The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
Bank Deposits and Credit As Sources of Systemic Risk R O B E R T A . E I S E N B E I S The author is senior vice president and director of research at the Atlanta Fed. T HERE ARE MANY DIFFERENT WAYS TO DEFINE A FINANCIAL CRISIS. INDEED, THE ECONOMICS AND FINANCE LITERATURE IS FILLED WITH TERMS LIKE PANIC, FINANCIAL CRISIS, RUNS, SYSTEMIC CRISIS, OR CONTAGION.1 THERE IS IN FACT LITTLE AGREEMENT ON EVEN THE RUDI- MENTARY DEFINITIONS OF A FINANCIAL CRISIS, THE SEQUENCE OF EVENTS CONSTITUTING A CRISIS, OR THE CAUSES OF THESE EVENTS. The professional discussion divides itself into two broad categories. Macroeconomists typically are concerned with explaining business cycle fluctuations and determining when a recession will degenerate into a depression.2 They are equally interested in the financial system’s role as a propagator of this process because most depressions have been accompanied by serious disruptions in the financial system, including banking failures and panics. Eichengreen and Portes, for example, define a financial crisis as “a disturbance to financial markets, associated typically with falling asset prices and insolvency among debtors and intermediaries, which ramifies through the financial system, disrupting the market’s capacity to allocate capital within the economy. . . . Our definition implies a distinction between generalized financial crisis on the one hand and bank failures, debt defaults and foreign-exchange market disturbances on the other” (1991, 10). Financial economists examine the micro behavior of market participants to explain disruptions in financial markets (see Diamond and Dybvig 1983; Chari and Jagannathan 1984). They have tended to focus on banking panics and runs and the reasons depositors withdraw funds rather than on the macro consequences for employment and output in the real economy per se. 4 While differing in their emphases, the micro and macro approaches to analyzing financial stability share several themes. The first focuses on alternative explanations for why a crisis occurs. One prominent thesis argues that the financial system is inherently unstable and is therefore vulnerable to random shocks. Shocks simultaneously cause market participants to lose confidence in the system and exchange their bank deposits for currency. Others believe that such herd behavior cannot be explained solely by shocks that, like animal spirits, randomly induce depositors to run from bank deposits to currency. They offer more behaviorally oriented explanations and models, the most prevalent being models based on the existence of information asymmetries between borrowers and lenders. These models attempt to show that it is sometimes rational for depositors to attempt to withdraw their funds in such a way that it creates a run on the banking system. Most of the analysis in the random shock and information asymmetries models concentrates on aggregate behavior, assuming essentially that all market actors— both depositors and institutions—are identical. It does not admit differences among depositors and institutions or even the presence of more than one institution in the financial system. When the analysis recognizes more Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 realistic features of market and financial structure, researchers are better able to examine the process by which a shock or problem in one part of the country or sector of the economy is transmitted to other sectors or the system as a whole. These transmission models, representing the second main theme in the literature, have not been the focus of much empirical work and tend to be relatively undeveloped. The third area investigates the causes of financial crises and their impact on the real economy. For example, do financial crises cause declines in real economic output, or are they instead manifestations of deeper problems in the real economy? What are the channels of transmission? Do deposit runs cause liquidity problems, which in turn induce contractions in lending, thereby affecting real output and production? The final area of analysis examines the role of government policies— both macro and micro—in generating financial crises as well as lessening their potential severity. The remainder of this article explores these issues in more depth. The discussion gives particular attention to the possible linkages between deposits and credit availability as the transmission mechanism for crises since runs on deposits and payments system disruptions are believed to be transmitted to the real economy through a credit channel. Random Shocks and Inherent Financial Instability t the macroeconomic level, models such as those proposed by Minsky (1982) and Kindleberger (1978) embody the claim that the banking system is inherently unstable. Minsky argued that a capitalist economy, and especially its banking system, is inherently unstable. Furthermore, this instability is endogenous, originating within the system itself. He defined instability as “a process in which rapid and accelerating changes in the prices of assets (both financial and capital) take place relative to the prices of current output” (1982, 13). Simply stated, Minsky assumed that during relatively stable times firms engage in balanced financing, by which he meant that cash flows are sufficient to cover principal and interest payments. However, as the economy grows and enters the expansion phase of the business cycle, firms begin to reach for profits, presumably because of management’s preference for short- A term gains. Firms start to leverage up, and banks, in particular, begin to shorten the maturity structure of their liabilities relative to their assets. Expanding returns by funding long-term investments with shortterm borrowing is driven by the desire to take advantage of an upward-sloping term structure with long-term interest rates exceeding short-term rates.3 This period of leveraging, which Minsky labels a period of speculative finance, is still one of relative stability. Cash flows from investment are still sufficient to cover principal payments as debts. This speculation ultimately degenerates into what Minsky calls a period of Ponzi finance, in which cash flows cover neither principal nor interest payments. Debt refunding requires new Do financial crises cause debt issuance, the proceeds of which are used declines in real economic to cover required interoutput, or are they instead est and principal debt manifestations of deeper payments. During this period, an exogenous problems in the real shock will result in a economy? collapse of both the financial system and the real economy. The shock, which can come from many different sources, serves as the trigger for collapse. Minsky was silent on the exact mechanisms by which this happens. Commenting on Kindleberger’s (1978) similar view, Schwartz observes that “those who regard banks as inherently unstable assume no connection between monetary policy and the price conditions under which economic agents make decisions. Proponents of inherent instability see a recurring historical pattern in which many bankers abandon conservative standards of asset management during business expansions only to be caught short when booms collapse. For them instability resides in economic agents. Benevolent government then comes to the rescue. This is the central thesis offered by Charles P. Kindleberger in his 1978 book” (1986, 11). Minsky puts forth certain stylized facts that would be observed, although they are not the outcome from any specified model.4 The first is that, during an expansion, credit expands at rates that exceed the growth of 1. For representative examples see Smith (1991), Kaufman (1995), Donaldson (1992), Bartholomew, Moe, and Whalen (1995), and Eichengreen and Portes (1991). See Benston and Kaufman (1995) for a review of the evidence on fragility. 2. Eichengreen and Portes (1991) require declines in real output for a true financial crisis to occur. 3. Before 1910, however, the most common yield curve in the United States was downward-sloping. 4. It is generally argued that the theory as put forth by Minsky is not a unified theory that yields testable hypotheses. See, for example, Sinai (1977), Lintner (1977), Mishkin (1991), and Schwartz (1986). Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 5 income or the capital stock. Second, interest rates and nominal asset prices must be increasing at a rapid rate. Third, debt maturities must become shorter, and, fourth, some exogenous shock must occur to cause a change in expectations. Finally, governments must fail to intervene in ways that cushion any asset reevaluations accompanying any changes in expectations.5 Bernanke and James (1991) suggest a different view of the causal relationship. For them, a precipitating force that could lead to a financial collapse is a deflation. Deflation adversely affects credit quality by reducing borrower equity cushions. When companies finally default, Micro random shock intermediaries become owners of illiquid, real models pay no particular assets. To reliquefy attention to the source, their balance sheets, or nature, of the random banks are induced to reduce lending and call shock that causes in loans. Those banks depositors to line up. that are unable to reliquefy fail, and, by implication, deposits will be destroyed. In this scenario, credit problems lead to a reduction in bank deposits, contracting the money supply. The chief distinction from the picture Minsky and Kindleberger paint is that Bernanke and James see banks as passive bystanders in the process. They are not required to take on more risk, nor do they have to misprice risk or adjust their balance sheets to take on more interest rate or maturity risk. The model also suggests that crises occur only during and after an exogenous shock has induced a deflation. Bernanke and James are careful to argue, however, that while deflation is a necessary condition for a crisis to occur, it is not sufficient. They highlight several aspects of banking structure that, if present, also help increase the likelihood that financial institutions would experience a crisis. These include (a) lack of branch banking, (b) universal banking and the commingling of banking and commerce, and (c) funding though short-term, foreign deposits. Thus, banking and financial structure can either mitigate or accentuate the likelihood that a financial crisis will result during a deflationary period. Unlike the macroeconomists’ models discussed, the random shock models of the financial economists, most closely associated with Diamond and Dybvig (1983), look more deeply at the structure of the deposit contract and the process by which it is redeemed.6 Because deposits are payable upon demand at par, they offer depositors nearly costless liquidity, provided that 6 not all depositors wish to withdraw their funds at the same time. With sequential servicing, in which depositors are treated on a first-come, first-served basis, depositors, especially if they are geographically dispersed, rationally know that not everyone can withdraw simultaneously. If bank loans are inherently not marketable, or cannot be easily liquefied, then at the first hint of potential trouble, it is rational for depositors to step to the head of the line rather than incur costs to determine whether and exactly when deposits will be paid. These micro random shock models pay no particular attention to the source, or nature, of the random shock that causes depositors to line up. Depositors just decide to run, and once they do, all depositors run. These models also do not consider the credit side of the balance sheet as a factor in crises, other than the fact that loans are less liquid than deposits so that banks cannot pay all claims in currency. Nonetheless, they make it easy to see that shocks affect depositors’ willingness to hold bank deposits, and, when that willingness is reduced, a contraction in credit follows as loans must be liquidated to meet the deposit-redemption demand. The Diamond and Dybvig model approximates the situation that prevailed in early U.S. history. Individual banks issued their own bank notes to the public, promising to redeem these notes at par for specie.7 Since note issues typically were not backed 100 percent by specie, periodic liquidity problems arose whenever noteholders became concerned that a bank might not be able to honor its redemption commitment and suspend convertibility of deposits into specie. Runs on individual banks and the system sometimes occurred, and these resulted, albeit infrequently, in cumulative contractions in the money stock.8 Suspension of convertibility of deposits into specie was a common way for early banks to deal with temporary liquidity problems. It often resulted, however, in a decline in purchasing power since the value of deposits declined. By shifting the cost of nonconvertability at least temporarily to the creditors (depositors) of the bank, they gave all liability holders an important incentive to worry about bank solvency. Diamond and Dybvig (1983) investigate the suspension of convertibility as one equilibrium solution to the problem of runs, but they do not consider the price level effects or how the costs of suspension of convertibility are distributed because their model has only a consumption good and no currency. Another weakness of their model is that there is only one bank in the system, and hence runs are on the banking system as a whole and involve flights to the currency rather than runs on one of many banks in the system.9 For these early banks, avoidance of runs meant maintaining public confidence. Depositors needed to Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 believe that the institution could convert notes into specie in sufficient amounts and would not need to suspend convertibility.10 Indeed, the first forms of public regulation designed to deal with the problems involving suspension of convertibility imposed reserve requirements specifying permissible ratios of notes to specie. The regulations sought to assure public confidence by requiring banks to engage in minimal maturity intermediation, maintain sufficient specie reserves, and have adequate capital and liquidity.11 Information Asymmetries Models he micro random shock models have been less than satisfying, both because they appear generally inconsistent with economic events, as will be discussed in the next section, and because many economists find it hard to believe that people randomly decide to run without some just cause rooted in economics. Recent modeling efforts have applied concepts of information asymmetries to derive conditions that might make it rational for depositors to engage in runs on banks. Under the information asymmetries models, banks are viewed as being “opaque” to depositors and thus costly for depositors to monitor. With imperfect and costly information, a type of Akerlof (1979) lemons model applies in which depositors have a great deal of difficulty distinguishing between healthy and unhealthy banks. Any shock or news event that might induce depositors to reassess their bank’s riskiness (in combination with the sequential servicing constraint) will cause depositors to assume that all banks are riskier than previously believed. Under these circumstances, it is more rational for depositors to withdraw funds than to seek out and evaluate costly information or risk los- T ing their funds by not withdrawing. In these models, as in the micro random shock models, the source of the shock is not specified, in that no particular cause is suggested for a failure. But usually it is hypothesized that the shock originates in credit markets and in releases of relevant news about bank asset quality. The model’s predictions are consistent with the view that shocks are more likely to result from disturbances in the real sector than from the default of a single borrower. Macroeconomists have articulated a form of this same asymmetric information hypothesis in attempting to counter the inherent instability arguments. As Schwartz describes it, “a widely held belief in the United States and the world financial community is that the default of major debtors—whether companies or municipalities or sovereign countries—could lead to bank failures that would precipitate a financial crisis. . . . A financial crisis is fuelled by fears that means of payment will be unobtainable at any price and, in a fractionalreserve banking system, leads to a scramble for highpowered money. It is precipitated by actions of the public that suddenly squeeze the reserves of the banking system. In a futile attempt to restore reserves, the banks may call loans, refuse to roll over existing loans, or resort to selling assets. . . . The essence of a financial crisis is that it is short lived, ending with a slackening of the public’s demand for additional currency” (1986, 11).12 Under this scenario, a banking crisis is precipitated by the failure of a major debtor, which induces a sudden shift in the public’s demand for currency. In turn, banks scramble for reserve assets by curtailing lending and selling assets. By implication the decline in lending and refusal to roll over existing credits leads to a decline in economic output. The process becomes systemic in that 5. Minsky argues that the ability to intervene is directly correlated with the size of government; and big government, with its revenue capacity, has the resources to support, through fiscal and monetary policies, a longer run-up of leverage. Also, through its lender-of-last-resort capabilities, it can soften the landing during an exogenous shock period by supporting a gradual rather than precipitous liquidation of assets. It thereby avoids the corresponding collapse of credit, bank failures, and destruction of the money supply. 