use the time-series properties of exchange rates to decompose dollar changes into their permanent and transitory components. Had Froot and Klemperer done so, however, they would have found that dollar changes are mostly predicted to be permanent because, as discussed earlier, exchange rates are close to random walks. In Federal Reserve Bank of Atlanta terms of Froot and Klemperer's model, dollar fluctuations would therefore be mostly passed on to imports prices. While it is beyond the scope of this article to provide a complete survey of the literature on pricing to market, some conclusions can be drawn f r o m the literature reviewed. 2 6 First, there is strong evidence that the law of one price fails—that is, that the prices of similar goods are different in different countries. By implication, firms charge different prices for the same goods sold at different locations. Second, there is also strong evidence that firms react to changes in exchange rates partly by adjusting prices and partly by adjusting their profit margins. This evidence implies, in particular, that There is strong evidence that firms react to changes in exchange rates partly by adjusting prices and partly by adjusting their profit margins. the effects of a falling dollar have a less than proportional effect on imports prices. Finally, while many theories have been developed to explain pricing to market, their empirical relevance is still to be convincingly established. Conclusion M a c r o e c o n o m i c evidence f r o m vector autoregressions implies that dollar shocks have a n o n z e r o but small effect on the U.S. inflation rate. The conclusion for policy is, perhaps, that exchange rate m o v e m e n t s are not normally a source of concern from the viewpoint of keeping inflation low, although an extraordinarily large fall in the dollar m a y have a significant impact on inflation. One reason for the small impact of the dollar on inflation was found in the pricing decisions of foreign firms selling in the United States. T h e failure of the law of one price implies that foreign firms may charge Economic Review 11 integrated to a general macroeconomic model. Such a model would recognize that exchange rates depend on market characteristics and firms' decisions and not only the other way around. Whether theories that explain pricing to market can be extended to macroeconomic models that yield predictions for exchange rates is a challenging, but necessary, step for future research. different prices for the same good in the United States and abroad. Maximizing firm profits may then require that c h a n g e s in t h e d o l l a r b e a b s o r b e d in p a r t by changes in profit margins. This factor dampens the response of dollar prices to the exchange rate. The theory and evidence on pricing to market focus on particular industries and have not been successfully Notes 1. See Obstfeld (1995) for a review of the post-1973 experience with floating rates. 2. This inflationary effect was cited, for example, in April 1995 to justify the intervention of the governments of the United States, Germany, and Japan to halt a 9 percent fall in the dollar. "Choosing his words carefully, [Treasury Secretary Rubin] said that Washington's primary objective in trying to reverse the dollar's slide was to prevent the rising costs of imported goods—from German machine tools to Japanese cars—from causing higher inflation in the United States" ("Currency Markets; U.S. and Allies Try to Prop up the Dollar but Fail Again," The New York Times, April 6, 1995, Dl). 3. For example, it was reported in April (hat "Administration officials, who declined to be quoted by name out of concern that their remarks about dollar policy might roil the markets, speak of the possibility of a rate increase later this year. . . . (Given the weakened dollar, imports] would become more expensive, raising the danger of a higher inflation rate. . . . Raising inflation would undoubtedly prompt the Federal Reserve to raise rates, helping the dollar in the process" ("U.S. Is Resisting Any Pressures for a Quick Fix to Aid Dollar," The New York Times, April 22, 1995, 37). 4. It must be emphasized that the inflationary effect of devaluation is not the only valid reason for government intervention in foreign exchange markets. Disorderly conditions in financial markets, such as those prevailing during the October 1987 stock market crash, may justify intervention. The focus of this article, however, is on the relationship between dollar movements and inflation. 5. This relationship has been considered by, among others, Hooper and Lowrey (1979), Pigott and Reinhart (1985), Sachs (1985), Koch. Rosensweig, and Whitt (1988), and Hafer (1989). 6. This statement assumes, of course, that low inflation is a main objective of economic policy. 7. This is not the only possible answer. An alternative view may be that dollar movements cause changes in relative prices between tradable and nontradable goods but not in the general price level and hence in inflation. An implication would be that f o l l o w i n g a dollar devaluation the dollar prices of nontradable goods must fall and the dollar prices of tradable goods must increase. This relationship sounds implausible, but formal empirical work may conclude the op- Economic Revieiv 12 posite. In any case, this article focuses on the better-developed literature on profit-margin adjustments. 8. The Plaza Agreement was the result of an economic meeting of the governments of the G-5 countries; after the meeting they announced that they would jointly take measures to bring the dollar down. 9. One measure of this lack of relationship is that the contemporaneous correlation between the two series is only 0.09. Contemporaneous correlations do not, however, capture the relationship between one variable and lagged changes of another. Evidence of the possibility of such dynamic relationships will be presented later. 10. The rest of the article treats the (logs of the) exchange rate, the CPI, and the price of gold as stationary in differences and not cointegrated. This assumption was not rejected at very convincing significance levels by the standard unit root tests and cointegration tests described, for example, in Hamilton (1994). 11. For discussions of the unpredictability of exchange rates, see, for instance, Meese and Rogoff (1983). 12. See, in particular, Lucas and Sargent (1981) and S i m s (1980). 13. This finding seems to contradict the evidence in Table 1 because changes in a random walk cannot be predicted even from its own past. However, there is not necessarily a contradiction. The Atlanta Fed dollar index is a monthly average of daily figures. Time averaging tends to induce some serial correlation (and hence Granger-causality) even if exchange rates follow random walks. For a discussion of this issue, see Roberds (1988). 14. All the VARs presented in this article were estimated with six lags, a constant term, and seasonal dummies. 15. T o see why this definition is appropriate, recall that the standard deviation of a variable is the expected value of the squared deviation of the variable from its mean. The standard deviation of a shock is, therefore, a measure of the expected size of the shock. 16. Note that the response of inflation to a dollar shock is of the expected sign: a positive dollar shock implies a dollar appreciation. Because an appreciation should help reduce the dollar prices of traded goods, the expected inflation response is negative. 17. The magnitude of these estimates is comparable to that of those reviewed by Hooper and Lowrey (1979) and Pigott September/October 1995 and Reinhart (1985) and the ones obtained by Sachs (1985) and Koch, Rosenswcig, and Whitt (1988). 18. This conclusion is consistent with the view that the dollar follows a random walk, as argued by M e e s e and R o g o f f (1983). See note 10. 21. These facts are documented by Krugman (1987), Dornbusch (1987), Mann (1986), and Krugman and Baldwin (1987). 22. As the discussion attempts to make clear, the " l a w " of one price is not a law but an empirical conjecture that has been refuted by empirical work. 19. This assumption is more arbitrary than the assumption that the aggregate price level is sluggish; changing it makes very little difference to the results. 23. An early example is given by Isard (1977). 24. The study uses geographical distance as a proxy for transport costs. 20. Even the four-variable VAR seems unsatisfactory in at least one respect: there is a "price puzzle" in that the response of inflation to a federal f u n d s rate positive shock is initially positive, a response inconsistent with the view that the funds rate is a measure of exogenous monetary policy. The natural next step would be to attempt a structural VAR identification, as in Bernanke (1986) and Sims (1986). For a discussion of the price puzzle and its possible solution, see Sims (1992) and Gordon and Leeper (1994). 25. Froot and Klemperer also use interest rate differentials as proxies for exchange rate expectations. Because interest rate differentials are known to be poor predictors of future exchange rate changes, however, it is probable that they arc also imperfect proxies. 26. In addition to the studies already mentioned, interested readers should see Giovannini (1988), Fischer (1989), Baldwin (1988), Marston (1990), and Kasa (1992). References Baldwin, Richard. "Hysteresis in Import Prices: The Beachhead Effect." American Economic Review 78 (September 1988): 773-85. Hooper, Peter, and Barbara R. Lowrey. "Impact of the Dollar Depreciation on the U.S. Price Level." Board of Governors of the Federal Reserve System. Staff Study 103, April 1979. B e r n a n k e , Ben. "Alternative Explanations of the M o n e y Income Correlation." Carnegie Rochester Conference Series on Public Policy 25 (1986): 49-100. Isard, Peter. " H o w Far Can W e Push the 'Law of One P r i c e ' ? " American Economic Review 67 (December 1977): 942-48. Kasa, Kenneth. "Adjustment Costs and Pricing to Market: Theory and Evidence." Journal of International Economics 32 (February 1992): 1-30. Bernanke, Ben, and Alan Blinder. " T h e Federal Funds Rate and the Channels of Monetary Policy." American Economic Review 82 (September 1992): 901-21. Dornbusch, Rüdiger. "Exchange Rales and Prices." American Economic Review 11 (March 1987): 93-106. Engel, Charles, and John H. Rogers. " H o w W i d e Is the Border?" National Bureau of Economic Research, unpublished manuscript, August 1994. Fischer, Eric. " A M o d e l of E x c h a n g e Rate P a s s - T h r o u g h . " Journal of Internationa! Economics 26 (February 1989): 119-37. Froot, K e n n e t h , and Paul Klemperer. " E x c h a n g e Rate PassThrough When Market Share Matters." American Economic Review 79 (September 1989): 637-54. Knetter, Michael M. "International Comparisons of Pricing to Market B e h a v i o r . " American Economic Review 83 (June 1993): 473-86. Koch, Paul D., Jeffrey A. Rosensweig, and Joseph A. Whitt. " T h e Dynamic Relationship between the Dollar and U.S. Prices: An Intensive Empirical Investigation." Journal of International Money and Finance (June 1988): 181-204. Krugman, Paul. "Pricing to Market When the Exchange Rate C h a n g e s . " In Real-Financial Linkages Among Open Economies, edited by S.W. Arndt and J.D. Richardson, 4970. Cambridge, Mass.: MIT Press, 1987. G i o v a n n i n i , A l b e r t o . " E x c h a n g e R a t e s and T r a d e d G o o d s Prices." Journal of International Economics 24 (February Krugman, Paul R., and Richard E. Baldwin. " T h e Persistence of the U.S. Trade Deficit." Brookings Papers on Economic Activity 1 (1987): 1-43. 1988): 45-68. Gordon, David B., and Eric M. Leeper. "The Dynamic Impacts of Monetary Policy: An Exercise in Tentative Identification." Journal of Political Economy 102 (December 1994): 1228-47. Lucas, Robert E., and T h o m a s J. Sargent. " A f t e r Keynesian Macroeconomics." In Rational Expectations and Econometric Practice, edited by Robert E. Lucas and T h o m a s J. Sargent, 295-320. Minneapolis: University of Minnesota Press, 1981. Granger, Clive W . "Investigating Causal Relations by Econometric Models and Cross-Spectral Methods." Econometrica 37 (July 1969): 424-38. M a n n , C a t h e r i n e L. " P r i c e s , Profit M a r g i n s , and E x c h a n g e Rates." Federal Reserve Bulletin 12 (June 1986): 366-79. Hafer, R.W. "Does Dollar Depreciation Cause Inflation?" Federal R e s e r v e Bank of St. Louis Review 71 ( J u l y / A u g u s t Marston, Richard. "Pricing to Market in Japanese Manufacturing." Journal of International Economics 29 ( N o v e m b e r 1990): 217-36. 1989): 16-28. H a m i l t o n , J a m e s D. Time Series Models. Princeton University Press, 1994. Meese, Richard, and Kenneth R o g o f f . " E m p i r i c a l E x c h a n g e Rate Models of the Seventies: D o They Fit Out of S a m p l e ? " Journal of International Economics 14 (1983): 3-24. Federal Reserve Bank of Atlanta P r i n c e t o n , N.J.: Economic Review 13 Obstfeld, Maurice. "International Currency Experience: New Lessons and Lessons Relearned." Brooking Papers on Economic Activity, forthcoming 1995. Pigott, Charles, and Vincent Reinhart. " T h e Strong Dollar and U.S. Inflation." Federal Reserve Bank of New York Quarterly Review (Autumn 1985): 23-29. Roberds, William. " A Quarterly Bayesian VAR Model of the U.S. E c o n o m y . " Federal Reserve Bank of Atlanta Working Paper 88-2, May 1988. Sims, Christopher A. "Macroeconomics and Reality." Econometrica 48 (January 1980): 1-48. . "Are Forecasting Models Usable for Policy Analysis?" Federal Reserve Bank of Minneapolis Quarterly Review 10 (Winter 1986): 2-16. . "Interpreting the Macroeconomics Time Series Facts." European Economic Review 36 (June 1992): 975-1011. S a c h s , J e f f r e y D . " T h e D o l l a r and the Policy M i x : 1 9 8 5 . " Brooking Papers on Economic Activity 1 (1985): 117-85. Economic Revieiv 14 September/October 1995 financial Crises and The Payments System: Lessons from the National Banking Era William Roberds A . . . f a c t o r that has been present in m o s t financial disruptions o f the past fifteen years is the threat of dislocation in p a y m e n t , settlement, or clearing syst e m s . T h i s has b e e n r e a s o n a b l y well d o c u m e n t e d in the c a s e of the s t o c k m a r k e t c r a s h , b u t very difficult and potentially very serious p r o b l e m s with p a y m e n t a n d s e t t l e m e n t s y s t e m s h a v e a l s o b e e n e n c o u n t e r e d in o t h e r e p i s o d e s . . . . | P ] a y m e n t and settlement s y s t e m s are of special i m p o r t a n c e bec a u s e such s y s t e m s can b e the vehicle through w h i c h a localized p r o b l e m can v e r y q u i c k l y b e t r a n s m i t t e d to o t h e r s , t h e r e b y t a k i n g o n s y s t e m i c i m p l i cations. — E . Gerald Corrigan (1991) A critical function of the Federal Reserve System is that of mainX I taining the integrity of the group of institutions k n o w n as the / I payments system. The Fed fulfills this role through two main / I channels. First, on a day-to-day basis, the Fed provides payA • Mment and settlement services through its check-clearing and electronic-funds transfer operations. Second, the Fed has historically intervened during financial crises in order to ensure that financial market participants were provided with the liquidity necessary to complete their desired transactions. 