6. For other examples of models in this mode see Haubrich and King (1984), Cone (1983), Jacklin (1987), Wallace (1988), Bhattacharya and Gale (1987), Smith (1991), and Chari (1989). 7. In the Diamond and Dybvig (1983) model there is really no nonbank money in circulation. Individuals deposit a real consumption good in the bank in exchange for a deposit or warehouse receipt. This consumption good is close, but not identical, to specie. In early U.S. banking, it was not uncommon for notes issued by out-of-area banks to trade at discounts, which reflected several factors, including transportation and transaction costs, lack of information on the issuing bank, and uncertainties about the creditworthiness of the issuing bank. This lack of par clearance in no way affected the ability of state bank notes to function as money. 8. For discussions of the evidence on runs see Kaufman (1988) and Gorton (1987). 9. Because of the way the model is constructed, runs necessarily have an adverse impact on the real economy. 10. For a discussion of these early bank runs see Kaufman (1988) or Bryant (1980). 11. Clearinghouses and other banks in the region often provided temporary credit to institutions experiencing liquidity problems (see Kaufman 1988). Kaufman (1994) notes that bank capital ratios were substantially higher during this period than they were after deposit insurance was introduced. 12. Although Schwartz articulates this view, she clearly does not believe it is correct or that the policies designed to protect against the events are appropriate. Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 7 problems in one or several major creditors raise questions about the quality of bank assets in general and induce the public to switch to holding currency. The hypothesis implies a direct linkage between increased demand for money and the availability of credit—and hence the ability to finance and maintain the real economy. The information the public perceives is not assumed to be bank specific; instead, it is the fact that the information concerns the quality of banking assets in the aggregate that increases the private sector’s demand for currency relative to deposits. The channel envisioned in this scenario results in banks calling in loans and building up liquidity to meet the public’s Recent modeling efforts desire for currency. have applied concepts of Empirically, three information asymmetries to elements are necessary for this view to hold. derive conditions that First, there must be a might make it rational for credit-related shock that depositors to engage in affects the public’s desire to hold currency runs on banks. relative to deposits. Second, this shock must induce a liquidation of deposits for currency by the public. Third, bank credit must contract. There are several important differences between the various random shock models and asymmetric information models. First, Minsky’s random shock model includes leveraging up of both bank and corporate balance sheets across the board, and, furthermore, it does not require an inflationary environment. Second, the collapse that results is not driven by runs forcing institutions to liquefy balance sheets to meet deposit withdrawals. Third, under this type of model financial institutions accommodate the leveraging up of balance sheets by underpricing credit risk. They also take on more interest rate and maturity risk by shortening the maturity structure of liabilities relative to assets. Fourth, no interdependence among either borrowers or lenders is necessary for a collapse to take place. Finally, the direction of causation, in terms of propagators of the crisis, appears to run through credit channels by eroding depository institution real equity values. Only if institutions fail is the money supply affected. In the random shock model of Diamond and Dybvig (1983), the crisis does not result from asset mispricing or from rational economic behavior but rather from an exogenous event. Since there is only one bank in the economy in this model, runs take the form of flights to currency (or more precisely, the consumption good) and not to other healthy banks. The panic is due solely 8 to the existence of the sequential servicing requirement discussed above and the fact that bank assets are not perfectly liquefiable. Like the micro random shock models, the asymmetric information models do not rest upon systematic ex ante asset mispricing or other problems of bank behavior. Changes in expectations and market assessment of bank asset quality, combined with the opaqueness of bank balance sheets and sequential servicing, make runs a rational customer response. Empirical Evidence on Systemic Risk hen examined carefully, many of these alternative explanations of panics and financial crises appear to overlap, differing only slightly in their details. Separating them empirically can therefore be very difficult. Empirical tests of various hypotheses about financial crises and panics have generally focused on the National Banking Era and the period of the Great Depression. The reason for studying these periods is that no broad-based panics have occurred since (in part because of the existence of federal deposit insurance and lender-of-last-resort actions followed by the Federal Reserve). In this section, the empirical evidence is examined to determine which of the models appear to be more consistent with the data. The question of whether this empirical work provides useful insights or is relevant today is a legitimate one, given the changes in financial structure and markets, the rise of technology, the proliferation of information, and the globalization of markets. This issue will be addressed in the next section. The Random Shock and Financial Fragility Hypothesis. Given the lack of precision in specifying the models, does the evidence suggest that one or more of the models may be correct? With respect to the macro models, critics of the Minsky financial fragility hypothesis argue that it does not yield testable hypotheses and is inconsistent with the data (see Sinai 1977; Lintner 1977; Mishkin 1991; Schwartz 1986). As mentioned previously, for the hypothesis to hold, a sequence of several factors must be present: debt burdens increasing faster than the growth of income or capital stock, interest rates and nominal asset prices increasing rapidly, debt maturities at depository institutions becoming shorter, an exogenous shock occurring to cause a change in expectations, and, finally, governments failing to intervene in ways that would provide a soft landing to any asset revaluation that must accompany the change in expectations. Unfortunately, data do not readily exist for examining a number of the conditions Minsky sets forward. As an alternative Table 1 lists the periods of economic recession with information on when panics took place and, where possible, what possible shocks may have W Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 TA B L E 1 P a n i c D a t e s a n d C a u s e s Panic Date October 1857 Business Cycle Peak and Trough Fall 1853– July 1857 Percentage (and Number) of National Bank Failures NA Prepanic Interest Rate Movement Rates fell until the recession began. Percent Change in Currency-toDeposits Ratio from Previous Year’s Average NA With failure of Ohio Life and Trust, reserves were pulled from New York City banks. First bank failures occurred in September. Several railroads went bankrupt in September, and major runs on New York banks in October culminated in specie suspensions in mid-October. 14.5 Crisis began when New York Warehouse and Security Co. failed on September 8. Other failures and suspensions followed: Kenyon, Cox & Co., Jay Cooke & Co., and Fisk & Hatch. Panic-selling on the New York Stock Exchange led to closing of the market for ten days. The initial failures appeared related to debt problems and problems with railroad bonds.a 8.8 On May 6 Marine National Bank failed. The Wall Street brokerage firm Grant and Ward was linked to a bank that failed on May 8. That failure was followed by a run on Metropolitan National Bank and suspension of several other banks. However, an inflow of foreign capital and the issuance of clearinghouse notes moderated the panic. It appeared the clearinghouse notes provided a signal to the market of bank solvency. 3.0 —— Spread on bonds did not widen until after the onset of the recession. September 1873 October 1873– March 1879 2.8 (56) Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 Rates rose about 5 percentage points in August, five months before the beginning of the recession. Spread on bonds did not rise until the month of the panic and did not rise prior to that. June 1884 March 1882– May 1885 0.9 (19) No obvious pattern preceded the panic. With the exception of a threemonth period, spreads on bonds declined steadily for two years prior to the panic. No panic March 1887– April 1888 0.4 (12) —— Possible Prepanic Exogeneous Shock (Continued on page 10) 9 10 TA B L E 1 P a n i c D a t e s a n d C a u s e s (cont.) Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 Panic Date November 1890 Business Cycle Peak and Trough July 1890– May 1891 Percentage (and Number) of National Bank Failures 0.4 (14) Prepanic Interest Rate Movement Rates did not rise appreciably until after the recession had begun. Percent Change in Currency-toDeposits Ratio from Previous Year’s Average 9.0 New York Stock Exchange prices began falling in early November. On November 11 Decker, Howell & Co. failed, involving the Bank of North America. On November 12 a stock broker failed, and on November 15 Baring Brothers failed in London. 16.0 On February 26 the Philadelphia & Reading Railroad went into receivership, and on May 4 National Cordage Co. failed and a stock market crash followed. New York banks weathered the situation until banks in the West and South experienced runs and began withdrawing reserves from New York City banks to meet liquidity needs. In August there was a general suspension of specie payments. National banks were reopened after examination and certification by the Comptroller of the Currency. 14.3 The period of 1895–96 was a mild, paniclike period. Although the New York Clearing House Association made emergency credits available in the form of loan certificates, none were used. 2.8 —— Spread on bonds was essentially flat for a year preceding the panic. May 1893 January 1893– June 1894 1.9 (74) Rates rose beginning in January 1892 approximately 2.5 percentage points in the seven months preceding the recession. Spread was essentially flat for more than one year preceding the recession. October 1896 December 1895– June 1897 1.6 (60) Rates rose only about 75 basis points in the three months preceding the peak of the expansion but rose substantially in the three months before the panic. Possible Prepanic Exogeneous Shock Spread on bonds did not widen until after the beginning of the recession and peaked just prior to the panic. No panic June 1899– December 1900 0.3 (12) —— No panic September 1902– August 1904 0.6 (28) —— –4.1 —— October 1907 May 1907– June 1908 0.3 (20) Rates were flat preceding the recession and rose only slightly thereafter. 11.5 Stock market declined in October. During the week of October 14, five New York members of the New York Clearing House Association and three outside banks required assistance. These banks had been used to finance speculation in copper-mining stocks. On October 12 Knickerbocker Trust Co. (third largest in New York City) began to experience clearing problems, and it suspended operations on October 22. Spread on bonds was essentially flat prior to the beginning of the recession. Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 No panic January 1910– January 1912 0.1 (10) —— –2.6 —— August 1914 January 1913– December 1914 0.4 (28) —— 10.4 This crisis was linked to problems in London and disruptions to payments on discount bills by foreign (European) borrowers. London stopped discounting foreign bills, and the effect was to disrupt New York banks, who were in a net debt position during the summer (as apparently was usual). New York banks were forced to remit gold, draining reserves. Both London and New York Stock Exchanges closed on July 31, and a panic threatened. a Sprague ([1910] 1968) notes that all the failures were due to criminal mismanagement or to neglect or violation of the National Banking Act and not to questions about bank solvency. Source: Gorton (1988) and Schwartz (1986), except for Panics of 1857 and 1893 (Mishkin 1991). 11 existed. Looking first at the timing of the panics relative to the peaks and troughs of the business cycles shows that in only one instance was there a panic before the peak of the business cycle. In most cases, the panic occurred anywhere from three to six months after the business cycle had peaked. Such long lags would seem to be logically inconsistent with Minsky’s view. Mishkin (1991) devotes considerable attention to the rate pattern and to risk premiums and their relationship to the onset of panics. In general, the spread between rates on high- and low-quality bonds rose before the panic began. However, these spreads generally widened after the recession started rather than before as the Minsky hypothesis would require.13 The Asymmetric Information Hypothesis versus Micro Random Shock Models. Gorton (1988), Mishkin (1991), Evidence suggests that and Donaldson (1992) specifically investigate recession, and not a the information asymtriggering bank failure, metric hypothesis in is the critical factor in detail. Examining the National Banking Era determining whether a and the post–Federal panic will occur. Reserve Era through 1933, Gorton models depositor behavior in terms of the currency/ deposit ratio. He poses two questions. First, if panics are random events, then is the model predicting a different currency/deposit ratio during panic periods than exists in other times? Second, are panics predictable in terms of movements of perceived risk? From these two questions Gorton suggests the following testable hypothesis: “Movements in variables predicting deposit riskiness cause panics just as such movements would be used to price such risk at all other times. This hypothesis links panics to occurrences of a threshold value of some variable predicting the riskiness of bank deposits” (1988, 751). Such predictive variables might be extreme seasonal fluctuations, unexpected failure of a large corporation (usually a financial corporation), or a major recession. A third question Gorton asks is whether certain predictors of risk stand out as important predictors of panics. Finding no evidence that panics are random events, he concludes that there is strong support for the asymmetric information hypothesis. Furthermore, panics appear to be predictable ex ante. Evidence also suggests that recession, and not a triggering bank failure, is the critical factor in determining whether a panic will occur. Gorton explains: “the recession hypothesis best explains what prior information is used by agents in 12 forming conditional expectations. Banks hold claims on firms and when firms begin to fail (a leading indicator of recession), depositors will reassess the riskiness of deposits” (1988, 778). In short, causation seems directed from the real sector to the financial sector rather than vice versa. Donaldson (1992) extends Gorton’s analysis using a somewhat different specification of the model and weekly data between 1867 and 1907 to determine whether panics are systematic and predictable events. Unlike Gorton, Donaldson rejects the conclusion that panics are systematic events and argues that the data are more consistent with the random shock model than the asymmetric information model.14 However, for the panics of 1914 and 1933 (which required expansion of the money supply during crisis periods), he finds behavioral patterns of earlier panics had been dampened. Given that the later panics followed the creation of the Federal Reserve in 1913 and passage of the AldrichVreeland Act of 1912, this finding suggests that government involvement to increase liquidity can truncate panic situations. He concludes that panics are therefore special events. But he also finds evidence that panics are more likely to occur when seasonal and cyclical factors are present. Mishkin (1991) formulates the asymmetric information hypothesis somewhat differently. He argues that key variables help to capture differences in depositor assessment of bank risk. In particular, during periods of financial distress high-quality firms will be less affected and lenders will have less uncertainty about the riskiness of such firms than they will have for low-quality firms. To the extent that these risks are priced, an important index of asymmetric information uncertainty should be captured by the spread between the rates on high- and low-quality bonds, by stock prices (as a measure of net worth and collateral value), and by interest rates (as a measure of agency costs and adverse selection). His analysis, like that of Gorton (1988), supports the information asymmetries hypothesis to the extent that the proxy variables are in fact good proxies. He concludes that most financial crisis periods begin with an increase in interest rates and a widening of the spread between high- and low-quality bonds and a decline in stock prices, rather than with a panic. “Furthermore,” Mishkin observes, “a financial panic was frequently immediately preceded by a major failure of a financial firm, which increased uncertainty in the marketplace” (1991, 97). He also asserts that the information hypothesis offers a better explanation than the macro theories of financial fragility for the pattern of rate spreads and stock market movements both before and after a panic as well as the panic’s likely occurrence. Finally, Park (1991) argues that the provision of bank-specific information overcame the information Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 asymmetries that played a role in runs on banks. In particular, by analyzing the panics of 1873, 1884, 1893, 1907, and 1933 he concludes that clearinghouse and government intervention were effective devices in settling panics but only when they provided information on bank-specific solvency.15 In the panic of 1884, a run was abated following certification of solvency by the Comptroller of the Currency and by subsequent extensions of clearinghouse certificates to Metropolitan National Bank, which was the bank suffering the greatest withdrawals. The panic of 1893 followed a long period of depression during which banks suffered prolonged periods of withdrawals of gold and uncertainty about U.S. adherence to the gold standard. Gold hoarding culminated in suspension of convertibility, and repeal of the Sherman Silver Act was promised by the president. Banks lifted the suspension of convertibility, and the runs stopped. Because no systematic attempt was made to release information on individual banks, public confidence in all banks remained low until the source of uncertainty—lack of confidence in U.S. maintenance of the gold standard—was removed. Park (1991) interprets the Comptroller of the Currency’s certification of individual bank solvency before their reopening following the panic of 1893 as the major information factor that quelled depositor uncertainty. In the panic of 1907, the problem began with runs on individual New York banks and trust companies that had been directly or indirectly associated with a failed attempt to corner the market in copper stocks. Only intervention by the New York Clearing House Association, which attested to the solvency of banks experiencing runs and provided financial assistance, resolved the situation. Again, release of firm-specific information appeared to have addressed the information asymmetries and helped stabilize the crisis.16 Unlike other cases, in the panic of 1907 runs did not affect all banks, and, indeed, some New York Clearing House member banks experienced reserve inflows (Park 1991). Transmission Models. Neither the basic random shock models nor the information asymmetry models specifically address the issue of which mechanisms transmit panics or financial crises through the economy. In fact, no models admit more than one institution, a condition that would be necessary to model customers simply transferring funds from an unhealthy to a healthy institution as distinct from retreating to currency.17 The models provide no information on what, if any, real impacts such funds transfers among banks have. Nor have the models addressed when depositors will run on one bank and when they will run on the entire system. Researchers have addressed the question of transmission mechanism more indirectly by attempting to generalize from the basic models. For example, Calomiris and Gorton (1991) maintain that it is the sequential servicing constraint imposed in the Diamond and Dybvig–type models that can induce banks to run on other banks. Such runs are especially likely when banks are geographically dispersed but are permitted to count interbank deposits as legal reserves, as under the National Banking system. Two other regulatory constraints—restrictions on branching and on the payment of interest on interbank deposits—have also been regarded as important.18 The structure of the National Banking system prior to creation of the Federal Reserve in 1913 added a further source of instability to the economy. Under that system, legal reserves for National Banks included not only cash in vault but also deposits in Reserve City and Central Reserve City banks. In such a fractionalreserve banking system that has pyramiding of reserves, a run on an individual bank could more easily have systemic, systemwide effects. Shocks originating in the countryside, for example, could induce country banks to 13. The exception is the panic of 1873. 14. As a robustness test, he also reruns the analysis using monthly data as Gorton does and gets similar results to those found by Gorton. He concludes that monthly data are too spaced out to provide a sharp test of the hypothesis. 15. A more complete test of the Gorton-Mishkin-Park hypothesis about information asymmetries would be provided by examining fund flows from individual solvent and insolvent institutions. Relying upon aggregate statistics can be only circumstantial, not conclusive. 16. The Roosevelt administration, following the declaration of the bank holiday on March 6, 1933, employed this same policy. 17. Smith (1991) does provide a model in which banks are permitted to hold funds at a Reserve City bank. Bhattacharya and Gale (1987) provide a model with geographically dispersed depositors and banks. Again, however, these models only look at the interdependence among banks through the interbank deposit markets. 18. See also the discussions in Haubrich (1990), Bordo (1986), and Williamson (1989). All emphasize the advantages over U.S. banks that banks in Canada and other countries that permitted branching had in weathering panics. Calomiris and Schweikart (1991) have explored in detail for the United States the effects that structure had on failure rates in different states with different branching statutes. They show that branch banks had both lower failure rates and in general paid lower premiums on their notes during the crisis of 1857 than banks in other parts of the country. Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 13 improve their liquidity positions by recalling interbank deposits from the Reserve City and Central Reserve City banks. Hence, panics were also endemic to the structure of the system as a whole, and it is clear how a panic or run in a rural region could blossom into a systemic crisis for healthy banks in Reserve Cities and Central Reserve Cities. Chari (1989) addressed this issue directly in considering a model of spatially separated banks. He argued that the most likely source of a shock that would cause country banks to withdraw reserve funds was seasonally related, with differences in currency demands rising significantly during planting and harvest times. Inferences about financial Calomiris and Gorton (1991) attemptcrises and systemic risk ed to determine specifidrawn from study of the cally whether panics banking situation during were transmitted from rural areas through the the National Banking Era National Banking sysand early 1900s are not tem, as the analysis particularly meaningful or suggested, and also whether the patterns relevant in today’s economwere more consistent ic environment. with the random shock or information asymmetries models. They found that three important differences between the models have empirical implications. The first concerns the origin of problems. The random shock model suggests that shocks would occur in rural areas because of seasonal demands for currency. In contrast, the asymmetric information model implies that adverse economic news related to asset-quality problems would precede a panic. Second, the two theories would seem to predict different patterns of failures during a crisis. The asymmetric information model suggests that banks whose asset portfolios were closely linked to the specific shock would be more prone to failure whereas the random shock model would predict that failures would be experienced in the areas suffering currency withdrawals. Finally, the models differ in the conditions required to resolve a crisis. In the random shock model, the key to resolving a panic is liquefication of assets. In the asymmetric information model, it is the effectiveness of mechanisms initiated to resolve depositor uncertainty about bank solvency. Calomiris and Gorton’s exhaustive investigation of the sources of panics between 1873 and 1907 led them to reject the idea that seasonal money-demand shocks were the cause of banking panics. Rather, their analysis suggests that panics originated in bad economic news and bank vulnerability to that news. Moreover, their 14 inspection of failure patterns shows virtually no support for the random shock model. Finally, they conclude that in terms of resolving crises, the mere availability of currency, which would provide the ability to liquefy assets, was not sufficient to stop panics during the periods studied. Again, this conclusion suggests that the asymmetric information model was more consistent with the data than was the random shock model. Smith (1991) provides some specific evidence on country banks’ behavior vis-à-vis their holdings of cash reserves as compared with reserves held in the form of interbank deposits when panics occurred. He provides analysis of some anecdotal and other evidence, derived mostly from Sprague ([1910] 1968), about the behavior of Reserve City banks during the crises of 1873, 1893, 1907, and 1930–33. Smith describes the situation leading up to the panic of 1873, indicating that interbank deposits were concentrated in seven of the New York City banks. These interbank deposits constituted about 45 percent of the sources of funds for the New York banks and were the base upon which their bond holdings and loans were built. These banks were clearly vulnerable to demands by country banks for withdrawal of reserves, and funds were especially tight in the few months before the crisis. When the key triggering events occurred (see Table 1), a combination of circumstances made the crisis severe. In addition to having virtually no excess reserves, several of the banks were in weak financial condition. As subsequent events would prove, several had been the victims of fraud and defalcations, probably accounting in part for their financial weakness. Clearly, however, the institutions’ problems stemmed primarily from reserve withdrawals and their inability to call in loans in that economic environment rather than from major credit problems in the New York Central Reserve City banks. The Reserve City banks experienced similar problems caused by currency outflows during the panics of 1893 and 1907. Thus, it seems clear that reserve outflows, coupled with the lack of excess currency reserves at the Central Reserve City banks in New York and Chicago, forced contractions in loans and finally resulted in the suspension of currency payments. Smith notes that currency suspension was the prime transmission mechanism of panics once a triggering mechanism occurred. He also concludes that the problems during the 1930–33 period originated in the rural agricultural areas as well and were intimately intertwined with the correspondent banking system. Despite a fairly clear pattern in the transmission mechanism of panics emanating from large reservedeposit withdrawals (rather than from uncertainties about credit quality in Reserve City banks, as the Minsky hypothesis would imply), a number of questions remain. For example, Tallman (1988) indicates that Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 looking at the data over longer time periods does not suggest a clear linkage between the incidence of panics and either increases in currency demand relative to deposits or contractions in loans. He presents evidence that loan contractions occurred at several intervals during the period between 1893 and 1907, for example, that exceeded the declines during periods when panics occurred. Similarly, during some periods of time between 1873 and 1930 the number of bank failures far exceeded those observed during panic periods. Finally, Tallman provides aggregate data on the growth in loans relative to high-powered money and on the growth of manufacturing output between 1873 and 1914. Two observations are important. First, loans do increase in the years prior to panic periods, but the panics occur after loan growth has fallen significantly. Second, numerous periods during the interval show the same patterns in loan and output growth and decline but are not accompanied by a panic. These aggregate data do not reveal whether there are differences in the loancontraction periods in terms of their concentration in particular parts of the country during episodes of panic and nonpanic periods. Causal Direction. The research evidence seems to indicate fairly consistently that the dynamics between financial panics and changes in real economic output begin in the real sector and move to the financial sector rather than starting in the financial sector. There are no examples in U.S. history of the economy operating at high levels of output when a financial crisis occurred that resulted in a contraction in the real economy. As the discussion in the previous sections suggests, however, banks were sometimes under pressure and were forced to call in loans. It seems reasonable to assume that once problems in the financial sector become severe, there could be negative feedback effects to the real sector. Indeed, Bernanke (1983) has made precisely this point. Financial crises can have real effects outside the normal reserve/loan transmission mechanism because of the disruptions to the intermediation process. Bank failures disrupt borrower/lender relationships and make attaining financing more difficult and costly. But this observation should not obscure the fact that financial crises are better viewed as creatures of recession and economic downturns rather than primary causal agents precipitating the downturns. The Role of Government. A substantial body of evidence indicates that government actions have played significant roles in contributing to crises as well as in mitigating them. For example, Sprague ([1910] 1968) notes that lack of access to a reliable lender of last resort to provide short-term liquidity can help escalate a period of financial tightness into one of crisis. Friedman and Schwartz (1963) argue that several Federal Reserve actions during the Great Depression contributed to both its duration and magnitude. For instance, they observe that the Federal Reserve’s failure to liquefy the assets of many small nonmember institutions (the Fed was not obligated to lend to nonmember banks), together with its insistence that it would lend only upon sound collateral, added to the number of bank failures. This policy, in conjunction with the Fed’s attempt to adhere to the rules of the gold standard, contributed to a 33 percent decline in the money supply and clearly exacerbated the severity of the recession. While it has become fashionable to criticize the Fed for its policy failures during the Great Depression, it is also the case that government interference affected financial soundness long before the Fed was created. For example, during the National Banking Era the monetary base was tied, except for a period of suspension, to gold and silver specie monies through the Treasury.19 When the United States adhered to the gold standard, fluctuations in the gold supply expanded and contracted the monetary base, directly affecting banks’ lending behavior. Decisions about how much in the way of international gold flows would be permitted before conversion could be suspended was a matter of Treasury and government policy. European central banks, and to a lesser extent the U.S. Treasury, often intervened to prevent loss of gold reserves by raising short-term interest rates. Government policies frequently exacerbated gold flows and, by implication, induced fluctuations in the monetary base. For example, following passage of the Sherman Silver Purchase Act in 1890, foreigners’ concern that the United States would remain on the standard precipitated gold outflows and contributed to the panic of 1890.20 Tallman and Moen note that each panic after 1897 was preceded by unusual gold flows. They conclude that political uncertainties concerning the U.S. commitment to the gold standard were important influences on gold flows and, hence, the U.S. monetary base. Political conditions outside the United States also affected gold flows. For example, in 1907 the Bank of England responded to problems in the London money markets by raising its discount rate to stem potential speculative 19. Specifically, the monetary base included gold coin, gold certificates backed 100 percent by gold, silver dollars, silver certificates, other small silver coins, U.S. notes and other Treasury fiat, and national bank notes. See Tallman and Moen (1993). 