1 The author is a research officer and senior economist in charge of basic research in the Atlanta Fed's research department. Federal Reserve Bank of Atlanta In recent years, this second part of the Fed's role in the payments system has come under increased scrutiny. The most important causal factor behind this reexamination of the Fed's "safety net" for the payments system has been the changes in financial institutions brought on by improvements in technology. Specifically, advances in computer and communications technology have led to increases in the liquidity of many types of financial claims and have led to the creation of new financial markets and new forms of payment (see, for Economic Review 15 example, Fred R. Bleakley 1994, John H. Boyd and Mark Gertler 1994b, Jane W. D'Arista 1994, Frederic S. Mishkin 1994, and Martin H. Wolfson 1994). Recent years have also seen passage of legislation designed to more precisely delimit the scope and administration of the Fed's safety net. 2 This essay will examine the current debate over the Fed's payments safety net role in light of the historical experience of the National Banking Era (1864-1914). Though somewhat remote from modern experience, this period is highly relevant for the study of financial crises and payment system disruptions for the following reasons. The first and most obvious reason is that a n u m b e r of m a j o r financial crises occurred during the National Banking Era. Most historians of this era mark at least five, and some, as many as six or seven. T h e second reason f o r considering this period is that there are certain noteworthy parallels between the financial history of the late nineteenth and early twentieth centuries and that of the late twentieth century. Specifically, the National Banking Era was also a time of tremendous growth and innovation in financial markets and institutions. Margaret G. Myers (1931) chronicled some of the important innovations during this period, which include the introduction of modern settlement systems for stock trades, the development of a call market or overnight funds market, and the development of the markets for trading commercial paper and foreign exchange. While these innovations were important in terms of improving the efficiency of financial markets, some of these were also potentially destabilizing in the sense that they facilitated highly leveraged bets on market outcomes. Important innovations in banking included the spread of checkable accounts as an alternative to circulating banknotes and the subsequent development of the clearinghouse system for clearing and settling checks. As has been the case in the late twentieth century, the f i n a n c i a l m a r k e t s of the N a t i o n a l B a n k i n g E r a were also globalized in the sense that s h o c k s were transmitted easily across national borders. Widespread adherence to the gold standard resulted in very tight international linkages a m o n g financial m a r k e t s and particularly a m o n g short-term interest rates. Market crashes in L o n d o n were quickly transmitted to N e w York (as occurred in 1890 and 1914) and vice versa (as occurred in 1907) (see Charles Goodhart 1995 and Oskar Morgenstern 1959). The third reason for considering the history of this period is that during the National Banking Era there was no public-sector a g e n c y charged with m a n a g e ment of the payments system in times of crisis, nor Economic 16 Review were there (with the exception of one financial panic) legally sanctioned private-sector remedies for disruptions in the payments system. In the absence of legal remedies for payment system disruptions, banks were able to improvise emergency p a y m e n t s systems that were, though strictly speaking illegal, nonetheless accepted as necessary both by the public and by governmental authorities. The application of these makeshift payments systems was in some instances effective in preventing severe disruptions to financial markets and to the e c o n o m y , but it was i n e f f e c t i v e in other instances. The variety of experiences during the different crises thus offers useful evidence on the type and scope of measures necessary to maintain an effective emergency payments system during a crisis. A fourth reason for considering the crises of the National Banking Era is that these have been very well studied. While there are some disagreements over matters of interpretation, the factual history of each crisis is well documented. A complete survey of the literature on these crises is beyond the scope of this essay; some of the m o s t i n f l u e n t i a l studies h a v e b e e n those of Alexander D. Noyes ([1909] 1980), O.M.W. Sprague (1910), Myers (1931), and Milton Friedman and Anna J. Schwartz (1963). More recently a number of studies, including Gary Gorton (1988), Gerald P. Dwyer, Jr., and R. Anton Gilbert (1989), Charles W. Calomiris and Gorton (1991), Mishkin (1991), and Jon R. Moen and Ellis W. Tallman (1992), have sought to reinterpret the National Banking Era experience in the context of contemporary banking theory. The Mechanics of the Payments System: Clearing and Settlement W h y is the integrity of the p a y m e n t s s y s t e m so critical to the functioning of the e c o n o m y ? T h e answer to this question lies in the mechanics of clearing and settlement. In developed economies, most goods and services are not directly purchased with cash (or technically, outside money). 3 Instead, most payments take the form of promises to deliver funds worth a certain amount. Ultimately these promises are converted to transferable claims on bank assets, such as checks. T h e s e c l a i m s are t a l l i e d and p r e s e n t e d to the appropriate b a n k for p a y m e n t , in a process k n o w n as clearing. C l a i m s cleared a m o n g b a n k s are, in most cases, only considered final payments once they have been settled, that is, offset by an opposite transfer of assets of equal value. May/June 1995 The set of rules governing clearing, settlement, and payment finality are thus of critical importance because they in effect determine an ordering of claims on the assets of any person or organization that purchases a good or service by transfer of the claims on bank assets. In the words of Corrigan (1990a, 131), once a payment becomes final, "the money in question is 'good m o n e y ' even if at the next instant the sending institution goes bust." In other words, once, a payment becomes final, the party receiving p a y m e n t (payee) has no need to stand in line with other creditors should the bank used by the party sending payment (payor) later go bankrupt. T r a d i t i o n a l rules c o n c e r n i n g finality of p a y m e n t in essence favor payees over other claimants on the payo r ' s bank's assets. 4 Such rules, necessarily somewhat arbitrary, are a prerequisite for mutually beneficial exchange in a free market economy. People are unlikely to provide goods and services without knowing that they will be paid "good funds" in exchange. In modern times, central banks form a crucial link in the process of clearing and settlement because in many instances settlement of claims among banks can be effected only via transfer of outside money in the f o r m of reserves held in accounts at a central bank."1 During the National Banking Era, the same function was served by the t r a n s f e r of reserve assets, which consisted of either specie or its equivalent (typically notes issued by the U.S. government payable in gold on demand). Today, the Fed generally provides the banking and payments systems with the amount of liquidity that is consistent with monetary policy goals. Under current Fed operating procedure, doing so means a c c o m m o dating short-term fluctuations in the d e m a n d for reserves in a manner consistent with a given target for the fed f u n d s rate. D e m a n d s for liquidity are accommodated along three main channels: the provision of r e s e r v e s v i a the f e d f u n d s m a r k e t ( s e e M a r v i n S. Goodfriend and William Whepley 1993 and Stephen A. Lumpkin 1993), the issuance of loans via the discount window (see David L. Mengle 1993a), and, on an intraday basis, the provision of " d a y l i g h t " credit over Fed wire, the F e d ' s large-value electronic f u n d s transfer system (see Mengle 1993b). Clearing and Settlement during Crises: The Policy Dilemma As a provider of funds to the banking and payments systems, the Fed faces certain informational problems Digitized forFederal FRASER Reserve Bank of Atlanta c o m m o n to all providers of credit. Like all lenders, the Fed must provide credit to borrowers—banks, in the case of the F e d — w i t h o u t perfect k n o w l e d g e of the borrowers' ability or desire to repay. In the absence of complete information on borrowers' creditworthiness, the recent literature on financial intermediation has shown that lending a r r a n g e m e n t s are subject to the problems of adverse selection (inability of the creditor to distinguish high-risk from low-risk borrowers) and moral hazard (incentives for borrowers to undertake actions that increase the likelihood of default). 6 Such problems are not insurmountable. T h e Fed, like all prudent creditors, uses a variety of time-honored techniques for minimizing its exposure to risks resulting from informational asymmetries. These include collateral requirements, close monitoring of borrowers' financial conditions (that is, supervision and regulation), and m i n i m u m net worth (capital) requirements on the part of the borrowers. 7 There is a key difference, however, between the informational problems faced by central banks and all other creditors. This difference is most apparent during times of financial crisis when i n f o r m a t i o n a l a s y m m e t r i e s b e t w e e n b o r r o w e r s and lenders are likely to b e e x a c e r b a t e d . 8 D u r i n g s u c h times, private-sector lenders can limit their risk exposure by charging markedly higher rates or by simply refusing to extend credit. Under contemporary institutional arrangements, however, it would be disastrous for a central bank to pursue such a strategy. Because outside money plays a critical and unique role in the modern-day payments system—that of settling transactions and providing payment finality—current institutional a r r a n g e m e n t s o f f e r no a l t e r n a t i v e to the provision of central bank money during crises. In such circumstances, central banks must be willing to provide adequate supplies of liquidity for transactions to be completed at a reasonable cost. This observation w a s the m a j o r t h e m e of Walter B a g e h o t ' s ([ 1873] 1991) treatise on central banking and is now regarded as standard central banking doctrine. Unfortunately, the obligation of central banks to provide liquidity during times of uncertainty does not diminish the potential adverse selection and moral hazard problems they face as lenders. To the contrary, this obligation can lead to the impression on the part of the private sector that, official policy statements to the contrary, liquidity will always be forthcoming at least to some banks during a crisis or even a situation with the potential to become a crisis. 9 The problem of h o w best to maintain the integrity of the payments system during times of crisis is thus seen to be a particularly delicate one. P o l i c y m a k e r s Economic Review 17 will always be confronted with the following dilemma: an overly restrictive approach to liquidity provision could easily hamstring the payments system and lead to a systemic crisis, while an overly generous policy can expose the central bank and ultimately the taxpayers to an excessive amount of credit risk. approaches to the Policy Dilemma Along one dimension of the policy dilemma, Fed policy in concert with that of the Federal Deposit Insurance Corporation (FDIC) and the Federal Savings and Loan Insurance Corporation must be seen as successful: no m a j o r disruption to the banking and payments systems has occurred in the United States since the 1930s. T h e r e s e e m s to be v i r t u a l l y u n a n i m o u s scholarly agreement that the lack of disruptions is due in large part to prompt and aggressive interventions on the part of the Fed. 1 0 There is considerably less agreement, however, over the issue of whether the Fed has been too generous with providing liquidity during potential "crisis" situations. The provision of credit through the discount window, in particular, has been the focus of criticism from the economics profession. 1 1 Critics of the F e d ' s discount window policy have seen a number of problems with discount window policy but have emphasized the problems resulting from extending discount window credit to b a n k s w i t h p o o r c r e d i t w o r t h i n e s s d u r i n g c r i s i s episodes. It should be noted that critics of Fed discount window policy have not suggested that the discount window has operated at a loss. Instead, the critics maintain that discount window loans have allowed troubled banks to remain open while uninsured depositors and other creditors r e m o v e their f u n d s . If the bank then fails, the FDIC is exposed to greater losses as a result. Recent legislation (FDICIA) seeks to limit this adverse selection problem by placing limits on the Fed's ability to make discount window loans to banks that do not meet m i n i m u m capital standards or are given the lowest supervisory rating. 1 2 S o m e a c a d e m i c critics, however, maintain that FDICIA does not go far enough. Their claim is that any benefits deriving from the provision of discount window credit are more than outweighed by costs associated with the potential adverse-selection problem. 1 3 In the view of these critics, liquidity demands of the private sector could be entirely accommodated by the Fed's open market operations. 1 4 In light of this debate, the National Banking Era experience is of interest because the contemporary ar- 18 Economic Review rangements for the provision of emergency liquidity were apparently very successful in dealing with the credit risk arising f r o m i n f o r m a t i o n a l a s y m m e t r i e s . During the National Banking Era there was no central bank and hence no provision of liquidity through open m a r k e t operations or t h r o u g h a discount w i n d o w . 1 5 During times of crisis, many banks faced with extraordinary d e m a n d s on liquidity did h a v e a c c e s s to an e m e r g e n c y s o u r c e of l i q u i d i t y — c l e a r i n g h o u s e loan certificates. In the National Banking Era, banks in most of the larger cities were organized into associations called "clearinghouses." In normal times, clearinghouses did exactly what the n a m e implies: they cleared checks drawn on their m e m b e r banks and facilitated settlement by either transfer of reserve assets or transfer of claims to reserve assets held at a central settlement account. 1 6 In times of crisis, however, it was c o m m o n for banks to completely suspend or at least curtail cash payments. To maintain the value of claims on their assets as means of payment, the clearinghouses would on occasion suspend the requirement that interbank settlement be effected in cash and would allow their members to settle in loan certificates. Loan certificates were simply debt issued against "good collateral"— that is, they were collateralized loans taken out by clearinghouse m e m b e r banks that were in turn held by other clearinghouse members. 1 7 As noted by Richard H. Timberlake, Jr. (1993, 204-7), most contemporary analysts recpgnized that the issue of loan certificates amounted to a private and most likely illegal issue of reserve money. However, at the time there was nothing approaching a consensus as to what arrangements might replace the issue of loan certificates. In many ways, the rules for issue of clearinghouse certificates resembled those that were imposed later for discount window loans under the Federal Reserve System. The most important rule was that the rate of discount on loan certificates was not indexed to a market rate but instead was set in advance at a rate typically below market rates (James G. Cannon 1910, 78). Like modern central banks, clearinghouses saw themselves as being able to "perfect a collateral interest"— that is, they could establish a senior claim on collateral posted by the banks issuing loan certificates. 1 8 In some crises, clearinghouses also took pains to keep the public from discovering which banks had taken out loan certificates. 1 9 It would be a mistake, however, to characterize the National Banking Era practice of loan certificate issue as equivalent to the operation of a private discount window. Settling payments in loan certificates, while May/June 1995 an accepted practice among clearinghouse banks, was still settlement in private debt, and debt of dubious legality at that. When issue of loan certificates was combined with a restriction of cash payments, the result was the emergence of a "currency p r e m i u m , " such that checks were accepted in payment but not valued on par with currency (see Friedman and Schwartz 1963, 110, 161-62). Access by banks to credit in the form of loan certificates was m u c h more restricted than that of present-day banks to the discount window. To be able to issue loan certificates, banks first had to belong to a clearinghouse. Clearinghouse membership requirements were typically quite stringent so that membership was far f r o m universal. M e m b e r s h i p w a s also localized in the sense that only banks in a given city were eligible for membership in that city's clearinghouse; contemporary branching restrictions made it impossible for most banks to operate in more than one locality. Finally, credit via the issue of loan certificates was simply unavailable during normal times. Loan certificates could only be issued once the governing body of the association (typically a representative committee) had met and agreed that a crisis situation existed that would require issue of the certificates. Once the decision to issue certificates had been made, all member banks were allowed to issue loan certificates upon posting eligible collateral (Cannon 1910, 77-78). The failure of a single institution was generally seen as insufficient cause for the issue of loan certificates. 71ien versus Now: A Comparison of the Two Approaches The National Banking Era approach to providing emergency liquidity provides a sharp contrast with the present-day safety net. (The salient features of the two s y s t e m s are s u m m a r i z e d in T a b l e I. 2 0 ) A s is w e l l known, the National Banking Era approach was unsuccessful in the sense that m a j o r financial disruptions were common during that period, several of which were followed by severe recessions. Yet the loan certificate system was successful in preventing some disruptions from developing into full-fledged panics. The loan certificates were also remarkably f r e e f r o m credit risk. Timberlake (1993, 210) notes that the m a x i m u m loss rate recorded on any single loan certificate issue was 1.8 percent, and for most issues there were no losses. O n e reason the loan system was so effective in curtailing credit risk is that it w a s a s y s t e m based o n clearing in private debt. That is, once a clearinghouse Federal Reserve Bank of Atlanta had decided to allow for the issue of loan certificates, m e m b e r banks had little choice but to accept other m e m b e r s ' debt (certificates) in settlement. This eventuality gave clearinghouse banks a strong incentive to monitor each others' creditworthiness and to expel insufficiently liquid or undercapitalized members. While settlement in private debt provided an almost perfect solution to problems of credit risk, this practice had its own problems. The first was the fact that settlem e n t in p r i v a t e debt l e s s e n e d the v a l u e of c h e c k s drawn on bank deposits as a medium of exchange. The loss of confidence in checks as m o n e y was explicitly manifested in some instances by the e m e r g e n c e of a currency premium. A second and more critical shortcoming of the loan certificate system was the flip side of the system's ability to contain credit risk. That is, the potential exposure of clearinghouse banks to each other's credit risk gave clearinghouses strong incentives to limit the breadth of access to loan certificate credit. In several instances, the provision of liquidity via loan certificates was either delayed too long or was insufficiently broad to stem the development of a maj o r panic. Goodhart (1988, 57-75) characterizes this problem as one of a potential conflict of interest between clearinghouse banks' desire on the one hand to maximize their own profits by avoiding the effects of panics and their desire on the other hand to avoid unnecessary aid to their commercial rivals. 21 In the case of the National Banking Era clearinghouses, the tendency of banks to withhold emergency access to liquidity was no doubt reinforced by the (strict) illegality of the loan certificates. In contrast to the National Banking Era system, in which private clearinghouses issued their own debt as emergency liquidity, our present-day system of emergency liquidity provision depends on the ready availability of outside money—that is, central-bank-issued debt—for purposes of settlement. This debt is generally provided through fully collateralized discount window credit. Such a system clearly a f f o r d s c o m p l e t e protection f r o m credit risk ~>n central bank m o n e y rec e i v e d in settlement. H o w e v e r , such a s y s t e m also necessarily transfers some portion of the credit risk associated with emergency liquidity to the F D I C and ultimately to the taxpayers. This system also necessarily removes at least a portion of the incentives for private participants in the payments system to monitor other participants' creditworthiness aggressively. Critics of the discount window (see note 10) have argued in effect that attempts to substitute the Fed's evaluation of creditworthiness for those of the private sector have not been entirely successful. Economic Review 19 Table 1 A S u m m a r y of the National Banking Era a n d Present Systems for the E m e r g e n c y Provision of Liquidity Feature National Banking Era System T y p e of liquidity provided Privately issued d e b t (clearingh o u s e loan certificates) Outside money (central b a n k liabilities) Value as a m e d i u m of e x c h a n g e O f t e n circulated (via check) at a d i s c o u n t ; only a c c e p t a b l e locally Universally a c c e p t e d at par Bearer of credit risk Clearinghouse banks Fed and Federal Deposit I n s u r a n c e C o r p o r a t i o n (in t h e c a s e of d i s c o u n t w i n d o w loans) Incentive to limit credit risk Strong d u e to self-interest of clearinghouse members Less so, though strengthened by provisions of FDICIA Breadth of c o v e r a g e Narrow W i d e , t h o u g h traditionally limited to depository institutions Availability of c o v e r a g e O n l y d u r i n g mutually a g r e e d u p o n "crises" Always Incentive to p r o v i d e timely coverage Mixed Strong Public c o n f i d e n c e in system Mixed Strong Efficacy in limiting systemic crises Mixed N o systemic crises since t h e 1930s Another possible advantage of the current system over the National Banking Era system is in breadth of coverage. Specifically, a central bank will have an incentive to intervene and provide liquidity in situations in which a clearinghouse or similar coalition of private banks would have incentives not to provide liquidity. Again, critics of the discount w i n d o w would argue that this option has been exercised too often and in cases where no real systemic crisis threatened. 20 Present System Economic Review Finally, an important advantage of the current system is the ability to credibly promise intervention even in cases for which little or no intervention actually takes place. A credible promise to intervene can lessen or even obviate the need to intervene, as was demonstrated during the October 1987 stock market crash. Various accounts of this episode (see note 10) agree that its successful resolution can be credited in large part to the Fed's public and credible c o m m i t m e n t to May/June 1995 p r o v i d e liquidity to b a n k s as n e e d e d . C o m m e r c i a l banks, freed from concerns about their own liquidity, were able to provide market participants with the extraordinary amounts of liquidity needed to settle financial m a r k e t t r a n s a c t i o n s d u r i n g t h i s e p i s o d e . An inevitable difficulty associated with such credibility is the familiar "too-big-to-fail" problem. To be effective, the Fed's ability to intervene in such situations cannot be taken as an implicit c o m m i t m e n t to intervene in any situation that could result in the failure of a single institution. bate concerning whether the behavior of depositors during the panics stemmed more from changing perceptions about the quality of their banks' assets or from a sort of herd instinct (see Calomiris and Gorton 1991, Mishkin 1991, and Moen and Tallman 1993). A Synopsis of Crises during the National Banking Era T h e Panic of 1873. T h e panic began with the September 8 failure of an insignificant warehousing firm because of bad loans to western railroad interests. By the following week, several large banks and investment houses had failed for essentially the same reason. In the face of a collapse in stock prices, the New York Stock Exchange closed on September 20 and did not reopen until ten days later. The closing of the stock market did much to spread the general perception of a panic. From the end of the Civil War until the outbreak of World War I, six crises required the issue of loan certificates by the N e w York Clearing House Association and other m a j o r clearinghouses. While details of the various episodes vary, each crisis followed the same basic pattern. The crisis would typically begin in N e w York with an event that investors would find unsettling, often the failure of one or more financial interm e d i a r i e s . T h e f a i l u r e s w o u l d lead to asset p r i c e d e c l i n e s and l i q u i d i t y p r e s s u r e s on the N e w York banks, pressures partially accommodated by the issue of clearinghouse loan certificates. In some cases (for example, 1884 and 1890) the issue of loan certificates w a s e n o u g h to forestall the d e v e l o p m e n t of a f u l l fledged money panic. O f t e n the issue of loan certificates was either too late or of insufficient quantity to reassure bank depositors, h o w e v e r . T h r o u g h o u t the country, d e p o s i t o r s w o u l d a t t e m p t to w i t h d r a w their d e p o s i t s in c a s h , causing banks outside of N e w York and the other major cities to attempt to liquidate their deposits with the city banks. The end result was often a nationwide suspension of cash p a y m e n t s by banks, as occurred in 1873, 1893, and 1907, followed by a severe macroeconomic contraction. Little d e f i n i t e is k n o w n c o n c e r n i n g the ultimate causes of the crises. However, certain factors associated with some or most of the incidents are often thought to have contributed to the likelihood of a crisis. A m o n g the factors that have been cited are seasonal (spring and fall) demands for liquidity resulting from agricultural activity and cyclical developments such as gold outflows, declining stock prices, rises in interest rates, and rises in the spreads on interest rates for low-quality versus high-quality investments. There is also some de Federal Reserve Bank of Atlanta The summary below does not attempt to attribute the occurrence of the panics to any particular cause. Instead it presents a brief description of the evolution of each crisis, the response by the N e w York Clearinghouse Association and other clearinghouses in terms of liquidity provision, the effectiveness of the liquidity provision in stemming the panic, and the macroeconomic consequences of the crisis. Information about the panics is further summarized in Table 2. The key factor behind the closing of the N e w York Stock Exchange was the primitive nature of contemporary systems for clearing and settling stock trades. At the time literally every stock trade had to be offset by a transfer of bank funds via certified check, a practice that resulted in liquidity demands far in excess of the funds that most trading firms had on deposit. The difference between the liquidity needs of the brokers and the funds they had on deposit was financed directly by the banks. Consequently, it was not unusual for banks to extend uncollateralized daylight credit to brokers in excess of ten times what they might have on deposit. 22 The d e m a n d s for such " c l e a r a n c e " f u n d s only g r e w with the onset of the panic, just as the banks became more reluctant to expose themselves to the risks associated with fluctuations in stock prices. Without access to sufficient funds to clear trades, the N e w York Stock Exchange had little choice but to cease operations. On September 24 the stock exchange proposed to the N e w York Clearinghouse Association that a special pool of liquidity be created by the clearinghouse for the sole purpose of settling stock trades; this proposal was summarily rejected (Sprague 1910, 39-40). Following the closing of the stock market, the N e w York Clearinghouse Association members continued to experience a liquidity drain. On September 20, 1873, they voted to issue up to $10 million in loan certificates. The issue of loan certificates was followed by a partial suspension of cash payments on September 24. Economic Review 21 Table 2 S u m m a r y of N a t i o n a l B a n k i n g E r a P a n i c s 3 Proximate Cause 1873 1884 1890 1893 1907 1914 Stock'Market Stock Market Stock Market Stock Markets and Run o n O u t b r e a k of Crash a n d Closure Crash Crash C o m m e r c i a l Failures NYC Trusts World War I Loan C e r t i f i c a t e Issue, N Y C ($ m i l l i o n s ) 26.6 24.9 16.6 41.5 101.0 NA Loan C e r t i f i c a t e Issue, NA NA NA 60-100 330 212b 46.9 70.7 92.5 99.0 181.0 NA 230/783 347/1,189 478/1,366 529/1,514 1,205/2,889 1,630/3,353b S u s p e n s i o n of P a y m e n t s ? Yes No No Yes Yes No Followed by Recession? Yes No No Yes, s e v e r e Yes, s e v e r e No N a t i o n w i d e ($ m i l l i o n s ) Bank Reserves, N Y C ($mil I ions) Bank Reserves/Monetary B a s e , N a t i o n w i d e ($ m i l l i o n s ) a All figures are approximate. The figures for loan certificate issues represent the total issue during the crisis. The amount of loan certificates outstanding at any given time was likely somewhat less than these numbers. The figures for reserves and the monetary base are for the dates that are closest to the crisis for which data are available. Estimates of reserves and money do not include loan certificates. h In addition to the issue of loan certificates, the 1914 panic saw the issue of about $400 million in Aldrich-Vreeland emergency currency. Some portion of this emergency currency is incorporated into the 1914 estimates of reserves and money. Sources: Friedman and Schwartz (1963), Sprague (1910, 1915), Timberlake (1993). Cannon (1910, 85) estimated the amount of the loan c e r t i f i c a t e s issued by the N e w York C l e a r i n g h o u s e Association m e m b e r s during the crisis at m o r e than $26 million. Some idea of the impact of this issue can be obtained by noting that aggregate reserves of New York banks before the onset of the crisis were on the order of $50 million (Timberlake 1993), and the entire stock of base m o n e y in the United States was about $800 million (Friedman and Schwartz 1963, 799). The 1873 issue of loan certificates was a success in the sense that bank failures during the crisis were limited to the failure of a few relatively unimportant institutions and in the sense that for most depositors checks drawn on bank deposits continued to serve as a medium of exchange. However, the provision of liquidity by means of loan certificates was, as explained above, insufficient to allow for operation of the stock market. Following the suspension of cash payments, the value of bank deposits was diminished by the emergence of a currency premium that peaked at roughly 5 percent by the end of September. The suspension of payments by the N e w York banks also wreaked havoc with banking and commerce across the country. Suspension in New York was followed by widespread suspension of payments by banks throughout the country. In some cases the shortage of cash led to layoffs because manufacturers were unable to meet their payroll obligations (Sprague 1910, 61-75). In some of the larger cities, clearinghouse associations followed the lead of the New York Clearinghouse Association and issued their own loan certificates (Cannon 1910, 86-90). . Given the paucity of data on this period, it is difficult to gauge the m a c r o e c o n o m i c costs of the 1873 panic. Friedman and Schwartz (1963, 40-41) suggest that aggregate output may not have contracted following the crisis. This conjecture is confirmed by Christina Romer (1989), who estimates that G N P grew by about 1 percent f r o m 1873 to 1874. Nathan S. B a l k e and Robert J. Gordon (1989), on the other hand, estimate that G N P fell by about 0.5 percent that year. In any case it is almost certain that for the year following the panic, output grew well below its intermediate-term trend value of about 4.5 to 6.5 percent, implying that the output cost of the panic was most likely substantial. The Panic of 1884. T h e 1873 panic had provided an almost perfect example of how not to manage a crisis. T h e provision of e m e r g e n c y liquidity had been made too late and in insufficient quantity to compensate for structural weaknesses in the banking system and in financial market settlement systems. These lessons were not lost on the N e w York Clearinghouse Association members. In early May 1884 the Federal Reserve Bank of Atlanta financial markets were rocked by the failure of a number of prominent firms due to fraud. On May 13 it was r e v e a l e d that the p r e s i d e n t of the S e c o n d N a t i o n a l Bank had absconded with more than $3 million of the bank's securities, and the following day a similar revelation of fraud started a run on the Metropolitan National Bank. On the afternoon of May 14 the N e w York Clearinghouse Association approved the issue of loan cert i f i c a t e s t o all m e m b e r s , i n c l u d i n g M e t r o p o l i t a n National. Ultimately about $25 million of loan certificates w e r e i s s u e d , as c o m p a r e d w i t h the c l e a r i n g h o u s e ' s r e s e r v e s of $ 7 0 million ( T i m b e r l a k e 1993, 204). Metropolitan National was able to r e o p e n its doors immediately, and the general feeling of panic subsided within a week, according to c o n t e m p o r a r y accounts. In contrast to the situation in 1873, there was no general suspension of cash payments, no currency premium, no closing of the stock exchange, and no nationwide disruption of banking and commerce; the macroeconomic effects of the panic also appear to have been minimal. 