20. Tallman and Moen (1993) indicate that this uncertainty was greatly reduced with the discovery of large gold supplies in the late 1890s. Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 15 outflows of gold to the United States. At that time London was the most important market for discounting U.S. trade bills. The increase in the discount rate not only disrupted the flow of gold to the United States but also discouraged the discounting of trade bills and caused a liquidity crisis in the United States. Since the debacle of the Great Depression, U.S. intervention in markets has often had as its objective providing liquidity to avoid a crisis. Numerous examples exist of emergency liquidity having been provided through the efforts of the Federal Reserve either directly or indirectly, such as during the Penn Central scare, the Chrysler problem, the collapse of Drexel-Burnham, and the failure of Continental Illinois Bank, to name just a few. The Federal ReWhile panics do appear to serve has on occasion be associated with recesattempted to provide liquidity not only to sions and deflationary cushion problems in periods, the direction of interbank markets but causation seems to run also to prevent disruptions in other markets. from the real sector to the financial sector rather than the other way around. Relevance in Today’s World t can be argued that inferences about financial crises and systemic risk drawn from study of the banking situation during the National Banking Era and early 1900s are not particularly meaningful or relevant in today’s economic environment. Pyramiding of legal reserves in private banks is not a structural feature of the present reserve requirement regime. Markets are no longer isolated, and informationavailability problems that might have resulted in the past in information asymmetries have been reduced significantly. Communications technology and new instruments have increased the liquidity of all banking assets and have given rise to new markets that make the kinds of liquidity crises that occurred in the National Banking Era unlikely today. Furthermore, the United States has abandoned the gold standard, and thus the domestic money supply is not subject to the random fluctuations and shocks that it was vulnerable to under strict adherence to the gold standard rules. Deposit rate ceilings of the 1930s have been phased out, and branching restrictions, which essentially prevented institutions from achieving geographical diversification, are a thing of the past. Certainly the focus on protecting the money supply from sudden shocks is no longer of prime policy concern for three reasons. First, it seems unlikely that significant runs to currency will occur (see Kaufman 1988). Deposits still have large advantages over currency for a I 16 variety of purposes, and there are many banks to chose from. Runs on individual banks would simply transfer reserves from one institution to another. Second, Federal Reserve policy is likely to provide emergency liquidity to prevent such runs from disrupting other institutions. Finally, while still accounting for the bulk of payment items, checks and currency are no longer the dominant forms in terms of dollar volume of transactions in the economy. The concerns and risks have shifted to other sectors of the payments system that did not exist during the National Banking Era. Today, the payments system is larger, has many more components (both private and public), and is subject to different risks than in the past. The check/ demand deposit system, which accounts for the bulk of individual payments (except for currency), and the one that the present regulatory structure was primarily designed to protect, is small in terms of the dollar volume of payments. The rest are made in the form of computerized transfers of reserve balances on the Federal Reserve’s Fedwire system and the privately owned Clearing House Interbank Payments System (CHIPS) and in the form of automated clearinghouse (ACH) transactions. Payments on the former two systems account for about 85 percent of the dollar value of transactions. Closely related to these systems are the automated transfers of book-entry Treasury securities, which also take place on Fedwire and involve substantial volumes of transactions. Finally, as markets have become increasingly global, timing differences and differences in clearing and settlement conventions can add temporal and other dimensions to credit risks not always found in the domestic markets that characterized earlier times. Many other significant sources of uncertainty can also be identified in the clearing and settlement processes in modern financial markets (see, for example, Eisenbeis 1997 and McAndrews 1997). Maintaining the integrity of payment flows is a substantially more complicated and difficult problem today than protecting the stock of demand deposits for a number of reasons. First, given the large size of transactions in the system and the size of the system itself, the resources required to support unwinding even a shortrun problem may be enormous and could exceed the capacity of private participants to self-insure. Second, because the transactions are electronic and occur instantaneously, monitoring them and the net position of each participant is critical to controlling participants’ credit risk exposure. Third, when the international activities of U.S. banks and the links between the U.S. domestic payments system and foreign banking organizations are recognized, it becomes difficult to conceive of ensuring domestic financial stability without also ensuring international financial stability. Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 Clearly, different types of uncertainties exist with respect to systemic risk exposures today than existed in the past. There is also reason to believe that liquidity problems for borrowers may be significantly different than they were for borrowers in the 1800s. The growth of new mortgage lending instruments and, particularly, the development of home equity lines of credit provide ways for borrowers instantaneously to liquefy previously illiquid assets during tight times. While this ability to liquefy assets more easily may enable borrowers to maintain payments on outstanding debts and lessen the severity of the credit component of a recession, it also suggests introduction of a new discontinuity that might systematically transfer risks to the banking system at a critical trigger point. If during times of financial distress borrowers draw down lines on home equity and similar lines of credit and are then forced into default, the burdens of these defaults will be shifted to the providers of the home equity lines. Should these losses be large, capital might be wiped out, with few options available to lenders to avoid the costs of those defaults. Examples of similar impacts in commercial and real estate lending markets occurred when commercial paper borrowers drew down banks’ back-up commitments during the Penn Central and Real Estate Investment Trust (REIT) crises. Summary and Conclusions his article has investigated the various theories of financial panics and crises with particular emphasis on the links between credit and deposits. The survey suggests that panics are not random events, as some of the theories may suggest, but neither are they perfectly predictable. Nevertheless, it does appear that information asymmetries about the ability to liquefy deposits were a major contributing factor to banking panics in the past. Moreover, while panics do appear to be associated with recessions and deflationary periods, the direction of causation seems to run from the real sector to the financial sector rather than the other way around. It is not that financial crises cannot exacerbate economic declines; rather, they are not primary causal agents of recessions. The analysis also suggests that government policies can affect the likelihood of a financial crisis as well as play a role in its solution. These considerations are as relevant today as they have been historically. At the same time, the article raises a cautionary note that the dynamics of crises and how they might play out may be significantly different in the future given recent, rapidly developing changes in the U.S. and world financial system. T Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 17 REFERENCES AKERLOF, GEORGE. 1979. “The Market for Lemons: Qualitative Uncertainty and the Market Mechanism.” Quarterly Journal of Economics 84:488–500. DONALDSON, R. GLEN. 1992. “Sources of Panics: Evidence from Weekly Data.” Journal of Monetary Economics 30 (November): 277–305. BARTHOLOMEW, PHILIP F., LARRY R. MOE, AND GARY WHALEN. 1995. “The Definition of Systemic Risk.” Office of the Comptroller of the Currency. Presented at the seventieth annual Western Economic Association International Conference, San Diego, California, July. EICHENGREEN, BARRY, AND RICHARD PORTES. 1991. “The Anatomy of Financial Crises.” In Financial Markets and Financial Crises, edited by R. Glenn Hubbard, 10–58. Chicago: University of Chicago Press. BENSTON, GEORGE J., AND GEORGE G. KAUFMAN. 1995. “Is the Banking and Payments System Fragile?” Journal of Financial Services Research 9 (December): 209–40. BERNANKE, BEN S. 1983. “Non-Monetary Effects of the Financial Crisis in the Propagation of the Great Depression.” American Economic Review 73 (June): 14–31. BERNANKE, BEN, AND HAROLD JAMES. 1991. “The Gold Standard, Deflation, and Financial Crisis in the Great Depression: An International Comparison.” In Financial Markets and Financial Crises, edited by R. Glenn Hubbard, 33–68. Chicago: University of Chicago Press. EISENBEIS, ROBERT A. 1997. “International Settlements: A New Source of Systemic Risk?” Federal Reserve Bank of Atlanta Economic Review 82 (Second Quarter): 44–50. FRIEDMAN, MILTON, AND ANNA J. SCHWARTZ. 1963. A Monetary History of the United States, 1867–1960. Princeton, N.J.: Princeton University Press. GORTON, GARY. 1987. “Banking Panics and Business Cycles: Data Sources, Data Construction, and Further Results.” The Wharton School, University of Pennsylvania. Photocopy. ———. 1988. “Banking Panics and Business Cycles.” Oxford Economic Papers 40:751–81. BHATTACHARYA, SUDIPTO, AND DOUGLAS GALE. 1987. “Preference Shocks, Liquidity, and Central Bank Policy.” In New Approaches to Monetary Economics, edited by William Barnett and Kenneth Singleton, 69–88. New York: Cambridge University Press. HAUBRICH, JOSEPH. 1990. “Non-Monetary Effects of Financial Crises: Lessons from the Great Depression in Canada.” Journal of Monetary Economics 25 (March): 223–52. BORDO, MICHAEL. 1986. “Financial Crises, Banking Crises, Stock Market Crashes, and the Money Supply: Some International Evidence.” In Financial Crises and the World Banking System, edited by Forest Capie and Geoffrey E. Wood, 190–248. London: Macmillan. JACKLIN, CHARLES. 1987. “Demand Deposits, Trading Restrictions, and Risk Sharing.” In Contractual Arrangements for Intertemporal Trade, edited by Edward D. Prescott and Neil Wallace, 26–47. Minneapolis: University of Minnesota Press. BRYANT, JOHN. 1980. “A Model of Reserves, Bank Runs, and Deposit Insurance.” Journal of Banking and Finance 4:335–44. KAUFMAN, GEORGE G. 1988. “Bank Runs: Causes, Benefits, and Costs.” Cato Journal 7 (Winter): 559–87. CALOMIRIS, CHARLES W., AND GARY GORTON. 1991. “The Origins of Banking Panics: Models, Facts, and Bank Regulation.” In Financial Markets and Financial Crises, edited by R. Glenn Hubbard, 109–73. Chicago: University of Chicago Press. HAUBRICH, JOSEPH, AND ROBERT KING. 1984. “Banking and Insurance.” NBER Research Working Paper No. 1312. ———. 1994. “Bank Contagion: A Review of the Theory and Evidence.” Journal of Financial Services Research 8:123–50. ———. 1995. “Comment on Systemic Risk.” In Research in Financial Services, vol. 7, edited by George G. Kaufman. Greenwich, Conn.: JAI Press. CALOMIRIS, CHARLES W., AND LARRY SCHWEIKART. 1991. “The Panic of 1857: Origins, Transmission, and Containment.” Journal of Economic History 51:807–34. KINDLEBERGER, CHARLES P. 1978. Manias, Panics, and Crashes. New York: Basic Books. CHARI, V.V. 1989. “Banking without Deposit Insurance or Bank Panics: Lessons from a Model of the U.S. National Banking System.” Federal Reserve Bank of Minneapolis Quarterly Review 13 (Summer): 3–19. LINTNER, JOHN. 1977. “Discussion.” In Financial Crises: Institutions and Markets in a Fragile Environment, edited by Edward I. Altman and Arnold W. Sametz, 204–207. New York: John Wiley and Sons. CHARI, V.V., AND RAVI JAGANNATHAN. 1984. “Banking Panics, Information, and Rational Expectations Equilibrium.” Northwestern University. Photocopy. MCANDREWS, JAMES J. 1997. “Banking and Payment System Stability in an Electronic Money World.” Federal Reserve Bank of Philadelphia Working Paper No. 97-9. CONE, KENNETH. 1983. “Regulation of Depository Institutions.” Ph.D. diss., Stanford University. MINSKY, HYMAN P. 1982. “The Financial-Instability Hypothesis: Capitalist Processes and the Behavior of the Economy.” In Financial Crises: Theory, History, and Policy, edited by Charles P. Kindleberger and Jean-Pierre Laffargue, 13–38. Cambridge: Cambridge University Press. DIAMOND, DOUGLAS, AND PHILIP DYBVIG. 1983. “Bank Runs, Liquidity, and Deposit Insurance.” Journal of Political Economy 91 (June): 401–19. 18 Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 MISHKIN, FREDERIC S. 1991. “Asymmetric Information and Financial Crises: A Historical Perspective.” In Financial Markets and Financial Crises, edited by R. Glenn Hubbard, 69–108. Chicago: University of Chicago Press. SPRAGUE, O.M.W. [1910] 1968. History of Crises under the National Banking System. National Monetary Commission. Reprint. New York: Augustus M. Kelley. PARK, SANGKYN. 1991. “Bank Failure Contagion in Historical Perspective.” Journal of Monetary Economics 28 (October): 271–86. SCHWARTZ, ANNA. 1986. “Real and Pseudo Financial Crises.” In Financial Crises and the World Banking System, edited by Forest Capie and Geoffrey E. Wood, 11–31. London: Macmillan. SINAI, ALLEN. 1977. “Discussion.” In Financial Crises: Institutions and Markets in a Fragile Environment, edited by Edward I. Altman and Arnold W. Sametz, 187–203. New York: John Wiley and Sons. SMITH, BRUCE D. 1991. “Bank Panics, Suspensions, and Geography: Some Notes on the ‘Contagion of Fear’ in Banking.” Economic Inquiry 29 (April): 230–48. TALLMAN, ELLIS W. 1988. “Some Unanswered Questions about Bank Panics.” Federal Reserve Bank of Atlanta Economic Review 73 (November/December): 2–21. TALLMAN, ELLIS W., AND JON MOEN. 1993. “Liquidity Shocks and Financial Crises during the National Banking Era.” Federal Reserve Bank of Atlanta Working Paper 93-10, August. WALLACE, NEIL. 1988. “Another Attempt to Explain an Illiquid Banking System: The Diamond and Dybvig Model with Sequential Service Taken Seriously.” Federal Reserve Bank of Minneapolis Quarterly Review 12 (Fall): 3–16. WILLIAMSON, STEVEN. 