2 3 The Panic of 1890. Sprague (1910, 143-44) traces the proximate cause of the 1890 crisis (in the United States) to the N o v e m b e r 7 decision by the B a n k of England to raise its discount rate from 5 to 6 percent. S h a r p declines in the L o n d o n and N e w York stock markets led to the failure of Decker, Howell, and C o m pany on November 11, which also involved the Bank of North America. The N e w York Clearinghouse Association met on the afternoon of that day and decided on a prompt issue of loan certificates. Ultimately $17 million in certificates was issued, as c o m p a r e d with $92 million in reserve holdings (Timberlake 1993, 204). As was the case in 1884, the prompt issue of loan certificates seems to have contained the effects of the 1890 panic. On N o v e m b e r 15 the N e w York financial markets were again shocked by the news of the nearfailure of Baring Brothers and Company, and almost thirty brokerage houses failed as a result of subsequent declines in the stock market (Mishkin 1991, 86). Despite these d e v e l o p m e n t s , no general suspension of p a y m e n t s was declared and the e f f e c t s of the panic were largely confined to N e w York. 24 The Panic of 1893. The 1893 crisis was noteworthy as an event that defied the conventional wisdom of the times. The success of the loan certificate issues of 1884 and 1889 had strengthened the m a r k e t s ' confidence in this system of providing emergency liquidity. T h e liquidity d e m a n d s of the stock market had also been reduced by the introduction in 1892 of a modern system for settling stock trades (Myers 1931, 303-5). Economic Review 23 H o w e v e r , early 1893 saw the b e g i n n i n g of a sequence of events that would expose a serious flaw in the loan certificate system. A recession began in February that was accompanied by an unprecedented number of c o m m e r c i a l f a i l u r e s . T h e stock m a r k e t c r a s h e d on M a y 4 f o l l o w i n g the f a i l u r e of a m a r k e t f a v o r i t e . Nineteen national banks failed (or at least suspended payments) during M a y and June, leading to liquidity pressures on other banks throughout the country and causing t h e m to withdraw their deposits f r o m N e w York banks. 2 5 Recognizing the danger of a panic, the New York Clearinghouse Association authorized the issue of loan certificates on June 15, 1893. As had been the case in There are certain noteworthy parallels between the financial history of the late nineteenth and early twentieth centuries and that of the late twentieth century. 1884 and 1890, payments were not initially suspended. However, in the face of a continued, nationwide crisis of confidence, the loan certificates were ineffective in slowing the panic. The loan certificates had validity only within New York City and could not be used as a substitute for cash outside the city. Gold continued to flood out of New York during July, and by the beginning of August virtually all banks had suspended cash payments (Sprague 1910, 170-78). Suspension led to a currency premium that peaked at around 4 percent and persisted until September (Sprague 1910, 182). The suspension of cash payments only increased the demand for loan certificates. The size of the 1893 N e w York Clearinghouse Association issue was unprecedented at $ 4 1 . 5 million as c o m p a r e d with p r e p a n i c r e s e r v e s of $ 9 9 m i l l i o n ( T i m b e r l a k e 1993, 2 0 4 ) . Sprague (1910, 182) estimates that during the month of August 95 percent of interbank settlements in N e w York were effected by transfer of loan certificates. The 1893 panic also resulted in unprecedented issue of loan certificates outside of N e w York. Most is- 24 Economic Review sues were in amounts under $1 million, although the Boston and Philadelphia c l e a r i n g h o u s e s issued $11 million each. Perhaps m o r e importantly, many clearinghouse issues were m a d e in small d e n o m i n a t i o n s and were widely used as hand-to-hand currency. Quite a f e w of these small-denomination clearinghouse loan certificates were issued by b a n k s or by commercial firms in small communities that had no clearinghouse (Cannon 1910, 95-116). Sprague (1910, 197) places the m a x i m u m amount of loan certificates outstanding at any time during the panic at about $60 million. With the aggregate stock of narrow money about $1.5 billion at the time, even by this conservative estimate the issue of loan certificates still amounted to a significant addition to this aggregate. Despite these ingenious efforts at liquidity provision, the economy continued to contract during and after the panic of 1893. Roughly 5 percent of all U.S. banks failed during the crisis, along with 15,000 commercial firms. The total money stock fell by 6 percent, and o u t p u t c o n t i n u e d to c o n t r a c t until J u n e 1894 (Friedman and S c h w a r t z 1963, 109; Mishkin 1991, 87). The Miron-Romer (1990) index of industrial production shows a 14 percent drop from 1892 to 1893, followed by a weak rebound of only 2.7 percent from 1893 to 1894. Balke and Gordon (1989) have estimated that real ( i n f l a t i o n - a d j u s t e d ) G N P was flat f r o m 1892 to 1893 and contracted by about 3 percent in 1894. R o m e r (1989) has estimated that the economy contracted by about 1 percent in both 1893 and 1894. Not all of these effects can be attributed to the panic, as a recession was apparently well under way when the panic began. However, the disruptions caused by the p a n i c d o u b t l e s s c o n t r i b u t e d to the l e n g t h and severity of the downturn. The Panic of 1907. Surprisingly, the 1893 panic did not result in any major initiatives toward increasing the resiliency of the banking and payments system. The years between 1893 and 1907 also saw the rapid ascent of a new and particularly illiquid type of financial intermediary k n o w n as trusts. Trusts were essentially banks that offered deposit contracts that appealed to individuals and corporations who were more interested in higher returns over the long run and less intere s t e d in d r a w i n g on t h e s e d e p o s i t s as a m e a n s of p a y m e n t . 2 6 E x p l o i t i n g l o o p h o l e s in c o n t e m p o r a r y banking law, trusts were able to increase the yield on their assets relative to banks by holding less-liquid, higher-yielding portfolios that included virtually no reserves. By 1907, the total assets of trusts in N e w York State had grown to roughly three-quarters the size of national bank assets in the state ($1,364 billion versus May/June 1995 $1.8 billion) and were more than twice as large as the assets of state banks ($541 million) (Moen and Tallm a n 1992,612-16). In spite of their evident importance in the banking industry, in 1907 most N e w York City trusts were not members of the N e w York Clearinghouse Association. Instead, payments drawn on the trusts were settled indirectly through accounts held by the trusts at the various clearinghouse banks. The proximate cause of the 1907 panic was the October 21 announcement by one of the N e w York C l e a r i n g h o u s e A s s o c i a t i o n b a n k s that it would no longer clear checks for Knickerbocker Trust C o m p a n y . A run on Knickerbocker began the next day, followed by runs on other trust companies (Sprague 1910, 251-56). Because the trust companies were not in the N e w York Clearinghouse Association, they were not eligible for assistance via loan certificates. By the end of that week J.P. Morgan had formed a syndicate of large banks to extend credit to the beleaguered trusts, but by this time the general feeling of uncertainty had spread. Loan certificates were issued in N e w York b e g i n n i n g on O c t o b e r 26. N e w York banks suspended cash payments in early November, which led to a currency premium and widespread suspension of p a y m e n t s by banks in other parts of the country (Sprague 1910, 257-77). The subsequent issue of loan certificates was unprecedented in amount and scope. The N e w York City issue alone totaled $101 million versus $181 million in deposits at the start of the panic (Timberlake 1993, 2 0 4 ) . O u t s i d e of N e w York, loan c e r t i f i c a t e s were widely used both as a means of interbank settlement and as hand-to-hand currency. A. Piatt Andrew (1908) estimated that the aggregate issue of loan certificates in cities with population greater than 25,000 totaled $330 million. As the total stock of narrow money was approximately $2.9 billion, these issues amounted to a significant expansion of the money supply (Friedman and Schwartz 1963, 706). Like the 1893 panic, the panic of 1907 began shortly after the onset of a recession (June 1907), during which the U.S. industrial output again fell by roughly 14 percent as measured by the M i r o n - R o m e r (1990) index. Real G N P contracted by about 5.5 percent as estimated by Balke-Gordon (1989) and 4 percent according to the R o m e r ' s (1989) estimates. The traditional interpretation of this episode has placed almost all of the blame for the severity of the macroeconomic d o w n t u r n on the f i n a n c i a l panic. A l t h o u g h this int e r p r e t a t i o n has been c h a l l e n g e d by s o m e w r i t e r s (Mishkin 1991, 89-91, for example), the output costs of the panic must be reckoned as large by any accounting. Federal Reserve Bank of Atlanta The Panic of 1914. The immediate legislative response to the 1907 panic was the passage in 1908 of the Aldrich-Vreeland Act. T h i s act essentially legitimized the issue of clearinghouse loan certificates as emergency currency. There were, however, some important differences between the emergency currency issued under Aldrich-Vreeland and the earlier clearinghouse issues. Under Aldrich-Vreeland, national banks in the larger cities were allowed to group into associations for the purpose of issuing currency during emergencies. State banks and trusts were not eligible to issue emergency currency even if they were members of the local clearinghouse. The emergency currency was to be issued against a relatively narrow class of marketable assets held as collateral (U.S. securities and commercial paper) and was to be taxed at an annualized rate of 5 percent for the first month of its issue, increasing each month thereafter by one percent to a m a x i m u m of 10 percent. This schedule was later reduced to an initial rate of 3 percent, increasing by h a l f - p e r c e n t a g e points to 6 percent (Timberlake 1993, 209; Sprague 1915, 518-19). A distinctive advantage of the AldrichVreeland currency over the clearinghouse loan certific a t e s w a s that the f o r m e r w a s l e g a l l y c o n s i d e r e d currency and hence valid for settlements among banks that were not members of the same clearinghouse. For the same reason the legal status of the Aldrich-Vreeland currency all but eliminated the possibility of a currency premium. The Aldrich-Vreeland Act was superseded by the 1913 Federal Reserve Act. Before the Federal Reserve System could begin operations, however, world financial markets were shaken by the start of the first World War. Although the United States was not directly involved in the war at that time, the outbreak of hostilities had devastating effects on U.S. financial markets. The markets' supply of short-term credit from London was entirely cut off, just as the stock market crashed when European interests sought to liquidate their U.S. holdings. The resulting liquidity strains led to the closure of the N e w York Stock E x c h a n g e o n July 31, 1914. The stock exchange remained essentially closed for the next two months, and unrestricted trading did not resume until April 1 of the following year (Friedman and Schwartz 1963, 172). Other domestic financial markets and the markets for foreign exchange were also severely disrupted (Sprague 1915, 521-31). Following the closure of the stock market, runs began on banks as they had in earlier panics. The resulting liquidity demands were met by issues of both loan certificates (in aggregate about $212 million nationwide) Economic Review 25 and Aldrich-Vreeland emergency currency (in aggregate about $400 million). These issues expanded the n a r r o w m o n e y supply by about o n e - e i g h t h over its prepanic levels, a factor of increase somewhat larger than that experienced in 1907. 27 transactions technologies of the times, the delay of a day or two was critical. The post-1930s record of crisis interventions by the Federal Reserve (see note 10) and by foreign central banks suggests that this lesson has been almost universally accepted. 2 8 Despite the parallels with earlier panics, the panic of 1914 did not precede any m a j o r economic contractions, nor was there any general loss of c o n f i d e n c e in the banking system. Part of the reason for this lay in the peculiar wartime nature of the crisis. In the months immediately following the crisis, the U.S. banking system was bolstered by gold inflows resulting f r o m the demands for U.S. exports (Schwartz [1986J 1992a, 20). Aided by these same export demands, U.S. industrial output rose by 20 percent from 1914 to 1915 (Miron and Romer 1990, 336), while real G N P grew by about 5 percent (see R o m e r 1989 and B a l k e and G o r d o n 1989). Yet various accounts agree that the successful resolution of the crisis resulted in large part from the prompt issue of the Aldrich-Vreeland currency. The panic of 1893 illustrates, however, that timing is not everything. In the 1893 crisis, timely provision of loan certificates could not m a k e up for structural weaknesses present within the banking industry at the time. In particular, the inability of the banks to geog r a p h i c a l l y d i v e r s i f y their d e p o s i t base m e a n t that their clearinghouse loan certificates could serve only as local currency. City banks were therefore susceptible to s h o c k s that d r a i n e d r e s e r v e s a w a y f r o m the large cities towards the interior; country banks did not have access to the liquidity provided by the issue of loan certificates in the cities. B a n k s were also constrained in their ability to diversify their assets. Some striking evidence on the fragility of the U.S. banking industry during the National Banking Era can be obtained by a comparison of the U.S. and Canadian experiences during the panics. For example, Michael D. Bordo, Angela Redish, and Hugh Rockoff (forthcoming) observe that Canadian banks, which generally enjoyed a geographically diverse deposit base as well as ready access to emergency liquidity, were not run during National Banking Era panics. 71ie Lessons of the National Banking Era Crises T h e variety of crises experienced during the N a tional Banking Era offer a number of lessons for the effective provision of liquidity during crises. The system of providing liquidity via loan certificates was in many ways an imperfect solution to this problem, "a poor tenth best," in the words of Charles P. Kindleberger (1989). Yet this makeshift system was clearly on the margin of success and failure. It was as effective in certain instances (1884, 1890, and, in its legitimized form, 1914) as it was a failure in others (1873, 1893, 1907). In addition, the loan certificate system was highly successful in limiting the amount of credit risk borne by providers of emergency liquidity, that is, clearinghouse members. T h e m o s t e v i d e n t and least c o n t r o v e r s i a l lesson f r o m the National Banking Era experience is that in crisis situations, timing is critical. Emergency liquidity, if it is to be provided, should be provided before bank depositors and market participants begin to perceive institutions as illiquid. In the panics of 1873 and 1907, much of the damage resulted f r o m failure on the part of the clearinghouses and particularly the N e w York Clearinghouse Association to provide adequate liquidity early in the crisis. Relatively large issues of loan certificates a f t e r the fact could not correct the damage once a panic was under way. Even with the 26 Economic Review Today, the U.S. banking industry is obviously much more diversified over both sides of the balance sheet than was the case during the National Banking Era. Depositors are also insulated from the direct effects of s h o c k s to the b a n k i n g industry b y the g o v e r n m e n t "safety net." However, ongoing structural changes in the banking industry have led to increased risk-taking on the part of b a n k s , as d o c u m e n t e d by B o y d and Gertler (1993, 1994a). Boyd and Gertler conclude that recent policy initiatives such as the 1991 passage of F D I C I A and the Basle Accord of 1988, which seek to limit such risk-taking behavior, are justified because they improve economic efficiency and lessen the likelihood of taxpayer bailouts. The National Banking Era experience only reinforces this conclusion. Another lesson from the National Banking Era derives f r o m the experience of the 1873 panic, in which ample provision of interbank liquidity could not overcome structural weaknesses in securities market settlement systems and particularly an overreliance on bank credit. At the time the operation of the N e w York stock market depended entirely on the willingness of banks to extend indefinite amounts of uncollateralized credit to market participants. When this credit was not forthcoming, the immediate result was complete breakdown May/June 1995 of the market. The ultimate result of the 1873 and subsequent panics was the development of the clearinghouse system for clearing stock market trades, in which m a j o r market participants pool resources in order to guarantee the liquidity of the market. Today, methods for clearing and settlement in financial markets are vastly more sophisticated than was the case in 1873. Yet incidents such as the 1974 failures of B a n k h a u s H e r s t a t t a n d F r a n k l i n N a t i o n a l Bank, the 1980 silver crisis, the 1987 stock market crash, and the 1995 failure of Barings have all placed acute stresses on financial markets and their settlement systems. 2 9 S o m e financial m a r k e t s , particularly the markets for foreign exchange, continue to require extraordinary amounts of bank credit for their daily operations (see N e w York Foreign Exchange Committee 1994). As indicated in the epigraph introducing this article, this problem is a continuing concern to policymakers. The Federal Reserve and other central banks h a v e u n d e r t a k e n m e a s u r e s d e s i g n e d to e n c o u r a g e proper management of risks associated with the settlement of financial market transactions. Recent policy initiatives along these lines include the adoption of fees for daylight overdrafts on Fedwire and the adoption of m i n i m u m s t a n d a r d s f o r private large-value payment systems (Board of Governors 1993, 98-99; 1994, 109). The panic of 1907 sharply illustrates a problem with financial innovation, which is the tendency of the markets for financial intermediation to innovate around existing, costly arrangements for liquidity provision, through a process that Edward J. Kane (1977, 1981) has characterized as a "regulatory dialectic." In the case of the 1907 panic, the trusts arose as a new form of financial intermediary that was successful in circumventing the contemporary restrictions placed on bank portfolios, particularly reserve requirements. The trusts operated under the impression that they could "free ride" on the liquidity-providing services of the banks and the clearinghouses. However, many trusts did not have access to liquidity sufficient to cover depositor demands during the 1907 panic. Had the N e w York trusts had access to loan certificates in 1907, these institutions most likely would not have been run and a widespread panic would not have occurred. Only after the panic had revealed the illiquidity of the trusts was there any significant change in the institutional mechanisms for emergency liquidity provision. As was cent years v a t i o n in debatable the case in the early twentieth century, rehave seen a tremendous amount of innothe b a n k i n g and p a y m e n t s y s t e m s . It is w h e t h e r t h e s e i n n o v a t i o n s p o s e a direct Federal Reserve Bank of Atlanta challenge to b a n k s ' traditional role as liquidity providers. 3 0 However, certain financial innovations have placed new and potentially large (in times of uncertainty) demands on the liquidity of the banking system. For e x a m p l e , Boyd and Gertler (1993) survey three types of off-balance-sheet activities that have become increasingly prevalent in the U.S. banking system: loan c o m m i t m e n t s and standby lines of credit, loan sales and securitization, and provision of derivative instruments (such as swaps). As was the case with the trusts, these activities have enhanced the efficiency of f i n a n c i a l i n t e r m e d i a t i o n by a l l o w i n g f i r m s a n d The most evident and least controversial lesson from the National Banking Era experience is that in crisis situations, timing is critical. households access to cash flows that m o r e closely suit their individual circumstances. During normal times these activities do not place large demands for liquidity on the banking system, as was also the case with the trusts. Because each of these activities involves some liquidity guarantees on the part of banks, however, the potential exists that each could result in sudden, large demands for liquidity. T h e 1907 experience suggests that the issue of policy response in such situations will be an important one as new f o r m s of intermediation continue to evolve (Tallman and Moen 1995). Finally, a comparison of the 1914 panic with earlier disruptions illustrates the p a y o f f to a w e l l - p l a n n e d , credible, and legally sanctioned system for providing liquidity in crisis situations. The extraordinary liquidity demands associated with the 1914 panic were as large if not larger than that of any previous panic, yet there was no general loss of confidence in the banking or payments system. The knowledge that liquidity would be forthcoming in the f o r m of the legally sanctioned Aldrich-Vreeland emergency currency did much to alleviate the possibility of a sustained economic contraction. Nor are there any instances of losses recorded for either the legally sanctioned emergency currency or the Economic Review 27 e c o n o m i c history. Rather, the evolution Banking Era liquidity crises and proposed these crises offers a valuable set of policy whose relevance seems to appreciate with traditional clearinghouse loan certificates that were issued in response to the panic. In summary, the experience of the National Banking Era offers more than an colorful chapter in U.S. of National solutions to experiments time. Notes 1. For a summary of the F e d ' s interventions during financial crises and an i n t r o d u c t i o n to the r e l e v a n t literature see Mishkin (1991). 2. The Federal Deposit Insurance Corporation Improvement Act (FDICIA) of 1991 has been particularly important. See Wall (1993) for an introduction. 3. For the purposes of this article, outside money may be defined as either cash or reserves held by banks at a central bank. 4. The legal treatment of depositors' claims on bank deposits in this regard is s o m e w h a t different f r o m that of similar claims on most other types of firms. For example, chapter 7 of the U.S. bankruptcy code (which does not apply to banks and certain other financial intermediaries) allows for reversal or "avoidance" of transfers by liquidating firms. Transfers to creditors that are m a d e up to ninety days before a bankruptcy filing, or up to one year before bankruptcy in the case of "insiders," may be avoided. See Woelfel (1994, 125). 5. Blommestein and S u m m e r s (1994) discuss some of the potential advantages of interbank settlement in reserves. A more formal rationale for this practice is offered in Kahn and Roberds (1995). 6. See Gertler (1988) for a survey of this literature. 7. Limitations on the availability of one of the most direct forms of Fed credit—discount window loans—are discussed by Garcia and Plautz (1988) and more recently by Mengle (1993a) and Wall (1993). Historically, discount w i n d o w credit generally has been provided only to depository institutions, although some limited discount window lending to nondepository institutions took place in the 1930s under the Federal Reserve's emergency lending authority. 8. M a n k i w ( 1 9 8 6 ) , C o r r i g a n ( 1 9 9 1 ) , M i s h k i n ( 1 9 9 1 ) , and Summers (1991) all discuss the tendency of informational asymmetries to become magnified in such situations. One manifestation of this heightened uncertainty is discussed in Clair, Kolson, and Robinson (1995), who provide evidence that during such times banks have a strong tendency to shift interbank payment processing out of private systems and into Fed payments systems. Brimmer, Mishkin, Davis, Calomiris), the 1974 failure of Franklin National Bank (Schwartz, Garcia and Plautz, Wolfson) the 1980 silver crisis (Garcia and Plautz, B r i m m e r , W o l f s o n ) , the 1984 Continental Illinois e p i s o d e (Davis, Wolfson), the 1987 stock market crash (all except Schwartz and Calomiris), and the 1991 failure of the Bank of N e w England (Wolfson; see also Schwartz 1992b). 11. E x a m p l e s i n c l u d e F r i e d m a n ( 1 9 6 0 ) , G a r c i a and P l a u t z (1988), Goodfriend (1988), Kaufman (1991), and Schwartz (1992b). 12. F D I C I A - i m p o s e d limits on discount w i n d o w lending are s u m m a r i z e d in R o b e r d s ( 1 9 9 3 ) and d i s c u s s e d at s o m e length in Smith and Wall (1992) and Wall (1993). FDICIA also contains a clause that would allow these limits to be exceeded during systemic crises. In addition to the FDICIA reforms, the O m n i b u s Budget Reconciliation Act of 1993 contains "depositor preference" provisions designed to reduce F D I C exposure to credit risk. The efficacy of these provisions has been viewed by some observers as questionable; see for e x a m p l e K a u f m a n (1993) and T h o m s o n (1994). 13. See Kaufman (1991) and Schwartz (1992b). For contrasting views see Summers (1991) or Calomiris (1994). 14. T h e Fed p r o v i d e s liquidity via the m a r k e t f o r o v e r n i g h t f u n d s (the repo or R P market). However, only a narrow, homogeneous class of bank assets (U.S. Treasury securities) are available for RP. 15. O n occasion the T r e a s u r y would attempt to increase the supply of liquidity via open market purchases. T h e s e attempts at a "monetary" policy are seen by most historians as ineffective during times of crisis; see, for example, Timberlake (1993). Myers (1931, 286-87) also records several instances in which syndicates of large commercial banks formed pools of liquidity that were made directly available to the financial markets via the call loan market. 9. Mengle (1990) discusses this problem in some detail. In the recent literature this is usually described as the problem of "too-big-to-fail." 16. See Cannon (1910, 36-57) for a description of contemporary settlement practices. 17. See Cannon (1910, 75-76), Myers (1931, 98-99), and Timberlake (1993, 199-200). Loan certificates taken out by any member bank were required to be accepted by other members in settlement upon penalty of expulsion f r o m the clearinghouse. 10. See, f o r e x a m p l e , S c h w a r t z ([1986] 1992a), Garcia and Plautz (1988), Brimmer (1989), Bernanke (1990), Mishkin (1991), Davis (1992), Calomiris (1994) and Wolfson (1994). A m o n g the more recent events requiring Fed intervention have been the 1970 Penn Central bankruptcy (see Schwartz, 18. Discussions of discount window lending—see Goodfriend (1990, 263-65), for example—emphasize the legal advantage enjoyed by government agencies in being able to establish senior claims against bank assets, although Goodfriend notes that this advantage may not exist in the case of repur- 28 Economic Review May/June 1995 chase agreements. It is unclear why National Banking Era clearinghouses felt that they could enforce seniority of their claims. Part of the answer probably lies in the fact that, in the case of the loan certificates, posted collateral was rarely liquidated. Cannon (1910, 237-39) does describe one Supreme Court case in which the seniority of clearinghouse claims was successfully defended. 19. See Gorton and Mullineaux (1987, 465). The experience of some New York banks during the 1893 panic suggests that this information did not always remain secret. See Sprague (1910, 181). 20. This- comparison is not meant to suggest that the two systems summarized in Table 1 are the only possible means of providing emergency liquidity to the banking and payment systems. 21. Goodhart cites the ability to overcome this conflict of interest as a natural advantage of central banks, a view that is disputed by Dowd (1994). 22. See Sprague (1910, 38-39) and Myers (1931, 282-85). Myers n o t e s that this practice; k n o w n as o v e r c e r t i f i c a t i o n , flourished in spite of its illegality. 23. On the events of 1884 see Sprague (1910, 108-23), Cannon (1910, 90-91), Noyes ([1909] 1980, 99-100), Myers (1931, 412-13), and Mishkin (1991, 83-84). 24. There were issues of loan certificates by the clearinghouses in Boston and Philadelphia. See Cannon (1910, 92-94). 25. See Sprague (1910, 168-171). The drain on bank reserves may have been exacerbated by political uncertainty over whether the United States would maintain the gold standard. See Friedman and Schwartz (1963, 108-9). 26. Deposits at trusts w e r e still essentially b a n k - l i k e in the sense that they were payable on demand at par. 27. See Friedman and Schwartz (1963, 172). Sprague (1915, 518) noted that almost all of the Aldrich-Vreeland currency issued d u r i n g the p a n i c w a s issued against g o v e r n m e n t bonds; in this sense the 1914 emergency currency issues m o r e closely resembled m o d e r n open market operations than discount window loans. 28. On the actions of foreign central banks during crises see Corrigan (1990b) or Davis (1992). 29. On the Herstatt failure see Davis (1992, 154-56) and Corrigan (1990b), on Barings see Falloon (1995); on the other incidents see note 10. 30. For t w o c o n t r a s t i n g p o i n t s of v i e w on this s u b j e c t see D ' A r i s t a and S c h l e s i n g e r (1993) and B o y d and G e r t l e r (1994b). References Andrew, A. Piatt. "Substitutes for Cash in the Panic of 1907." Quarterly Journal of Economics 22 (August 1908). Bagehot, Walter. Lombard Street: A Description of the Money Market. N e w York: Scribner, A r m s t r o n g , and C o m p a n y , 1873. Reprint. Philadelphia: Orion Editions, 1991. Balke, Nathan S., and Robert J. Gordon. " T h e Estimation of Prewar Gross National Product: Methodology and New Evidence." Journal of Political Economy 97 (February 1989): 38-92. Bernanke, Ben S. "Clearing and Settlement during the Crash." Review of Financial Studies 3 (1990): 133-51. Bleakley, Fred R. "Fast Money: Electronic Payments Now Supplant Checks at M o r e Large Firms." Wall Street Journal, April 13, 1994. Blommestein, Hans J., and Bruce J. Summers. "Banking and the Payment System." In The Payment System: Design, Management, and Supervision, edited by Bruce J. Summers, 15-28. Washington, D.C.: International Monetary Fund, 1994. Board of Governors of the Federal Reserve System. Annual Report. Washington, D.C., 1993, 1994. Bordo, Michael D„ Angela Redish, and Hugh Rockoff. " T w o Long Roads to Stability: Canadian and American Banking, 1870-1925." Financial History Review (forthcoming). Boyd, John H., and Mark Gertler. "U.S. Commercial Banking: Trends, Cycles, and Policy." In NBER Macroeconomics Annual, 1993, edited by Olivier J. Blanchard and Stanley Fischer, 319-68. Cambridge, Mass.: M I T Press, 1993. Federal Reserve Bank of Atlanta . " T h e Role of Large Banks in the Recent U.S. Banking Crisis." Federal Reserve Bank of Minneapolis Quarterly Review 18 (Winter 1994a): 2-21. . " A r e Banks Dead? Or Are the Reports Greatly Exaggerated?" Federal Reserve Bank of Minneapolis Quarterly Review 18 (Summer 1994b): 2-23. Brimmer, Andrew F. "Central Banking and Systemic Risks in C a p i t a l M a r k e t s . " Journal of Economic Perspectives 3 (Spring 1989): 3-16. Calomiris, Charles W . "Is the Discount Window Necessary? A Penn Central P e r s p e c t i v e . " Federal R e s e r v e Bank of St. Louis Review 76 (May/June 1994): 31-55. C a l o m i r i s , C h a r l e s W . , and G a r y G o r t o n . " T h e O r i g i n s of Banking Panics: Models, Facts, and Bank Regulation." In Financial Markets and Financial Crises, edited by R. Glenn Hubbard, 109-74. Chicago: University of Chicago, 1991. C a n n o n , J a m e s G . Clearing Houses. Report by the National Monetary Commission to the U.S. Senate. 