1989. “Bank Failures, Financial Restrictions, and Aggregate Fluctuations: Canada and the United States, 1870–1913.” Federal Reserve Bank of Minneapolis Quarterly Review 13 (Summer): 20–40. Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 19 Decision Time for European Monetary Union J O S E P H A . W H I T T J R . The author is an economist in the macropolicy section of the Atlanta Fed’s research department. He thanks Peter Abken, Roberto Chang, and Mary Rosenbaum for helpful comments and Mike Chriszt and Jeff Johnson for research assistance. T RAVELERS IN THE UNITED STATES TAKE FOR GRANTED THEIR ABILITY TO USE THE SAME DOL- LAR BILLS TO PAY FOR MEALS, TAXIS, AND OTHER GOODS AND SERVICES THROUGHOUT THE NATION WHETHER THEY ARE IN HOWEVER, IN EUROPE NEW YORK, LOS ANGELES, THE SITUATION IS QUITE DIFFERENT. OR ANYPLACE IN BETWEEN. EVEN FAIRLY SHORT TRIPS OFTEN INVOLVE TRAVELING THROUGH MORE THAN ONE COUNTRY, AND, EACH TIME A BORDER IS CROSSED, TRAVELERS MUST USE COMPLETELY DIFFERENT CURRENCY AND COINS. This situation may change dramatically in the next few years. If the plans of European governments for economic and monetary union (EMU) are realized, within five years a new common currency called the euro will replace the money currently in use in at least a few western European countries. A traveler going to Paris, Amsterdam, Berlin, and Rome might be able to use euros in all four places. Even earlier, starting in 1999, a new European Central Bank is slated to take control of monetary policy in the initial member countries. At that time, exchange rates between the initial members will be fixed permanently. Eventually, in two or three decades, the euro may be in use throughout most of western and central Europe, from Ireland to Greece and from Portugal to Finland. The choice of initial members in the monetary union is scheduled to be made early in 1998, but as of this writing major hurdles remain that could delay or possibly kill the whole plan. The biggest stumbling 20 block involves budget deficits. To be eligible to join the proposed monetary union, countries are supposed to have budget deficits of no more than 3 percent of gross domestic product (GDP) in 1997, but it now appears likely that many prospective members, including the largest, Germany, will violate that limit. This article examines the economic and political factors that will determine whether monetary union proceeds on schedule and, if so, which countries will be initial members. The first section provides background and lays out the current official timetable for monetary union. The second section reviews the convergence criteria to be used in determining which countries are ready to join, with special emphasis on the fiscal or budget deficit criterion that is proving to be the biggest problem. Because so many countries are in danger of failing to satisfy all the convergence criteria, the third section describes several quite different scenarios that even at this late date are still under consideration, especially in Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 unofficial forums. The final decision on EMU will be made through political bargaining among the leaders of the prospective member countries. The next section outlines the results of economic models of such a bargaining process. The prospect of monetary union is already affecting financial markets. The final part of the discussion shows how financial market data can be used to try to infer the market’s assessment of the likelihood that certain countries will enter monetary union. Such estimates are no doubt imprecise, but as of early 1997 the pattern of market interest rates appeared to embody a substantial likelihood that a widespread monetary union will begin operation in the next few years. Current Timetable for Monetary Union t the beginning of 1999, Europe is scheduled to begin a major experiment in monetary arrangements. The Maastricht Treaty, which was signed by the members of the European Union (EU) in 1991, provides for economic and monetary union and creation of a European central bank by the end of this decade.1 In many respects this undertaking is highly unusual; we are accustomed to thinking of each nation as having its own government, its own money, and its own central bank, which is either directly or indirectly a part of the government. The Maastricht framework would create both a new money that would be legal tender in all participating countries and a central bank that would not be an agency of any one government. The Maastricht Treaty specified that monetary union between those countries that were ready would begin on January 1, 1999.2 On that date the new European Central Bank would begin carrying out monetary policy in the uniting countries, and exchange rates between their individual currencies would be fixed permanently. The countries that joined the monetary union initially would continue to use their national currencies for a time, but their bilateral exchange rates would be fixed irrevocably and their monetary policy would be set by the new European Central Bank, which is modeled on the German Bundesbank and is supposed to carry out monetary policy with the aim of ensuring price stability. By January 2002, notes and coins denominated in the new monetary unit, the euro, would be put into circulation, and after a short time the old national currencies and coins would be withdrawn from circulation. At A that point, the euro would be the single currency in circulation throughout the monetary union. The Maastricht Treaty was not approved merely to make life easier for travelers. To some extent monetary union was just one part of a more general move toward closer economic integration of the EU that also included the Single European Act or Europe 1992, which called for the elimination of many regulatory barriers to the free flow of goods, capital, and workers within the EU. European leaders hoped that greater economic integration would help rejuvenate their economies, many of which were plagued by high unemployment. Moreover, a larger, more integrated Europe would, it was hoped, benefit from economies of scale and be able to compete more effectively against economic Many preparatory steps rivals such as the United States and Japan. have been taken, but a The decision to disnumber of key decisions, mantle restrictions on notably about which councapital flows gave particular impetus to montries will be part of the etary union because union at its beginning, still free movement of capiremain to be made. tal is incompatible with fixed (or managed) exchange rate systems such as the European Monetary System (EMS) and national autonomy in the formation and implementation of monetary policy (see PadoaSchioppa 1994). When the proposals that became the Maastricht Treaty were under discussion in the late 1980s and early 1990s, the exchange rates of many of the European countries were already linked in the EMS. Fluctuations of each member’s exchange rate relative to other members were limited to ranges defined by fairly narrow target bands. If one country—for example, Belgium—tried to exercise national autonomy in its monetary policy by lowering its interest rates significantly below those of other members, capital outflows could become so large that they would push Belgium’s exchange rates to, if not beyond, the limits of the target bands. Free movement of capital would make the exchange rate target bands even harder to maintain. To proponents of monetary union (for example, Sutherland 1997), Europe would need to move ahead to monetary 1.When the Maastricht Treaty was signed, the EU had twelve members: Belgium, Denmark, France, Germany, Greece, Ireland, Italy, Luxembourg, the Netherlands, Portugal, Spain, and the United Kingdom. Austria, Finland, and Sweden were added in 1995. For reviews of the literature on European monetary union, see Bean (1992), Kenen (1992), and Eichengreen (1993). 2.The treaty envisioned an earlier start-up date for monetary union, at the beginning of 1997, but only if a majority of members were ready in time. If a majority were not ready in time for that earlier date (as actually occurred), then the treaty specifies the currently planned start-up date at the beginning of 1999, with no requirement that a majority be ready. Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 21 union, or the attempt to create a single European market (including free movement of capital) would fail. Another reason for monetary union would be the resulting reduction in transactions costs on intraEuropean trade. Travelers would no longer have to exchange one money for another each time they crossed a national border. Over time, all the workers, computers, and other equipment currently used simply to convert one European money into another could be redeployed. The European Commission (1990) has estimated the gain from lower transactions costs to be modest, 0.3 to 0.4 percent of GDP every year, with the largest proportional gains going to countries with relatively unsophisticated and inefficient financial sectors. Political considerations also played an important role. Countries such as France may have regarded monetary union as a way of gaining greater influence over their own monetary policy, as compared with the existing EMS, which is often interpreted as being dominated by Germany.3 For its part, Germany may have supported monetary union in exchange for benefits on other issues, notably the acquiescence of its European neighbors in its rapid unification with the former East Germany (see Garrett 1994 and Woolley 1994). While political considerations in some countries may have favored monetary union, critics such as Feldstein (1992) and Dornbusch (1996) argue that, from an economic perspective, monetary union would be a mistake for Europe. In their view, a country hit by a decline in worldwide demand for its output has two main ways of adjusting. One involves reducing relative prices and wages in the country affected. However, given the rigidities in European labor markets, a reduction in wages relative to those in other countries might occur only after a long and painful period of recession. The other way involves offsetting changes in economic policy, such as a loosening of monetary policy or a depreciation of the exchange rate that would reduce the need for nominal wage reductions. As far as individual countries are concerned, monetary union would eliminate the possibility of changing monetary policy or the exchange rate, forcing adjustment back primarily onto the labor market. From this perspective, the economic recoveries that occurred in Italy and Britain after those two countries allowed their currencies to depreciate in 1992 illustrate the advantages of retaining separate currencies and autonomy in economic policymaking. T A B L E 1 Maastricht Convergence Criteria, 1996 Country Criteria Annual Inflation Rate (Percent) Long-Term Interest Rate (Yield) Government Budget Deficita Government Debta 3.0 60.0 2.6 8.9 Austria 4.3 71.7 1.7 6.5 Belgium 3.3 130.6 1.6 6.7 Denmark 1.4 70.2 2.2 7.4 Finland 3.3 61.3 0.9 7.4 France 4.0 56.4 2.1 6.6 Germany 4.0 60.8 1.3 6.3 Greece 7.9 110.6 8.4 15.1 Ireland 1.6 74.7 2.1 7.5 Italy 6.6 123.4 4.7 10.3 +0.9 7.8 1.3 7.0 The Netherlands 2.6 78.7 1.2 6.3 Portugal 4.0 71.1 3.0 9.4 Luxembourg Spain 4.4 67.8 3.8 9.5 Sweden 3.9 78.1 1.6 8.5 United Kingdom 4.6 56.3 3.0 8.0 a As a percentage of GDP Source: European Commission (1997) 22 Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 Other economists counter that the absence of monetary union can itself be a source of problems. Buiter (1996) argues that if exchange rates are flexible, financial shocks will move them and result in temporary changes in international relative prices and wages that are not required by the underlying real fundamentals and that have negative effects on economic performance. In a similar vein, Obstfeld (1996) argues that as long as a discretionary devaluation is possible, a country has more than one economic equilibrium, some of which are worse for the country than a permanently fixed exchange rate. He suggests that Italy was in such a “bad equilibrium” prior to its devaluation in 1992, with unemployment and real interest rates both at suboptimally high levels. Regardless of their motivations, European policymakers no doubt were hoping for a smooth transition to monetary union. The policymakers who signed the Maastricht Treaty in December 1991 probably believed that with minor exceptions, EMS exchange rates would be kept within the existing target bands until monetary union was achieved; in effect, monetary union would simply shrink the width of the bands down to zero. However, those expectations of exchange market tranquility were dashed just a few months after the signing. The British pound and Italian lira came under enormous pressure as investors bet that those two governments would not maintain their exchange rates in the face of domestic economic weakness. In September 1992 both currencies dropped out of the EMS and soon fell well below their previous values. The following year further speculative pressures on other currencies led to a substantial widening of most of the EMS exchange rate bands from plus or minus 21⁄4 percent to plus or minus 15 percent. At that point, prospects for achieving monetary union on the Maastricht schedule looked dim. However, since the crisis of 1993, exchange markets within Europe have been generally stable. In the case of members of the EMS the 15 percent bands allow for fairly large exchange rate movements, but most of the time central banks have succeeded in keeping actual exchange rates within much narrower boundaries. For example, throughout 1996 the French franc was kept within or very close to the narrow pre-1993 target band, even though officially the wide bands were in effect. As of this writing, with the exchange markets fairly tranquil and political leaders in key countries, notably France and Germany, still publicly committed, the odds on monetary union starting up in 1999 have improved considerably from what they were during the crisis year of 1993. Nevertheless, major issues remain unresolved. One of the major unresolved issues is the question of which countries will be part of the initial monetary union. The current plan is for the political leaders of the fifteen countries in the European Union to meet early in 1998 in order to make this decision. At that time, each country’s economic data for 1997 should be available and could be compared with the convergence criteria in the treaty that provide guidelines for assessing a country’s readiness for monetary union. The Convergence Criteria hen the treaty was signed, economic conditions in the various EU members differed substantially. The treaty specified that to be considered ready for monetary union a country’s inflation rate and long-term interest rates should first converge to values similar to those of other prospective members. The treaty also set targets for fiscal policy and exchange rate behavior for each prospective member. The specific convergence criteria are as follows: inflation in each prospective member is supposed to be no more than 11⁄2 percent above the average of the inflation rates in the three countries with lowest inflation rates; long-term interest rates are to be no more than 2 percent above the average interest rate in those countries; the exchange rate is supposed to have been kept within the target bands of the European Monetary System with no devaluations for at least two years prior to joining monetary union; and, importantly for the current debate, there are two requirements regarding fiscal policy. One fiscal criterion is that the budget deficit in a prospective member should be at most 3 percent of GDP; the other is that the outstanding amount of government debt should be no more than 60 percent of a year’s GDP. Table 1 shows each country’s performance in 1996 relative to the criteria for inflation, long-term interest rates, fiscal deficit, and level of government debt.4 A majority of the members of the EU satisfied the inflation and interest rate criteria, but nearly all were in violation of at least one of the two fiscal criteria. The economic rationale for the fiscal criteria is that such limits are needed to ensure the support and commitment of all monetary union members to the goal of low inflation enshrined in the treaty. Historically, governments have sometimes used inflation as a way of W 3.For evidence on whether Germany dominates the EMS, see von Hagen and Fratianni (1990) and Herz and Roger (1992). 