61st Cong., 2d session. Doc. 491. Washington, D.C.: Government Printing Office, 1910. Clair, Robert T „ Joanna O. Kolson, and Kenneth J. Robinson. " T h e Texas Banking Crisis and the Payments System." Federal Reserve Bank of Dallas Economic Review (First Quarter 1995): 13-21. Corrigan, E. Gerald. "Perspectives on Payments System Risk Reduction." In The U.S. Payments System: Efficiency, Risk, and the Role of the Federal Reserve, edited by David B. Economic Review 29 Humphrey, 129-40. Boston: Kluwer Academic Publishers, 1990a. . "Response of Four Central Banks to Institutions in Distress." International Currency Review (1990b): 43-59. . "The Risk of a Financial Crisis." In The Risk of Economic Crisis, edited by Martin Feldstein, 44-53. Chicago: University of Chicago Press, 1991. D'Arista, Jane W . The Evolution of U.S. Finance. Vol. 2, Restructuring Institutions and Markets. Armonk, N.Y.: M.E. Sharpe, 1994. D'Arista, Jane W., and Tom Schlesinger. " T h e Parallel Banking S y s t e m . " E c o n o m i c Policy Institute B r i e f i n g Paper, March 1993. Davis, E. Philip. Debt, Financial Fragility, and Systemic Risk. Oxford: Oxford University Press, 1992. D o w d , Kevin. " C o m p e t i t i v e B a n k i n g , B a n k e r s ' C l u b s , and Bank Regulation." Journal of Money, Credit, and Banking 26 (May 1994): 289-308. Dwyer, Gerald P., Jr., and R. Anton Gilbert. "Bank Runs and Private Remedies." Federal Reserve Bank of St. Louis Review 71 (May/June 1989): 43-61. Falloon, William. " W h o ' s Missing f r o m the Picture?" Risk 8 (April 1995): 19-22. F r i e d m a n , Milton. A Program for Monetary Stability. New York: Fordham University Press, 1960. Friedman, Milton, and Anna J. Schwartz. A Monetary History of the United States, 1867-1960. Princeton, N.J.: N B E R , 1963. Garcia, Gillian, and Elizabeth Plautz. The Federal Reserve: Lender of Last Resort. Cambridge, Mass.: Ballinger, 1988. Gertler, Mark. "Financial Structure and Aggregate Economic Activity: A n O v e r v i e w . " Journal of Money, Credit, and Banking 20 (August 1988, pt. 2): 558-88. Kane, Edward J. "Good Intentions and Unintended Evil: The Case against Selective Credit Allocation." Journal of Money, Credit, and Banking 9 (February 1977, pt. 1): 55-69. . "Accelerating Inflation, Technological Innovation, and the Decreasing Effectiveness of Banking Regulation." Journal of Finance 36 (May 1981): 355-67. K a u f m a n , George G. "Lender of Last Resort: A Contemporary Perspective." Journal of Financial Services Research 5 (October 1991): 95-110. . " T h e New Depositor Preference Act: T i m e Inconsistency in Action." Loyola University of Chicago, unpublished paper, October 1993. Kindlebcrger, Charles P. Manias, Panics, and Crashes: A History of Financial Crises. Rev. ed. N e w York: Basic Books, 1989. L u m p k i n , Stephen A. " R e p u r c h a s e and Reverse R e p u r c h a s e Agreements." In Instruments of the Money Market, edited by Timothy Q. Cook and Robert K. LaRoche, 59-74. 7th ed. Richmond, Va.: Federal Reserve Bank of Richmond, 1993. Mankiw, N. Gregory. " T h e Allocation of Credit and Financial C o l l a p s e . " Quarterly Journal of Economics 101 (August 1986): 455-70. Mengle, David L. "Legal and Regulatory Reform in Electronic Payments: An Evaluation of Payment Finality R u l e s . " In The U.S. Payments System: Efficiency, Risk, and the Role of the Federal Reserve, edited by David B. Humphrey, 145-80. Boston: Kluwer Academic Publishers, 1990. . "The Discount W i n d o w . " In Instruments of the Money Market, e d i t e d by T i m o t h y Q. C o o k and R o b e r t K. LaRoche, 22-33. 7th ed. R i c h m o n d , Va.: Federal Reserve Bank of Richmond, 1993a. . "Behind the Money Market: Clearing and Settling Money Market Instruments." In Instruments of the Money Market, edited by Timothy Q. Cook and Robert K. LaRoche, 173-87. 7th ed. Richmond, Va.: Federal Reserve Bank of Richmond, 1993b. Goodfriend, Marvin S. " C o m m e n t on Gertler." Journal of Money, Credit, and Banking 20 (August 1988, pt. 2): 589-96. . "Money, Credit, Banking, and Payment System Policy." In The U.S. Payments System: Efficiency, Risk, and the Role of the Federal Reserve, edited by David B. Humphrey, 24777. Boston: Kluwer Academic Publishers, 1990. Miron, Jeffrey A., and Christina D. Romer. "A New Monthly Index of Industrial Production, 1884-1940." Journal of Economic History 50 (June 1990): 321-37. G o o d f r i e n d , M a r v i n S., and W i l l i a m W h e l p l e y . " F e d e r a l Funds." In Instruments of the Money Market, edited by Timothy Q. Cook and Robert K. LaRoche, 7-21. 7th ed. Richmond, Va.: Federal Reserve Bank of Richmond, 1993. Mishkin, Frederic S. " A s y m m e t r i c Information and Financial Crises: A Historical Perspective." In Financial Markets and Financial Crises, edited by R. G l e n n H u b b a r d , 6 9 - 1 0 8 . Chicago: University of Chicago, 1991. G o o d h a r t , C h a r l e s . The Evolution of Central Banks. C a m bridge, Mass.: M I T Press, 1988. . "Financial Globalization, Derivatives, Volatility, and . "Preventing Financial Crises: An International Perspective." N B E R Working Paper No. 4636, February 1994. the Challenge for the Policies of Central Banks." London School of Economics Financial Markets Group, Special Paper No. 74, October 1995. Gorton, Gary. "Banking Panics and Business Cycles." Oxford Economic Papers 40 (1988): 751-81. Gorton, Gary, and Donald Mullineaux. " T h e Joint Production of C o n f i d e n c e : E n d o g e n o u s R e g u l a t i o n and N i n e t e e n t h C e n t u r y C o m m e r c i a l B a n k C l e a r i n g h o u s e s . " Journal of Money, Credit, and Banking 19 (November 1987): 457-68. Kahn, Charles M „ and William Roberds. "On the Efficiency of Cash Settlement." Federal Reserve Bank of Atlanta Working Paper 95-11, November 1995. 30 Economic Review Moen, Jon R „ and Ellis W. Tallman. " T h e Bank Panic of 1907: The Role of Trust Companies." Journal of Economic History 52 (September 1992): 611-30. . "Liquidity Shocks and Financial Crises during the Nat i o n a l B a n k i n g E r a . " F e d e r a l R e s e r v e B a n k of A t l a n t a Working Paper 93-10, August 1993. Morgenstern, Oskar. International Financial Transactions Business Cycles. Princeton, N.J.: N B E R , 1959. and Myers, Margaret G. The New York Money Market. Vol. 1, Origins and Development. New York: C o l u m b i a University Press, 1931. New York Foreign Exchange Committee. "Reducing Foreign Exchange Settlement Risk." Unpublished paper, October 1994. May/June 1995 Noyes, Alexander D. Forty Years of American Finance. New York: G.P. Putnam's Sons, 1909. Reprint. New York: Arno Press, 1980. Roberds, William. " T h e Rise of Electronic Payments Networks and the Future Role of the Fed with Regard to Payment Finality." Federal Reserve Bank of Atlanta Economic Review 78 (March/April 1993): 1-22. R o m e r , Christina. " T h e Prew Business Cycle Reconsidered: N e w E s t i m a t e s of G r o s s National Product, 1 8 6 9 - 1 9 0 8 . " Journal of Political Economy 97 (February 1989): 1-37. Schwartz, Anna J. "Real and Pseudo-Financial Crises." In Financial Crises and the World Banking System, edited by F o r r e s t C a p i e and G e o f f r e y E. W o o d , 1 1 - 3 1 . L o n d o n : M a c m i l l a n , 1986. Reprinted in Financial Crises, vol. 1, edited by M i c h a e l B o r d o , 3 - 2 3 . L o n d o n : E d w a r d Elgar, 1992a. the U.S. Senate. 61st Cong., 2d session. Doc. 538. Washington, D.C.: Government Printing Office, 1910. . " T h e Crisis of 1914 in the United States." American Economic Review 5 (September 1915): 500-533. Summers, Lawrence H. "Planning for the Next Financial Crisis." In The Risk of Economic Crisis, edited by Martin Feldstein, 135-58. Chicago: University of Chicago Press, 1991. Tallman, Ellis W., and Jon R. Moen. "Private Sector Responses to the P a n i c of 1907: A C o m p a r i s o n of N e w Y o r k and Chicago." Federal Reserve Bank of Atlanta Economic Review 80 (March/April 1995): 1-9. Thomson, James B. " T h e National Depositor Preference L a w . " Federal Reserve Bank of Cleveland Economic Commentary, February 15, 1994. T i m b e r l a k e , Richard H., Jr. Monetary Policy in the States. Chicago: University of Chicago, 1993. United . "The Misuse of the F e d ' s Discount W i n d o w . " Federal Reserve Bank of St. Louis Review 74 (September/October 1992b): 58-69. Wall, Larry D. "Too-Big-to-Fail after F D I C I A . " Federal Reserve Bank of Atlanta Economic Review 78 (January/February 1993): 1-14. Smith, Stephen D., and Larry D. Wall. "Financial Panics, Bank Failures, and the Role of Regulatory Policy." Federal Reserve Bank of Atlanta Economic Review 11 (January/February 1992): 1-11. Woelfel, Charles J., ed. The Fitzroy Dearborn Encyclopedia of Banking and Finance. Chicago: Fitzroy Dearborn, 1995. Wolfson, Martin H. Financial Crises: Understanding the Postwar U.S. Experience. 2d ed. Armonk, N.Y.: M.E. Sharpe, 1994. Sprague, O . M . W . History of Crises under the National Banking System. Report by the National Monetary Commission to Federal Reserve Bank of Atlanta Economic Review 31 Essay Risk-Based Bank Capital: Issues and Solutions Robert R. Bliss anks are subject to many forms of risk, of which credit and market risk are perhaps the most important. Credit risk involves the risk that a counterparty to a contractual obligation, be it a mortgage, loan, or swap agreement, will default on the promised payments. Market risk is the risk that the values of assets or the cash flows from assets will change in response to movements in broad market factors, such as interest or exchange rates. Traditionally, banking regulation has focused on credit risk, the quality of assets, and internal control systems. But financial markets have changed so that market risk has become increasingly important. 32 The author is a senior economist in the financial section of the Atlanta Fed's research department. He thanks Peter Abken, Larry Wall, and Jerry Dwyerfor helpful comments. Risk is an integral part of bank business. In assessing the creditworthiness of a loan applicant the bank makes a j u d g m e n t about the riskiness of the loan. In taking a position in the foreign exchange market the bank takes on a risk that it factors into the price quoted to its customer. Regulatory interest is not in controlling the risk a bank can take on per se but in limiting the chances that adverse outcomes will exceed the bank's capacity to bear losse s — h e n c e the regulatory focus on bank capital, which provides a b u f f e r against the potential for losses inherent in the bank's conduct of its normal business. For exposures to market risks from trading desk operations, however, existing regulations for the determination of bank capital, based on the quality of assets held, are not appropriate. Economic Revieiv September/October 1995 Derivatives—financial assets whose value and payoffs are determined by the value of an underlying asset or index—are used to transfer risks from one party to another (at a cost, of course) and thus are a means of risk m a n a g e m e n t , m u c h like insurance. Derivatives such as forward foreign exchange contracts, interest rate swaps, commodity and financial futures, together with, more exotic variants such as caps, swaptions, and structured notes, have grown explosively in the past twenty years, though some types of derivatives are as old as financial markets themselves. These instruments are now an integral part of international trade, finance, and corporate financial risk management. A s the market has grown, certain large commercial banks have become lead players, competing directly with investment banks to create and sell derivative "products" in order to meet the risk-management needs of their customers. Increasingly, providing these products, either directly or through correspondent relations, will become important for smaller banks as well. Derivatives also have a dark side. They have been the subject of widespread and sometimes lurid publicity. S o m e consider t h e m u n i m a g i n a b l y c o m p l i c a t e d , dangerously risky, even a threat to the financial system. Rightly or wrongly, derivatives have been associated with a number of well-publicized financial disasters in recent years. 1 The resulting furor has led to demands that Congress and regulatory agencies " d o something!" S o m e might argue for an outright ban on derivatives trading by commercial banks. But such a ban would only drive the market (and its associated revenues) offshore or into nonbank institutions. It would therefore be futile and would simply hamstring U.S. commercial banks in the global financial marketplace. One can no more ban derivatives than the Luddites could ban power looms in the early nineteenth century. While fear drives public calls for regulation, there are also sensible reasons for reevaluating the current approach to regulating banks' trading activity. Bank involvement in derivatives trading represents a new and very different business from the traditional role of credit assessment and loan origination, and traditional methods of assessing bank capital are not appropriate to this new business line. For a trading desk's portfolio, the primary sources of risk are market factors—interest rates, exchange rates, mortgage prepayment rates—not credit factors. This environment has led to the discussion of "risk-based capital" assessment. Risk-based capital fits into a larger framework of the bank's overall capital. The Federal Deposit Insurance Corporation Improvement Act contains provisions for increasingly stringent supervisory intervention as capi Federal Reserve Bank of Atlanta tal ratios fall. The concern is that trading desk activities may lead to rapid changes in bank capital because of the potential volatility of the trading portfolio's value. An additional concern is that failure of large banks involved in derivatives origination and market making may have systemic implications. For these two reasons, regulators are subjecting trading risk to special scrutiny. Beginning in 1993, the Basle Committee on Banking Supervision (1993) outlined the need for requiring the assessment of capital to cover trading-portfolio risk and discussed means of doing so. The current regulatory discussion follows from that initiative. The basic goal of risk-based capital assessment is to determine the optimal level of risk-based capital a bank should hold against possible losses in its trading portfolio. Determining what is optimal involves trading off the costs of i m p l e m e n t a t i o n and h o l d i n g e x c e s s i v e amounts of capital, on the one hand, against the need to ensure that sufficient capital is available to cover reasonably likely outcomes given the bank's positions, on the other hand. Because portfolio positions are changing rapidly, it is desirable to have a means of assessing capital r e q u i r e m e n t s that is responsive to these changes. Fixed-capital requirements are likely to be too high or too low. If too high, b u r d e n s o m e capital requirements place banks under a competitive disadvantage relative to o f f s h o r e and n o n b a n k c o m p e t i t o r s . Alternatively, if too low, capital requirements do not provide adequate protection from losses, thus placing the bank's other activities at risk and ultimately passing the risk on to deposit insurers. T h e r e are three m a j o r p r o p o s a l s f o r d e t e r m i n i n g risk-based capital. These are (1) the standard or supervisory model approach, (2) the internal m o d e l s approach, and (3) the precommitment approach. Both the standard and internal models proposals are concerned with regulating, to a greater or lesser degree, the models used internally by banks for risk assessment and management. These are referred to herein as the modelsbased approaches. The precommitment approach is not models-based regulation in that it does not attempt to regulate models. It e m p h a s i z e s incentives and goals while leaving modeling issues entirely to banks. Models-Based Approaches to Risk-Based Capital It is w o r t h w h i l e r e f l e c t i n g o n the c o m p l e x a n d d y n a m i c n a t u r e of m o d e r n s e c u r i t i e s and s e c u r i t y markets. C o m p e t i n g pricing and h e d g i n g m o d e l s are Economic Review 33 d e v e l o p e d by so-called rocket scientists within investment banks and a c a d e m i c s at universities. These m o d e l s are based on financial-asset pricing theory, are necessarily cast in terms of sophisticated m a t h e matics, and their i m p l e m e n t a t i o n involves c o m p l e x statistical issues. T h e relative merits of these m o d e l s are hotly debated, and their d e v e l o p m e n t is ongoing. Not only are they d y n a m i c and subject to disagreement a m o n g experts, but the securities these m o d e l s are used to price and hedge are also rapidly evolving in response to c h a n g i n g m a r k e t forces and the efforts of intermediaries to