4.The exchange rate criterion is not shown because it does not have a numerical value. As discussed earlier, the exchange rate criterion requires that during the two years prior to a country’s entry into monetary union, its exchange rate be kept within the target bands of the EMS, with no devaluations. Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 23 raising revenue to maintain spending on politically popular programs. Once monetary union is achieved, inflation could not be confined to one member but would necessarily involve the entire membership. A single member wishing to pursue an inflationary policy might exert pressure on the European Central Bank to raise inflation throughout the union. Alternatively, if a member government with large debts had financial difficulty, the European Central Bank might feel obliged to bail it out to avoid a financial crisis, at the cost of compromising its low-inflation goal. The fiscal criteria were intended to ensure that only govCritics argue that, from ernments with sound finances would be able an economic perspective, to enter the union. monetary union would be Moreover, to guard a mistake for Europe. . . . against future problems the members of the EU It would eliminate the poshave agreed to set limsibility of changing moneits on deficits even tary policy or the exchange after monetary union is achieved. According rate, forcing adjustment to the Stability Pact back onto the labor market. approved by European finance ministers in December 1996, members of the monetary union will be fined if they consistently violate the 3 percent limit on budget deficits (J.P. Morgan and Company 1996b).5 The economic rationale for the fiscal criteria has been questioned by many observers (see Bean 1992 and Kenen 1992). One issue is whether the specific numbers in the treaty are optimal. Buiter, Corsetti, and Roubini (1993) and Eichengreen (1994) argue that the numerical limits of 3 percent for deficits and 60 percent for debt are arbitrary and give little indication of whether a country is suitable for monetary union. According to Bean (1992), the only historical justification for these limits is that they happen to be close to the average that prevailed at the time the treaty was signed. Masson (1996) argues that efforts to meet the deficit criterion have diverted attention from other important fiscal problems. He points out that since 1992 governments have often used the Maastricht criterion as justification for imposing tax increases. However, considering the sluggish growth and high unemployment that have prevailed in many European countries, he suggests that reductions in high tax burdens and cutbacks in social transfers—for example, generous early retirement benefits and high unemployment benefits that discourage job seeking—would seem preferable methods of reducing budget deficits because they would encourage an expansion of economic activity and, by so 24 doing, raise the tax base. In many European countries, high taxes are needed to finance government spending of 50 percent or more of national output. Moreover, like the United States, many European countries will face major fiscal pressures soon after the turn of the century as demographic factors cause soaring increases in the cost of retirement programs.6 More generally, it is debatable whether restrictions on fiscal policy are needed for a successful monetary union. Eichengreen and von Hagen (1995, 1996) note that many existing monetary unions, including the union of Belgium and Luxembourg, impose no debt or deficit limits on the members. In the United States, there is no nationwide agreement that limits the budget deficits of individual states. Some states do have limits on deficit spending, but these were adopted on a stateby-state basis and were not motivated by the desire to make monetary union viable. Eichengreen and von Hagen study various monetary unions around the world, including the United States, Canada, and Australia. They argue that restrictions on borrowing by subunits of a monetary union are most common when those subunits have little control over their own sources of revenue: for example, in some countries almost all revenue is raised by the central government, with part being passed on to subunits to finance their activities. When the sub-units control their own sources of revenue, restrictions on borrowing are often not imposed. In the European context, the national governments will be subunits after monetary union is achieved. Currently and for the foreseeable future, the national governments are financed predominantly with their own sources of revenue: very little spending is or would be financed by central EU institutions based in Brussels. Eichengreen and von Hagen conclude that the EU would therefore not need fiscal limits on the national governments in order to have monetary union. Buiter (1996) also downplays the need for fiscal restrictions on members of the monetary union. In his view, a default or rescheduling by one member country—for example, Italy—would not necessarily be a problem for the EU as a whole. Its costs should properly fall on either the owners of the debt, Italian taxpayers, or those who benefit from Italy’s public spending. The European Central Bank would become involved if a financial or banking crisis ensued, but in his view the way to avoid such a snowballing crisis would be to use bank supervision and regulation to set upper limits on the exposure of financial institutions to risks of default by European governments. Of course, such limits on exposure might require significant portfolio shifts because relative to their capital many banking systems currently have large exposures to their home-country governments. Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 Regardless of its economic desirability, the deficit criterion has taken on political importance, especially because of German leaders’ insistence that only governments satisfying it be allowed into the monetary union. Germany’s highest court has ruled that the fiscal criteria are an integral part of the treaty: if they are violated, Germany may be required to renounce monetary union (Gros 1995, 4). Given Germany’s economic and political importance in Europe, its withdrawal of support would probably spell the end of monetary union. The actual language of the treaty is not in fact so rigid. It indicates that political leaders can exercise some judgment rather than having to apply the convergence criteria mechanistically to determine a country’s readiness. In particular, the treaty states that a deficit above 3 percent of GDP should not be considered excessive if the deficit “has declined substantially and continuously and reached a level that comes close” to the limit or if the deficit “is only exceptional and temporary” (European Commission 1997, 17). With so many prospective members in violation of at least some of the convergence criteria, and little chance that major progress can be made toward meeting them in the few months remaining before decisions about initial membership are scheduled to be made, the shape of the initial union remains an open question. The Initial Union: Maxi-, Mini-, or Delayed? he combination of doubts about whether all or even most EU members would actually satisfy all the convergence criteria and the insistence of some countries (notably Germany) that the criteria be strictly enforced has generated continuing debate over what will actually transpire when the treaty’s deadline for beginning the monetary union is reached. There are three main possibilities: a maxi-union, a mini-union, or delay. Maxi-union. A maxi-union would be a broad monetary union that would cover most of the EU, including at least three of the four largest members, namely, Germany, France, Italy, and Britain. This alternative would generate the greatest benefits in terms of T reduced transactions costs, but the diversity of its membership might produce severe internal strains as different countries push for different monetary policy choices. For example, according to Shilling (1997), 90 percent of mortgages in Britain have adjustable interest rates compared with only 30 percent in Germany. As a result, changes in interest rates have much stronger direct effects on homeowners in Britain than in Germany, effects that could at times result in divergent views about monetary policy in the two countries. As far as entry is concerned, the biggest question marks are whether Italy and Britain will join. Italy very much wants to join the monetary union, in part as a matter of national pride.7 Along with France and Germany, Italy was one of the founding members of the European Community in the 1950s, and it opposes the idea of being left behind by the others.8 Unfortunately for its chances, Italy has for several years been in violation of the convergence criteria (European Commission 1997). Perhaps the biggest hurdle is its fiscal problem: its budget deficit has exceeded the 3 percent limit for some years, and its stock of debt has exceeded 120 percent of GDP, double the convergence criterion of 60 percent. In recent years Italy has managed to reduce its inflation rate and long-term interest rates enough to more nearly satisfy those two criteria than in the past, and it has also cut its budget deficit substantially. Moreover, it reentered the EMS in late 1996, albeit with new target bands that implied a substantial devaluation (more than 30 percent) from the rate prevailing just before its departure in 1992. If Italy can keep its exchange rate within the new target bands until late 1998, it will come into compliance with the exchange rate criterion just prior to the scheduled start of monetary union. Nevertheless, only loose interpretations of the convergence criteria will make it likely that Italy can qualify for monetary union in 1999, given its fiscal problems. Britain is in a relatively good position as far as the convergence criteria, but it has a traditional diffidence about tying itself to its continental neighbors as shown, for example, by its decision not to join the Common 5.The fines may range from 0.2 to 0.5 percent of a country’s GDP and are imposed by vote of the political leaders on the European Council, the highest decision-making body of the EU. No fine is to be imposed if the deficit occurs during a severe recession or because of “exceptional circumstances.” 6.The Organisation for Economic Cooperation and Development (OECD) (1995) shows that of the G-7 countries (the United States, Germany, France, Italy, the United Kingdom, Japan, and Canada), all except the United Kingdom will face major budgetary pressures from the interaction of demographics and retirement programs after the turn of the century. 7.See the Financial Times, October 11, 1995, and July 21, 1997. 8.For example, when the German finance minister stated that Italy would not be one of the initial members of monetary union, his remarks created a furor in Italy. The Italian prime minister responded by insisting that his country was committed to being an initial member and also suggested that the entire project be delayed rather than go forward without Italy. See the Financial Times, September 23 and 25, 1995. Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 25 Market when it was founded in the 1950s.9 In addition, Britain is particularly cautious about pegging its exchange rate because, rightly or wrongly, several periods of recession or sluggishness in Britain during this century have been blamed on pegging the pound at an overvalued rate.10 The Labour government that took power in May 1997 seems inclined to take a wait-and-see attitude, with the intention of joining monetary union only after it has proven to be a success. However, it is possible that Britain might decide to go ahead if nearly all the other members, including Italy, decide to start together in 1999.11 Mini-union. The second possibility is a mini-union that would leave out much of the EU. Most discussion of this option has focused on a possible union involving Germany, France, the Benelux countries, and perhaps one or two other small countries like Ireland or Austria. These countries have been in compliance with several of the convergence criteria for the past several years and are likely to continue in compliance in 1997 (European Commission 1997). Moreover, exchange rates within this smaller group have been kept within fairly narrow bands for the past several years. Nevertheless, even the mini-union faces obstacles. One problem is that several of these countries are in violation of at least one of the convergence criteria, usually in the fiscal area. In particular, for some time Belgium has had outstanding debt of more than 120 percent of GDP. Moreover, budget deficits in Germany and France were above the convergence limit of 3 percent of GDP in 1995 and 1996, partly because sluggish growth has boosted spending on such items as unemployment benefits while cutting into tax receipts (European Commission 1997). Delayed. The problems with the maxi- and miniunion options have led to speculation about a third option: delay. Perhaps the start of monetary union could be put off for two or three years in the hope that by that time Germany and France (and perhaps Italy) would be in compliance with the deficit criterion. This option would, however, cause acute political embarrassment for the leaders of France and Germany, who have been instrumental in pushing monetary union forward. Another risk is that during the interim other obstacles could arise that would indefinitely delay monetary union. The main risk, though, is the possibility that the new deadline might not be credible to market participants, resulting in financial market turmoil. The Bargaining Process ow will the bargaining process turn out? Economic models of strategic behavior, in which the offers made by bargainers are influenced by their expectations about the future behavior of others, H 26 offer some insights. Chang (1995) develops a model in which two countries may benefit from monetary union but each wants to maximize its share of those benefits. In some cases the two countries will reach an immediate agreement and unify their currencies, but in other cases agreement will be delayed by a number of periods of bargaining. Moreover, private market expectations affect the length of delay. Alesina and Grilli (1993) examine whether a mini-union is a good first step toward complete monetary union. They consider the case in which a few countries in the EU proceed with monetary union and then, by majority vote, decide whether to allow additional countries to join. Because a maxi-union is politically feasible in their model, in the sense that every country is better off with a maxi-union than with no union, one might logically expect a miniunion to be a stepping stone to full union. Alesina and Grilli show, however, that a mini-union may in fact prevent complete monetary union because it may be in the interests of the initial members to veto the others. This analysis provides a rationale for Italy’s reluctance to be left behind at the beginning of monetary union. Other scenarios are possible. The Maastricht treaty specifies that the decision to admit an individual country into monetary union will not be made by a simple majority vote but by the vote of a “qualified majority.”12 Under this procedure, less than half the fifteen members of the EU could block a country’s admission. De Grauwe (1996b) argues that if the most commonly discussed version of a mini-union is proposed (consisting of Germany, France, the Netherlands, Belgium, and Luxembourg, plus possibly Austria or Ireland), such a proposal will be blocked by negative votes from some of those left out. For example, the group of the four southern European members (Italy, Spain, Portugal, and Greece) have enough votes to block the mini-union. De Grauwe concludes that the only politically viable choices will be maxi-union or postponement. Countries such as Germany might find maxi-union more palatable if a suggestion by De Grauwe (1995) and Gros (1995) were adopted. They propose that, rather than putting so much emphasis on whether a country meets the convergence criteria prior to the start of monetary union, as in the current transition process, the treaty be changed to take away a country’s vote on the European Central Bank’s governing board after the start of monetary union whenever that member violates the deficit limit. Under this approach, countries with large fiscal deficits could join and remain members of the union but would be unable to vote to bail themselves out of fiscal difficulty through higher inflation. J.P. Morgan (1996a) suggests another possibility. A mini-union could start in 1999, with disappointed would-be members such as Italy or Spain assuaged by a conditional commitment that they would be allowed to Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 C H A R T 1 Eurocurrency Interest Rates (Italy minus Germany, Weekly Average) 10 Spread (percent) 8 6 4 2 0 1990 1991 1992 1993 enter a year or two later, as long as they made further progress toward meeting the convergence criteria. Because the members of the EU are constantly negotiating over a wide range of issues, there are many such options that countries strongly favoring an initial mini- or maxi-union can use to try to win over their opponents. Accordingly, as long as political leaders in the two largest countries in the EU, Germany and France, are committed to going ahead, the prospects for at least a mini-union beginning in 1999 seem favorable. Market-Based Probabilities of Monetary Union he prospect of monetary union has implications for the patterns of interest rates in Europe that can be used to make rough estimates of whether market participants expect monetary union to go forward. In the past, interest rates often differed considerably, even for the same borrower, depending on the denomination of the debt’s currency. For example, Chart 1 shows that over the last several years shortterm interest rates denominated in Italian lire have usually been at least 200 basis points above those T 1994 1995 1996 1997 denominated in deutsche marks, thereby compensating for expected depreciation of the Italian currency. Once exchange rates are fixed permanently at the start of monetary union, currency of denomination should no longer affect interest rates in the member currencies because (for example) Italian lire and German deutsche marks would both be convertible into the new euros at a fixed and unchanging rate: expected depreciation of the lira vis-à-vis the deutsche mark would become zero. Long-term interest rate contracts that are written before monetary union but apply to periods after it should reflect this lack of future depreciation. A special version of a concept called covered interest parity can be used to estimate the probability that market participants attach to monetary union going ahead on schedule. Intuitively, covered interest parity states that as investors seek the highest returns on their liquid assets in different countries, foreign returns that are hedged or “covered” against future changes in exchange rates will be equal to the returns on similar domestic assets. 