provide products t o sell to customers. Bank involvement in derivatives trading represents a new and very different business, and traditional me thods of assessing bank capital are not appropriate to this new business line. There is a continuing dynamic between the regulators, attempting to devise regulations that will meet social o b j e c t i v e s , and the r e g u l a t e d , a t t e m p t i n g to m a x i m i z e their profits within and around regulatory c o n s t r a i n t s . T h i s c o n f l i c t in g o a l s and i n c e n t i v e s sometimes leads to a contest that regulators rarely win. Regulations must necessarily be general and written so that c o m p l i a n c e is u n a m b i g u o u s . R e g u l a t i o n s take time to write, inevitably involve compromises, and tend to evolve slowly. In contrast, firms can respond quickly to changing markets and can adjust their business practices to the regulations in ways that are difficult to anticipate and that m a y p r o d u c e u n i n t e n d e d social consequences. Responding to these innovations only leads to another round of regulations and innovations, with the regulations becoming increasingly complex, burdensome, and costly to monitor. Viewed in terms of incentives, some approaches m a y b e counterproductive while others minimize the asymmetry between the goals of regulators and the regulated. T h e Standardized Model A p p r o a c h . The standardized model approach would have a single model, designed by regulators, applied to all banks. This ap- 34 Economic Revieiv proach is designed to keep the reporting burden from being excessive and to provide a f r a m e w o r k that supervisory personnel can verify. By defining the model to be used for determining risk-based capital and by deciding many of the judgment questions that keep model builders occupied, the standardized model might, in principle, be free of the temptation to " g a m e " the system to reduce capital set-asides. T h e underlying philosophy of this model is to divide securities into broad categories and then to assign weights to these categories. Unfortunately, in practice this approach is an invitation to gaming. For instance, one question raised in the proposed regulations was whether undiversified equity portfolios should be assessed an additional 8 percent risk-based capital set-aside. Clearly, "diversification" is not an either/or quality, and an 8 percent additional capital set-aside would be too much for some portfolios and too little for others. A t t e m p t s to use such rules of thumb to reduce complex and continuously varying properties into a f e w discrete categories are apt to lead to unsatisfactory results. Hugh Cohen (1994) assessed a related, and similar, regulatory proposal for measuring loan-portfolio interest rate risk. He showed that even for default-free bonds, the simplest of securities, portfolios that are "equivalent" under the proposed regulations will have widely varying risk characteristics. Adding complexity to a standardized model to attempt to solve these problems is an exercise in futility. Attempting to adapt the model to such circumstances will make it increasingly complex, unwieldy, and costly to implement and monitor. T h e "one-size-fits-all" approach implicit in the standardized model approach does not reflect the diversity of portfolios and strategies that exist. Neither is it likely to keep up with changing circumstances. Portfolio positions change rapidly, requiring real-time monitoring. Contrast the rapidity of Baring's decline with the once-a-quarter reporting requirements for most banks. Financial innovation and customization of financial products means that a standardized model may be outdated before it is promulgated. While the model may address, inadequately, "vanilla" securities, it will not be able to h a n d l e either c u s t o m i z e d or n e w l y c r e a t e d types. The regulatory agencies will be challenged to develop the highly technical models needed to be suitably firm-specific, rapidly evolving, and flexible. The Internal Models Approach. As a result of criticisms of the standardized model approach, the internal models approach has been advanced as an alternative (see Board of Governors 1995a). The assumption underlying this approach is that banks are in a better position to devise models suitable to their risk-management September/October 1995 needs than are regulators. Risk-management models already exist within banks. The proposed internal models approach seeks to piggyback on a bank's existing riskmanagement model to determine levels of risk capital to be held. At the heart of the internal models approach is the "value-at-risk" calculation, whereby the m a x i m u m loss that a portfolio is likely to experience in a given time interval is quantified to a certain level of probability. The output of such a model is a measure of value at risk, or VAR. For example, a 5 percent VAR of $1 million means that a loss exceeding $1 million is expected to occur one period out of twenty, at most. Of course, the bank expects to be making profits on average. Such calculations are performed routinely by banks with active trading portfolios to limit their e x p o s u r e s over short time intervals. Investment and commercial banks use daily VARs because that horizon fits into their riskm a n a g e m e n t systems, which monitor and adjust the bank's overall position risk on a daily basis. Under the internal models approach, regulators would then adjust the b a n k ' s daily VAR to reflect the longer period of regulatory interest, say, a quarter, by some fixed factor to arrive at the required capital level. Unfortunately, over longer horizons, the nonlinear payoffs of options in many portfolios make it impossible to extrapolate risk exposures linearly on the basis of one-day VAR calculations. Consider a portfolio that has written in an out-of-the-money option on interest rates that are very unlikely to become in-the-money in a single day. This position contributes nothing to the one-day VAR. But the possibility of a large interest rate m o v e over, say, a week may be such that the probability of the option going in-the-money becomes important. Thus, the potential losses from the option position over a week are not just a simple multiple of the potential losses over a single day; the potential losses are a nonlinear function of the time interval. Additionally, extrapolating from single-day potential losses to longer periods assumes a static portfolio position. In reality, a trading desk would be constantly adjusting its portfolio to reflect changing market conditions. Because there is no economic model for determining how to extrapolate daily VARs, which banks' internal m o d e l s p r o d u c e , to the h o r i z o n s of interest f o r capital assessment, the proposed regulations simply pick a multiplier number: 3.16. 2 This number would be the s a m e for all b a n k s r e g a r d l e s s of their portfolio composition or internal model performance. In order to ensure the adequacy of most banks' capital this multiplier will likely be conservative (that is, high). Although for a few banks with risky portfolios the risk- Federal Reserve Bank of Atlanta based capital will be too low, for most banks this approach imposes a risk-based capital requirement that is unduly burdensome given the actual risk of their portfolios. Since capital is expensive the requirement will impose unnecessarily high costs on banks and place them at a disadvantage relative to their competitors. It will then be natural for banks to reduce their effective capital costs by increasing their multiperiod risks relative to their daily VARs, for instance by increasing the use of securities with nonlinear payoffs, thus gaming the regulations and frustrating their intent. T h e proposed internal-models regulations seek to constrain banks' internal models in various ways by defining acceptable inputs, limiting permissible relations, segregating various types of securities for separate treatment, and so forth. The proposed regulations a d d r e s s the nonlinearity issue by directing that the banks internal models incorporate the nonlinearities. Of course, the internal models currently do just that, but over a one-day horizon. A different model would be needed to adjust for nonlinearities over different horizons. Backsliding into modeling, of course, runs counter to the premise that "banks know best" when it comes to constructing models. Because banks are going to maintain for their own internal uses models that reflect their best judgments as to how to build models, regulatory restrictions may well lead to a second set of models maintained only for risk-based capital determination. This development invites banks to "adjust" these regulatorily constrained models to minimize their capital requirements. Thus, by m i c r o m a n a g i n g m o d e l i n g , the internal m o d e l s a p p r o a c h will suffer f r o m the s a m e "gaming" problem as the standardized model approach. Unfortunately, the VAR approach is also inherently flawed. The primary flaw is that it creates adverse incentives for banks. Because the output of the bank's internal risk-management models, even if banks are left to their o w n best j u d g m e n t , is to be used to impose costs (capital set-asides) on the bank, banks will, perfectly rationally, weigh the costs p r o d u c e d by their m o d e l s against the benefits of h a v i n g " b e t t e r " riskmanagement systems. Management and model builders alike will be conscious of the dual objectives, and this ambiguity cannot help the pursuit of optimal internal risk-management systems. In an extreme case a bank may maintain separate internal models, one for capital assessment, another for risk management. This separation would destroy the premise of the internal models approach, which is that banks are best able to design models to m e a s u r e their portfolio risk. O n c e a bank starts keeping separate books, so to speak, the persons building the model used for capital assessment will no Economic Review 35 longer be trying to measure risk. They will be trying to reduce costly capital set-asides, at least to the extent that they can get away with adjusting their models to do so. Such extreme g a m i n g is apt to be rare but is more likely in those banks that are of the greatest regulatory concern, thus making regulatory oversight all the more difficult. Another challenge for regulators is verifying the accuracy of the b a n k ' s VAR m o d e l to ensure that the bank is not trying to game the regulations by adjusting its models. Verifying the accuracy of these m o d e l s — particularly in measuring low-probability events—will be difficult and unreliable, as Paul Kupiec (1995) argues. The apparent solution to the inadequacies of the internal models approach is to mandate very conservative, and hence costly, capital requirements. The second flaw in the internal models approach is inherent in the concept of VAR itself, namely, that it focuses solely on the probability of losses greater than a specified amount. It totally ignores how large those losses are expected to be when the bound is violated. Suppose that a bank had losses exceeding its 99 percent VAR 2 percent of the time but never by more than 1 percent of the bank's total capital. The VAR standard would say that the bank's model was inadequate, inviting regulatory intervention. Meanwhile, another bank that violates its 99 percent VAR only 0.5 percent of the time but on average by 20 percent of its capital would have an "adequate" model (and hence level of risk capital) by the VAR standard. The magnitude of the second b a n k ' s losses clearly indicates that this bank is riskier. The fixed VAR multiplier inherent in the internal models approach is unable to distinguish between these situations: it can address only the average-loss distribution. B a n k s are therefore invited to g a m e the system by investing in projects that trade slightly higher expected returns for larger (though no more likely) potential losses. No single-figure measure of risk can capture an entire probability distribution or even the tail of a distribution. Thus, the attempt to leverage the risk-based capital set-aside off the VAR n u m b e r will not capture the distribution of potential losses and will lead to inadequate or b u r d e n s o m e capital set-asides. VAR is useful, but it should not be used in isolation. Because regulators and regulations are ill-adapted to regulating model building, it is worthwhile to consider whether they should and why they would want to. Returning to the basic purpose of regulation—to provide sufficient capital to ensure that, in most situations, the trading losses do not endanger the bank's solvency—it b e c o m e s obvious that model-based regulation is one step removed from the ultimate objective. 36 Economic Revieiv 71ie Precommitment Approach to Risk-Based Capital An alternative to models-based regulation with its inherent conflict between the social goals underlying regulation and the profit incentives of the regulated is to attempt to devise regulations in such a way as to align the firms' incentives with the regulators' goals. Such an approach to achieving regulatory goals has been proposed by Kupiec and James O ' B r i e n (1995b), t w o Federal Reserve Board economists. T h e y argue that the models-based approach fails to satisfy two critical requirements: "(1) that an internal model can accurately measure the bank's risk exposure over a holding period of concern to regulators and (2) that the regulatory authority can verify that each bank's model is indeed providing such an accurate measure of the bank's exposure" (Kupiec and O ' B r i e n 1995a, 43). 3 Kupiec and O ' B r i e n ' s alternative approach focuses on goals—namely, maintaining sufficient capital to cover trading losses—and leaves it to banks to determine the best models and inputs to achieve the goals. Banks would determine their optimal amount of capital indirectly by determining a maximum-loss precommitment over a reporting period. This precommitted m a x i m u m loss would then be used to determine the appropriate capital set-aside. Bankers' incentives to announce reasonable precommitments and thus set aside sufficient capital lies in the penalties that w o u l d be i m p o s e d should a bank's trading losses in a reporting period exceed the amount previously chosen by the bank as its m a x i m u m expected loss. Banks that have good riskmanagement systems or conservative portfolios could precommit to lower m a x i m u m loss levels and hold less capital because of their confidence that they will not breach their precommitted m a x i m u m trading losses. Conversely, banks with fewer resources in risk management, or less confidence, or simply more conservative (risk-averse) preferences would choose a higher precommitment level and consequently higher capital levels. Under the precommitment scheme, regulators would not intrude on banks' market-risk models and control procedures. An additional benefit of the Kupiec and O ' B r i e n approach is that costs are determined by the banks rather than imposed on them directly. If they feel that devoting m o r e resources to their models is worthwhile in terms of reduced probability of penalties, they can invest those resources. If not, they need not. In any case, the agencies simply provide a set of incentives (schedule of penalties) to m o t i v a t e their decisions. If the September/October 1995 penalties are properly designed, they reduce or eliminate the temptation to game the system to reduce capital set-asides. Neither of the two model-based approaches is able to do this. Conceptually the Kupiec and O ' B r i e n precommitment approach has m u c h appeal. A n u m b e r of details in the implementation are, however, quite important to m a k e it w o r k in practice. T h e s e include (1) the form the penalties should take, (2) the penalty schedule, (3) the reporting frequency, (4) the reporting and regulatory burden this approach will entail, (5) whether the precommitment approach should be used alone or in combination with another approach, and (6) the link between precommitment and capital. What Form Should Penalties Take? The key to the precommitment approach lies in the incentives that, for the approach to work, should focus the bank's efforts on risk assessment and loss reporting, not on gaming the capital set-aside. Penalties need to be large enough in proportion to precommitment violations to provide an effective deterrent to deliberately underestimating potential losses but not so large as to force banks to overcommit, thus maintaining unnecessarily high levels of capital or, worse yet, providing incentives to hide losses so as to avoid penalties. It is necessary that the bank be prevented from manipulating its precommitment level to minimize its capital setaside, so incentives unrelated to the capital set-aside itself should b e used to ensure that p r e c o m m i t m e n t s accurately reflect portfolio risk or at least do not underestimate potential losses. W h y C a p i t a l - B a s e d P e n a l t i e s Will N o t Work. Careful analysis suggests that the use of increased future capital requirements as a penalty cannot achieve the desired focus. If violation of a precommitment level results only in an increase in the ratio of precommitment to capital set-aside in the future, there is no certain penalty today. The bank then responds to a precommitment violation either by reducing next period's reported precommitment or increasing the risk of its portfolio, in either case n u l l i f y i n g the penalty. It is worth r e m e m b e r i n g that p r e c o m m i t m e n t s are worstcase, so actual violations will (should) be rare and gaming therefore hard to detect. If instead of adjusting the ratio of precommitment to capital set-aside the penalty takes the f o r m of an increase in capital unrelated to the f u t u r e precommitments, then the entire premise of the precommitment Federal Reserve Bank of Atlanta approach is undermined. Banks will have no incentive whatsoever to report accurate risk measures. W h y Fines Will Work. What is needed is a penalty that is certain and that cannot be " u n w o u n d " by any action the bank takes. Fines meet these criteria. Once a p r e c o m m i t m e n t level is b r e a c h e d , the b a n k w o u l d incur an immediate cost (the fine). There is then no incentive to game the system for the next reporting period. Any attempt to do so by underreporting future potential losses (to reduce capital requirements) would only increase the probability and magnitude of potential fines in the next period. Banks cannot profitably play a multiperiod game to reduce capital set-asides in proportion to their trading risk. Discipline through capitalbased penalties cannot do this. 71ie Need for Nonlinear Penalties A second characteristic that effective penalties must have is that they increase nonlinearly in the size of the precommitment violation. This requirement would provide disincentives to deferring t o d a y ' s losses in the hope that future outcomes will reverse them. The hiding of losses with the concomitant taking of increasingly risky positions in an attempt to "bail out the boat" characterized debacles such as Barings and Daiwa, and any risk-management system, including fixed-capital and m o d e l s - b a s e d a p p r o a c h e s , has the potential for such problems. A linear penalty function means that banks face no downside in deferring today's losses until tomorrow rather than reporting them today. In the worst case the bank will have additional losses tomorrow and, if it gives up the g a m e and reports all losses then, it will h a v e penalties equal to the sum of the p e n a l t i e s it would have paid had it reported losses each day. This deal is a good one for banks—but not for society. It defers the penalty payment, for one thing. For another, deferral avoids for one period the fixed costs associated with loss reporting, such as increased regulatory attention, i m p a c t s on stock prices, and m a n a g e m e n t compensation. T h e real p r o b l e m , though, lies in the question, Why not continue the game indefinitely? Increasing penalties nonlinearly breaks d o w n this perverse incentive to defer losses. If a bank defers losses until tomorrow and then loses again it will have to pay a much larger penalty than the total it would have had to pay if it recognized the losses each period. Encouraging banks to report small losses early wipes the slate clean each period and has losses flowing through the financial Economic Review 37 statements while they are still small in comparison with capital levels. A schedule of fines can easily be devised to provide an appropriate penalty schedule. 4 The Need for Frequent Reporting and Assessment To further reduce incentives for deferring, or hiding, losses it is necessary to have losses and fines assessed frequently. Kupiec and O'Brien's (1995b) original precommitment proposal suggested a regulatory period of three months. This period is still long enough that if a bank has excessive losses early in the quarter it may be tempted to undertake m o r e risky positions in an attempt to reverse the results before it has to pay the piper. As the regulatory period is shortened, it becomes increasingly difficult for banks to pursue such a strategy. In addition, except for extreme cases, losses over shorter intervals may be expected to be smaller in proportion to available capital, the per-period costs of recognizing losses (and paying any fines) will be less, and therefore the temptation to defer (or hide) losses will be correspondingly reduced. /Reporting Burden The inherent costs of frequent reporting need not be excessive. Each reporting period a bank would report two numbers: its gains/losses for the previous period and its precommitment for the next period. The gains/ losses should be readily available from the bank's own profit-and-loss tracking programs. The raw inputs for the precommitment numbers should be available from its internal risk-management systems. T h e bank will have to decide how it will arrive at the actual precommitment numbers. The t w o reported n u m b e r s would then be fed into a computer program that would determine if a fine needed to be assessed and in addition would look for repeated precommitment violations. In the latter case supervisory personnel would be alerted to the potential problem. In the beginning it is envisioned that only about thirty banks would be subject to these regulations, m o s t others having no significant trading activity. The p r e c o m m i t m e n t approach compares favorably with the work hours required for imp l e m e n t i n g the standardized m o d e l or a d a p t i n g the internal model and then having these checked in detail by examiners. 38 Economic Revieiv Other Issues Some concerns have been raised that the penalties inherent in the precommitment approach will be destabilizing and may push a marginal bank into insolvency. Obviously this result is not the intention of the penalties. However, a bank that experiences losses that make it unable to pay its fines comfortably is a bank that is in trouble, whether or not the fines are enforced. In such cases waiving the fines and letting trading proceed only invites compounding the problem. Frequent reporting will reduce the chance that fines will be large relative to capital and provide incentives for the bank to alter its strategy before losses become large. Regulatory forbearance with respect to fine assessment should be avoided except in times of systemic stress (such as a market crash). Such episodes should be determined by the central regulatory authority. A request for bank-specific relief highlights the bank-specific nature of the problem. A policy of permitting bank-specific f o r b e a r a n c e in times of normal market volatility undermines the incentive structure on which the precommitment approach is built. The use of fines, particularly on a regular and automatic basis, seems strange to current regulatory practice. It need not be. The fines envisioned in this paper are not punishments, per se, for malfeasance, but rather are an incentive device and a device for imposing costs in proportion to regulatory risk. The schedule of fines imposes costs on those banks that maintain low levels of capital set-asides relative to their actual portfolio risks. These banks are the ones most at risk of becoming regulatory burdens, and it is perfectly fitting that they should pay a higher cost for this increased risk. Meanwhile banks that are less at risk will not pay the costs. Fines only m a k e explicit and bank-specific the costs that the one-size-fits-all models-based approaches implicitly impose on all banks. T h e amounts assessed are, in effect, risk-based insurance premiums, wherein risk is revealed in the adequacy of the individual bank's precommitment levels relative to their realized losses. A potential weakness of any system that seeks to address market risks is the need to mark to market assets that may not have readily observable market prices. M a r k i n g - t o - m a r k e t is d o n e in the n o r m a l course of business for the bank's internal profit-and-loss tracking by means of models or traders' "sense of the market." The less liquid and more specialized a security is, the more uncertainty surrounds this valuation. Obviously the result is adverse incentives for traders and perhaps firms to shade their valuations to their own advantage. September/October 1995 Countering this incentive is the firm's interest in its own long-term survival, which makes it important to have the best possible picture of its positions. The valuation problem is inescapable. It applies equally to any approach to capital management. With volatile financial securities, recourse to accounting cost numbers is inappropriate, or worse. Examiners simply need to be cognizant of the problem, as they are of other potential problems such as inflated real estate appraisals, and check for clues that valuations are being manipulated. In Kupiec and O ' B r i e n ' s proposal and in the proposed regulations, the multiplier used to determine capital set-aside f r o m p r e c o m m i t t e d m a x i m u m losses is fixed at unity. This stipulation is unduly restrictive. Adjustments to the ratio of precommitment to capital setaside, used in combination with fines, should be one of the additional tools available to regulators to reward or penalize banks. While the precommitment system is being implemented, and both banks and regulators are gaining experience, a higher initial multiplier would provide an additional degree of safety. This precaution m a y be dispensed with as banks gain experience and banks with good risk-assessment experience are identified by their history of staying within their precommitted maximum loss levels. Setting the base multiplier to a higher number also provides the flexibility to reward banks with good (conservative) loss estimates by lowering their multiplier. An additional benefit of a variable multiplier is that it provides examiners and other regulators with a degree of flexibility in implementing the regulation without reducing the certainty of the penalties, which is essential for them to be credible. That said, adjustments to the ratio of precommitment to capital set-aside and other regulatory sanctions should be used only as an addition to a fixed, certain schedule of fines. Without fines, gaming of the other sanction procedures will become feasible. Summary and Conclusions Neither the standardized model nor the internal model approach permits the design of a system of rewards and penalties that would align the incentives of bank management with those of regulatory authorities. All these approaches can do is attempt to shortstop the effects of the adverse incentives they create, which requires more costly and burdensome supervision. T h e Federal Reserve Bank of Atlanta precommitment approach, however, if carefully implemented, can achieve the desired incentive compatibility. It is reasonable and proper for bank regulatory agencies to specify goals and penalties. Regulators should avoid the temptation to micromanage banks' models, though, and should focus instead on the outputs. Both the standardized and internal m o d e l s a p p r o a c h e s to risk-based capital have serious disadvantages. In particular, both lead to gaming. In the standardized model the opportunity to game the system flows from the inflexible and static nature of regulations that, because they are not adaptable, are standardized across a variety of differing portfolios and market conditions. In the internal model the temptation to game flows from the adverse incentives the approach imposes on banks. In both cases adjustments to the problems are likely to make the models expensive to the regulated and burd e n s o m e to the regulators. If an internal models approach is to be adopted, model implementation should be left to the banks, and regulators should concern themselves with how well the models work. Thus, regulations regarding how actual losses are to be compared with a model's VAR predictions are reasonable, as are actions to be taken and penalties to be imposed when the model's forecasts prove inadequate. The precommitment approach represents a radical departure f r o m the detailed, check-list, look-over-theshoulder approach to regulation. It recognizes that focusing on results is more important and, in this case, m o r e feasible than double-checking the work of the regulated. This approach does not mean, of course, that there is no need for supervision beyond the reporting and tracking of precommitments and results. But, in the absence of evidence of gross problems—for example, a trader hiding losses or a pricing model deferring losses while recognizing gains—the focus should be on the precommitment levels and loss experience. When precommitments are frequently violated, a more intrusive regulatory intervention is warranted, including raising the capital set-aside multiplier or ultimately prohibiting the bank from trading. In a world in which financial modeling is becoming increasingly complex and esoteric and supervisory resources are increasingly stretched, the precommitment approach can enable regulators to identify and focus their efforts on problem banks while providing incentives for the majority of banks to stay out of trouble without constant, detailed oversight. Economic Review 39 Notes 1. It may be well to remember that none of the recent highly publicized debacles has arisen from errors in pricing models. Barings and Daiwa resulted from internal control problems, and bank regulators should pay strict attention to these. Orange County's disaster grew out of the trader's certainty that he knew what interest rates would do (and such hubris can hardly be regulated) together with lax supervision. Metallgesellschaft had to do with hedging, not valuation, and is still being vigorously debated by academics on both sides. One case involving valuation, that of Bankers Trust, has less to do with whether Bankers Trust's models were right than with whether Bankers Trust fully informed its customers. The moral here is that customers should treat marketmakers like car salesmen and verify valuations with a neutral third party or build their own models. This is not to say that models are never a prob- lem. In the early 1990s, collateralized mortgage obligation (CMO) valuation models were found to be painfully inadequate when a sudden drop in interest rates caused unanticipated rates of prepayment with corresponding losses to holders of some C M O s (and, of course, corresponding gains to others). 2 . T h e daily V A R , or a v e r a g e of the last sixty d a y s ' daily VARs, whichever is greater, is first multiplied by 3.16 (the square root of 10) to convert to a ten-day VAR. This number is then multiplied by a minimum factor of three to arrive at the required capital set-aside. 3. Their alternative, introduced in Kupiec and O'Brien (1995b), is open for public comment in the Federal Register (Board of Governors 1995b). 4. Kupiec and O'Brien (1995b) have proposed one such nonlinear penalty schedule. References Basle Committee on Banking Supervision. "'Proposal to Issue a Supplement to the Basle Capital Accord to Cover Market Risks." April 30, 1993. Board of G o v e r n o r s of the Federal Reserve S y s t e m . " R i s k Based Capital Standards: Market Risk" (Docket No. R-0884). Federal Register 60 (July 25, 1995a): 38082-142. . "Capital Requirements for Market Risk" (Docket No. R-0886). Federal'Register 60 (July 25, 1995b): 38142-44. Cohen, Hugh. "Data Aggregation and the Problem of Measuring a Bank's Interest Rate Exposure." Federal Reserve Bank of Atlanta Working Paper 94-6, August 1994. 40 Economic Revieiv Kupiec, Paul. "Techniques for Verifying the Accuracy of Risk M e a s u r e m e n t M o d e l s . " Journal of Derivatives 3 (Winter 1995): 73-84. K u p i e c , Paul, and J a m e s O ' B r i e n . " I n t e r n a l A f f a i r s . " Risk 8 (May 1995a): 43-47. . " A Pre-commitment Approach to Capital Requirements for M a r k e t R i s k . " Federal R e s e r v e B o a r d of G o v e r n o r s Working Paper 95-36, July 1995b. September/October 1995 Bulk Rate U.S. Postage PAID RESERVE Atlanta, GA Permit 292 B A N K OF ATLANTA Public Affairs Department 104 Marietta Street, N.W. Atlanta, Georgia 30303-2713 ( 4 0 4 ) 5 2 1 - 8 0 2 0 ® printed on recycled paper