9.For instance, see Dornbusch (1996), Sutherland (1997), and the Economist (September 21, 1996). 10.One such episode occurred in the 1920s when Britain returned the pound to its pre–World War I value in terms of gold and the dollar (see Ingram 1983, 140–56). Another occurred in the mid-1960s, culminating in the devaluation of the pound in late 1967 (see Cohen 1969, 143–49). 11. See Sutherland (1997). One factor in Britain’s decision is the concern of some economists and financial executives that London’s role as a financial center would suffer if monetary union, especially a maxi-union, occurs and Britain stays outside. Their influence on the new government may be sufficient to overcome the doubters. See “Growing Fears in Britain of Single-Currency Isolation,” New York Times, August 22, 1996, D2. 12.Qualified majority voting is a special system of weighted voting used by the EU. Under this procedure, large countries such as Germany and France have more weight than small ones such as Ireland and Finland. To be approved by a qualified majority, a proposal must win roughly 70 percent of the weighted votes. Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 27 More technically, covered interest parity states that in the absence of capital controls, the difference between the spot exchange rate (the rate on conversions of money from one currency to another that settle immediately) and the forward exchange rate (the rate on conversions of money that are agreed upon today but do not settle until some time in the future) is just enough to offset cross-country differences in interest rates, thereby making investors indifferent between investing at home and abroad. Covered interest parity, discussed in more depth in many international economics textbooks, can be represented as follows: Rt*,T − Rt,T = Ft,T − St , St (1) where Rt,T is the interest rate in the home country at time t on securities (for example, Treasury bills or certificates of deposit) that mature at time T (for example, three months in the future); Rt,T* is the interest rate in the foreign country on similar securities with the same maturity date; St is the spot exchange rate at date t, measured in foreign currency per unit of domestic money; Ft,T is the forward exchange rate, measured in foreign currency per unit of domestic money, that is agreed upon on date t but does not settle until the forward contract matures on date T (which coincides with the term of the interest rates Rt,T and Rt,T* ). Equation (1) is the most common form of covered interest parity; if it holds, and various researchers such as Frenkel and Levich (1975) and Taylor (1987) have found that it holds quite well when capital controls are not in force, an investor gets the same rate of return on a foreign security that is covered for exchange rate risk as on a domestic security. Covered interest parity should also hold in terms of forward interest rates. A forward interest rate is an interest rate that pertains to a time period that begins not today but sometime in the future. For instance, suppose some investors have bonds that mature five years in the future, and they wish to lock in a return on those funds for three additional years. Some banks are willing to agree today to accept the investors’ deposit of the bond proceeds five years from now and to pay them interest at a rate agreed today for the following three years. The interest rate on such a contract would be a forward interest rate with settlement five years in the future and a maturity of three years. Suppose such investors were considering two options: investing at the forward deutsche mark interest rate or investing at the forward Italian lira interest rate. If deutsche marks are the home currency for these investors, at the maturity date of the forward interest 1 contract they would receive (1 + RDt,t1,T)T–t deutsche marks for each mark they deposited, where RDt,t1,T is the annualized forward interest rate on deutsche marks today (at time t) for settlement at time t1 with maturity date T; hence the funds would actually be on deposit for (T – t1) years. Alternatively, if the investors made a covered investment in Italian lire, they would convert each deutsche mark to Ft,t1 lire at time t1 (where Ft,t1 is the forward exchange rate in lire per mark prevailing today [at time t] for settlement at time t1). They would then invest the proceeds until time T, receiving at the end the 1 amount (Ft,t1) 3 (1 + RLt,t1,T)T–t in lire; RLt,t1,T is the annualized forward interest rate on lire today (at time t) for settlement at time t1 with maturity date T. Finally, to be fully covered against exchange rate risk, the investors C H A R T 2 Forward Interest Rates (Italy minus Germany, Weekly Average) 10 Spread (percent) 8 6 4 2 0 –2 –4 –6 –8 1990 28 1991 1992 1993 Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 1994 1995 1996 1997 would have to make a contract today to convert their final proceeds in lire back into deutsche marks: each lira would yield (1/Ft,T) deutsche marks. For investors to be indifferent between the two investment alternatives, the ultimate return on them should be equal, or (1 + RDt, t1, T )T− t = 1 Ft, t1 Ft, T (1 + RLt, t1, T)T − t , 1 (2) where the term on the right-hand side is the covered return (in deutsche marks) from investing at the Italian forward interest rate. Equation (2) is another version of covered interest parity, expressed in terms of forward interest rates. De Grauwe (1996a) observes that if market participants were convinced that permanent monetary union involving Italy and Germany would occur on schedule and that the times t1 and T were both after the scheduled beginning of monetary union, then the forward exchange rates Ft,t1 and Ft,T should be identical to one another. This equality would hold even if the postunion conversion ratio between lire and deutsche marks, St1, was uncertain as of time t. In this case, equation (2) indicates that the forward interest rates RDt,t1,T and RLt,t1,T would be equalized as well, even if they were observed at time t well before monetary union. As an example, Chart 2 shows the difference between lira and deutsche mark forward interest rates from 1990 through January 28, 1997. These are five-year forward interest rates on five-year interbank loans.13 The horizontal axis gives the date of observation, t in terms of the notation in equation (2), where settlement date t1 is five years after t and maturity date T is five years after t1. The earliest observations were made well before the Maastricht Treaty was signed, at a time when expectations of monetary union were presumably low. Moreover, for the early observations, most of the fiveyear periods covered by the contracts (the period between the settlement date t1 and the maturity T) fell before the scheduled 1999 start date for monetary union. For example, the points on the chart for January 1991 represent contracts made in January 1991 with settlement dates in January 1996 and maturity dates in January 2001, implying that roughly three-fifths of the period covered by these particular contracts fell before 1999. The treaty was signed in late 1991, and as time passed the fraction of the period covered by these contracts that fell after the scheduled beginning of monetary union gradually rose, making expectations about monetary union more and more important in their deter- mination. After January 1, 1994, the entire period covered by these contracts fell after the scheduled beginning of monetary union. During the first few months shown in Chart 2 Italy’s five-year forward interest rate spread showed tremendous volatility, perhaps in part because capital controls were still in effect, though slated for removal. By the second half of 1990 volatility lessened, and the spread usually ranged between 300 and 500 basis points. Over the next few years the forward spread tended to rise and fall along with the short-term interest rate spread shown in Chart 1, but in the second half of 1996 the forward spread fell well below the short-term spread. In January 1997 the forward spread averaged only 87 basis points, even though the short-term spread for the next three months was still high, 411 basis points. The drop in the forward spread so far below the short-term spread is consistent with a market expectation that in the future Italian interest rates will be much closer to German ones than they are today. A rough estimate of the probability of EMU can be derived from the forward interest rate spreads, as described by De Grauwe (1996a). Suppose the forward interest rate spread observed before the beginning of monetary union is a weighted average of the spreads that would prevail in two alternative scenarios—namely, that monetary union occurs on schedule or it is delayed indefinitely—with the weights being the market’s assessment of the probability of each. That is, sto = pt × stu + (1 − pt ) stN , (3) where sto is the forward interest rate spread versus the deutsche mark observed at time t (prior to monetary union), pt is the market’s assessment at time t of the probability that monetary union will proceed on schedule, stu is the spread that would be expected to prevail if monetary union proceeds, and stN is the spread that would be expected to prevail if monetary union does not proceed. As discussed earlier, if the forward interest rates pertain entirely to the period after monetary union is scheduled to begin, they should be equalized, implying that stu would be zero. In this case, equation (3) simplifies to the following expression for the probability of monetary union pt: pt = 1 − ( sto / stN ). (4) As noted earlier, data on sto are available. The problem is estimating stN, the forward interest rate spread that 13.Market quotations on forward interest rates in various European currencies were provided by J.P. Morgan and Company. Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 29 C H A R T 3 Probability of Entering EMU France Percentage 100 80 60 40 20 0 1994 1995 1996 1997 1996 1997 1996 1997 1996 1997 1996 1997 Britain Percentage 100 80 60 40 20 0 1994 1995 Belgium Percentage 100 80 60 40 20 0 1994 1995 Italy Percentage 100 80 60 40 20 0 1994 1995 Spain Percentage 100 80 60 40 20 0 1994 30 1995 Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 would be expected to prevail if monetary union does not proceed. De Grauwe suggests using the average spreads on five-year forward interest rates that were observed during 1990, a year before the Maastricht Treaty was signed and a time when monetary union seemed a remote possibility. However, charts of the data on spreads show great volatility for Italy (see Chart 2) and Spain in early 1990, including some sizable negative values (meaning that Italian and Spanish five-year forward interest rates fell substantially below German ones). The position taken here, however, is that these negative values reflect market imperfections or capital controls and do not represent the market’s true expectation about future interest rates. Indeed, Italy did not abolish its remaining restrictions on capital flows until May 14, 1990 (Ungerer and others 1990). Accordingly, instead of using all of 1990 to estimate stN, this study uses the average spread during the second half of that year. The resulting estimates of pt for France, Britain, Belgium, Italy, and Spain are shown in Chart 3. The charts start in January 1994, when the five-year forward interest rates used in the calculations began to apply solely to the period after the scheduled commencement of monetary union in January 1999. The probabilities pt calculated using equation (4) were converted to percentages by multiplying by 100. A value of zero corresponds to zero probability that the country will be part of monetary union, while a value of 100 corresponds to virtual certainty that the country will participate.14 The results for France indicate that the markets have usually regarded that country as being almost certain to participate in monetary union. The results for Britain are surprising: since 1994 market estimates of the probability of British participation have usually been well above 50 percent and in January 1997 some 80 percent. Considering the political opposition that exists in Britain, these probabilities seem high. De Grauwe (1996a, 11–12) obtains similar results but argues that 1990 was a poor benchmark year for such calculations in the British case because in October of that year the pound entered the EMS after months of market turbulence. Moreover, many observers claimed at the time that the pound had entered the EMS at an overvalued exchange rate and that sooner or later a devaluation was certain. The forward interest rate spread for pound sterling that prevailed during 1990 may therefore not be an accurate proxy for the market’s estimate of the spread that would prevail if Britain stayed out of monetary union. Another explanation of the surprising British results is given in the caveat below. Belgium, Italy, and Spain all show notable increases in the last year or so of the period. De Grauwe (1996a), whose data sample ended in March 1996, reported that as of the end of his sample Belgium’s probability of joining was about 60 percent while that of Italy and Spain was much less, perhaps 20 or 30 percent. Using the slightly different proxy for s tN (only the last half of 1990) and extending the sample through January 1997, the charts in this article show end-of-sample probabilities of about 100 percent for Belgium, 90 percent for Spain, and more than 80 percent for Italy. Strictly speaking, these are probabilities that the country in question will be in monetary union five years after the date of the observation. Accordingly, the observations from January 1997 indicate a high likelihood that these three countries will enter monetary union either at the scheduled beginning in January 1999 and certainly no later than January 2002. Using a somewhat different approach, J.P. Morgan (1997) has also estimated market expectations of EMU that are quite high for most of these countries. This approach uses non-European financial data to estimate the forward interest rate spread for potential EMU members that would prevail if there were zero probability of the country joining.15 The company’s results indicate that market perceptions of the likelihood of a maxiunion increased noticeably in the second half of 1996, probably because of progress toward the Stability Pact at two political summits during the period. As of early February 1997, which was approximately the end of the data sample used in Chart 3, the Morgan approach yielded the following probabilities of joining EMU: 100 percent for France and Belgium, 85 percent for Spain, 66 percent for Italy, and 40 percent for Britain.16 These 14.In some time periods, the estimated probability obtained from equation (4) is either negative or above 100 percent. Negative values can occur if the observed spread sto is larger than the average spread that prevailed in the second half of 1990, when monetary union was presumably considered a remote possibility. Values above 100 percent can occur if the observed spread sto is negative, meaning that the country’s forward interest rate is actually lower than Germany’s. Because probability is normally defined only in the range between zero and 100 percent, the chart was drawn showing such observations as falling at either 100 or zero. 15.For example, J.P. Morgan regressed French franc forward interest rate spreads vis-à-vis Germany onto financial variables not directly affected by EMU, such as the U.S. three-month rate, the Japanese three-month rate, and the difference between ten- and two-year interest rates in the United States. The regression was estimated using data from the late 1980s and early 1990s, a period when expectations of monetary union should have been very low. The estimated coefficients were then applied to recent data on U.S. and Japanese rates in order to generate a proxy for the current value of stN. 16.These probabilities were reported in the Financial Times, March 4, 1997. In later weeks the probabilities for Spain, Italy, and Britain dipped somewhat. Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 31 estimated probabilities are lower than those shown at the end of Chart 3 but are still quite substantial. If either set of estimates is correct, market participants have concluded that a mini-union, at least, seems virtually certain in the next few years and a maxi-union including the two largest countries in Southern Europe is quite likely. The similarity of the probabilities obtained by different methods is comforting, but an important caveat is in order. All these estimates depend critically on the accuracy of the measure of the unobservable spread that would prevail if monetary union did not occur. If the proxy is not correct, the estimated probability derived using it will of course be inaccurate as well. Consider the case of Britain. In the chart, as noted earlier, the spread that actually prevailed between fiveyear forward rates on sterling and deutsche marks during the second half of 1990 was used as a proxy—321 basis points. By January 1997 that spread had shrunk to only 57 basis points, implying, using equation (4), approximately an 80 percent probability that Britain would join monetary union by 2002. However, the shrinkage in the spread has another possible interpretation: perhaps Britain’s commitment to continuing low inflation became substantially more credible during the years between 1990 and 1997. In that case, one would expect a reduction in the spread vis-à-vis deutsche marks even if the market really was not expecting Britain to participate in monetary union. While the caveat suggests that all these probability estimates should be treated with caution, the view here is that it does not vitiate the entire exercise. Perhaps the most interesting results are those for Italy and Spain. Though still not in compliance with the convergence criteria, these two countries have made important progress in reducing their budget deficits during the past few years. Italy has cut its deficit from 9.6 percent of GDP in 1993 to 6.6 percent in 1996, and Spain has cut its from 6.8 to 4.4 percent over the same period (European Commission 1997, 12). The governments of both countries have consistently supported their membership in the proposed monetary union. According to the results shown here, market participants think they have a substantial likelihood of joining the union. Conclusion ore than five years ago the members of the EU decided to form a monetary union by the end of the decade. Many preparatory steps have been taken, but a number of key decisions, notably about which countries will be part of the union at its beginning, still remain to be made. These choices cannot be put off much longer. There is little chance that most of the countries will comply with a strict reading of the convergence criteria for membership, but evidence from the financial markets suggests that by early 1997 market participants were leaning toward the belief that the political impetus in favor of a broad union might prevail in the end, resulting in a monetary union that would encompass a substantial portion of western Europe. The recent election in France has injected new uncertainty into the process, though, and final decisions about monetary union may remain up in the air until the last possible moment. M REFERENCES ALESINA, ALBERTO, AND VITTORIO GRILLI. 1993. “On the Feasibility of a One- or Multi-Speed European Monetary Union.” National Bureau of Economic Research Working Paper No. 4350, April. DE GRAUWE, PAUL. 1995. “The Economics of Convergence towards Monetary Union in Europe.” Centre for Economic Policy Research Discussion Paper No. 1213, July. BEAN, CHARLES. 1992. “Economic and Monetary Union in Europe.” Journal of Economic Perspectives 6 (Fall): 31–52. ———. 1996a. “Forward Interest Rates as Predictors of EMU.” Centre for Economic Policy Research Discussion Paper No. 1395, May. BUITER, WILLEM H. 1996. “The Economic Case for Monetary Union in the European Union.” Cambridge University. Photocopy, November. ———. 1996b. “The Prospects of a Mini Currency Union in 1999.” Centre for Economic Policy Research Discussion Paper No. 1458, September. BUITER, WILLEM, GIANCARLO CORSETTI, AND NOURIEL ROUBINI. 1993. “Excessive Deficits: Sense and Nonsense in the Treaty of Maastricht.” Economic Policy 16 (April): 58–100. DORNBUSCH, RUDIGER. 1996. “Euro Fantasies.” Foreign Affairs 75 (September/October): 110–24. CHANG, ROBERTO. 1995. “Bargaining a Monetary Union.” Journal of Economic Theory 66:89–112. COHEN, B.J. 1969. Balance of Payments Policy. Baltimore: Penguin Books. 32 EICHENGREEN, BARRY. 1993. “European Monetary Unification.” Journal of Economic Literature 31 (September): 1321–57. ———. 1994. “Fiscal Policy and EMU.” In The Political Economy of European Monetary Unification, edited by Barry Eichengreen and Jeffry Frieden, 167–90. Boulder, Colo.: Westview Press. Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 EICHENGREEN, BARRY, AND JÜRGEN VON HAGEN. 1995. “Fiscal Policy and Monetary Union: Federalism, Fiscal Restrictions, and the No-Bailout Rule.” Centre for Economic Policy Research Discussion Paper No. 1247, September . KENEN, PETER B. 1992. EMU after Maastricht. Washington: Group of Thirty. MASSON, PAUL. 1996. “Fiscal Dimensions of EMU.” Economic Journal 106 (July): 997–1004. ———. 1996. “Federalism, Fiscal Restraints, and European Monetary Union.” American Economic Review Papers and Proceedings 86 (May): 134–38. EUROPEAN COMMISSION. 1990. “One Market, One Money: An Evaluation of the Potential Benefits and Costs of Forming an Economic and Monetary Union.” European Economy 44. ———. 1997. “Report on Convergence in the European Union in 1996.” European Economy (January, Supplement A, no. 1): 1–31. FELDSTEIN, MARTIN. 1992. “The Case against EMU.” Economist, June 13, 23–26. FRENKEL, JACOB A., AND RICHARD M. LEVICH. 1975. “Covered Interest Arbitrage: Unexploited Profits?” Journal of Political Economy 83 (April): 325–38. GARRETT, GEOFFREY. 1994. “The Politics of Maastricht.” In The Political Economy of European Monetary Unification, edited by Barry Eichengreen and Jeffry Frieden, 47–65. Boulder, Colo.: Westview Press. GROS, DANIEL. 1995. “Towards a Credible Excessive Deficits Procedure.” Centre for European Policy Studies Working Document No. 95, July. ORGANISATION FOR ECONOMIC COOPERATION AND DEVELOPMENT. 1995. “Effects of Ageing Populations on Government Budgets.” OECD Economic Outlook 57 (June): 33–42. PADOA-SCHIOPPA, TOMMASO. 1994. “The EMS Is Not Enough: The Need for Monetary Union.” In The Road to Monetary Union in Europe, edited by Tommaso Padoa-Schioppa, 117–34. Oxford: Clarendon Press. SHILLING, A. GARY. 1997. “Good Luck, Tony.” Forbes 159, February 24, 184. SUTHERLAND, PETER. 1997. “The Case for EMU.” Foreign Affairs 76 (January/February): 9–14. TAYLOR, MARK P. 1987. “Covered Interest Parity: A HighFrequency, High-Quality Data Study.” Economica 54 (November): 429–38. UNGERER, HORST, JOUKO HAUVONEN, AUGUSTO LOPEZ-CLAROS, AND THOMAS MAYER. 1990. “The European Monetary System: Developments and Perspectives.” International Monetary Fund Occasional Paper No. 73, November. HERZ, BERNHARD, AND WERNER ROGER. 1992. “The EMS Is a Greater Deutschemark Area.” European Economic Review 36:1413–25. VON HAGEN, JÜRGEN, AND MICHELE FRATIANNI. 1990. “German Dominance in the EMS: Evidence from Interest Rates.” Journal of International Money and Finance 9 (1990): 358–75. INGRAM, JAMES C. 1983. International Economics. New York: John Wiley and Sons. J.P. MORGAN AND COMPANY. 1996a. “EMU: Impact on Financial Markets.” European Fixed Income Research, August. ———. 1996b. “FX Markets and Europe’s Stability Pact.” Foreign Exchange Research, December. OBSTFELD, MAURICE. 1996. “Destabilizing Effects of ExchangeRate Escape Clauses.” University of California at Berkeley, Center for International and Development Economics Research Working Paper No. C96-075, December. WOOLLEY, JOHN T. 1994. “Linking Political and Monetary Union: The Maastricht Agenda and German Domestic Politics.” In The Political Economy of European Monetary Unification, edited by Barry Eichengreen and Jeffry Frieden, 67–86. Boulder, Colo.: Westview Press. ———. 1997. “EMU Calculator Handbook.” Global Foreign Exchange Research, January. Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 33 The Role of Currency Derivatives in Internationally Diversified Portfolios P E T E R A . A B K E N A N D M I L I N D M . S H R I K H A N D E Abken is a senior economist in the financial section of the Atlanta Fed’s research department. Shrikhande is an assistant professor of finance at the Georgia Institute of Technology and a visiting scholar in the financial section of the Atlanta Fed’s research department. The authors thank Rob Bliss, Jerry Dwyer, Larry Wall, and Tao Zha for helpful comments and Elizabeth Bram of Salomon Brothers for providing bond return data. T HE POWER OF DIVERSIFICATION IN REDUCING RISK IS WIDELY UNDERSTOOD AND PRACTICED BY INVESTORS. IN RECENT YEARS INVESTORS HAVE BEEN TURNING TO FOREIGN MARKETS TO OBTAIN EVEN GREATER SCOPE FOR DIVERSIFICATION THAN IS POSSIBLE IN A DOMESTIC MARKET. WITH THE INTERNATIONALIZATION OF SECURITY PORTFOLIOS, HOWEVER, ALSO COMES AN ADDITIONAL RISK—FOREIGN EXCHANGE RISK.1 FOREIGN EXCHANGE RATE FLUCTUATIONS INDUCE CHANGES IN PORTFOLIO RETURNS BECAUSE UNCERTAIN FUTURE EXCHANGE RATES TRANSLATE RETURNS ON FOREIGN-CURRENCY-DENOMINATED INVESTMENTS INTO DOLLAR RETURNS.2 DIVERSIFICATION OF PORTFO- LIO HOLDINGS ACROSS SEVERAL COUNTRIES CAN HELP MITIGATE FOREIGN EXCHANGE RISK. DERIVATIVE SECURITIES ARE INSTRUMENTS THAT ALTER THE CASH FLOWS OF A PORTFOLIO. THE USE OF CURRENCY DERIVATIVES CAN FURTHER REDUCE RISK IN INTERNATIONALLY DIVERSIFIED PORTFOLIOS. This article investigates the impact of currency hedging on internationally diversified stock and bond portfolios. It explains how currency hedging works and shows how hedging affects actual historical portfolio returns. The focus of the analysis is on index portfolios of stocks and bonds from markets in seven industrialized countries. Portfolio diversification eliminates the influence of what is called idiosyncratic risk—the unpredictable losses specific to individual security returns—from a securities portfolio. Domestic diversifi34 cation, however, leaves exposure to “systematic” risk, the unpredictable losses that affect all domestic securities.3 Because domestic systematic risks are likely to differ from country to country, international diversification can further reduce the volatility of portfolio returns by mitigating country-specific risk. Several studies have suggested that hedging against foreign exchange risk has little effect on expected return, or may even enhance it, while reducing the variability of portfolio returns (Perold and Schulman Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 1988; Thomas 1988, 1989; Eun and Resnick 1988, 1994; Kaplanis and Schaefer 1991; Eaker and Grant 1991; Glen and Jorion 1993; Levich and Thomas 1993).4 The advantage of hedging has even been described as a “free lunch” (Perold and Schulman 1988) because currency hedging appears to deliver benefits at no cost. This article reexamines the data for international equity and bond returns and foreign exchange rates for sample periods running from 1980 to 1996 for equities and from 1986 to 1996 for bonds. Most of the previous studies include sample periods that were dominated by the dollar’s appreciation against most major currencies during the first half of the 1980s. After the Plaza Accord of 1985, the dollar began a long depreciation that lasted until the mid-1990s.5 This more recent period is also characterized by a different structure of security return and foreign exchange volatilities and correlations. This change had a significant impact on the apparent performance of hedging. These relationships and their effects are explained in the following sections. The results in this article are derived from an analysis of “efficient” portfolios, securities portfolios that offer the greatest feasible return for a given level of risk. The apparent risk-reducing benefits of currency hedging of equity portfolios in the early 1980s are not confirmed for the 1986–96 period overall or for subperiods. In contrast, foreign long-term bond portfolios consistently exhibited dramatically lower variability of hedged returns compared with the variability of unhedged returns, a finding that agrees with results from the earlier studies of currency hedging based on earlier sample periods. However, even for bond portfolios, the case for currency hedging is not decisive because, historically, the lower variability of hedged return is associated with lower returns. The decision to hedge depends on the investor’s preference for risk and return. Diversification odern portfolio theory dates back to the work of Markowitz (1952). Markowitz started with the assumption that a portfolio’s riskiness may be measured by the variance of its returns. He showed that an investment in a portfolio of securities offers investors risk and return combinations that are not possible from individual securities. In most cases, diversification allows an investor to obtain higher expected return for the same risk or lower risk for the same expected return relative to the return available from a single security. Markowitz’s insight is easily seen by considering the formulas for the mean and variance of return from a two-asset portfolio: M rp = w1r1 + w2r2 sp2 = w12 s21 + w22 s22 + 2w12 w22 rs1s2, (1) where ri is the security return on security 1 or 2 or the portfolio p, s 2i is the variance of the corresponding return, w1 and w2 are portfolio weights, and r is the correlation coefficient between the individual security returns. Because the portfolio weights are assumed to sum to one, the portfolio mean return is a weighted average of the returns on the two assets. However, because of the covariance term for s p2 in equation (1), a portfolio containing both securities will usually have a lower standard deviation (square root of the variance) than simply a weighted sum of their individual standard deviations.6 An efficient portfolio has the greatest feasible return for a given standard deviation of return. The later empirical section focuses principally on determining the proportions of international stock or bond portfolios that generate the efficient “frontier,” which is a graph of efficient portfolios’ standard deviations against their returns. A particular investor’s taste for risk and 1. Stock prices themselves may reflect the foreign exchange exposures of firms with multinational operations. However, firms can reduce this risk using derivatives or other risk-management techniques. See Chow, Lee, and Solt (1997). 2. A security’s return is the rate of price appreciation, including associated cash flows such as dividend or interest payments. 3. The measure of risk used in this article is the standard deviation of a security’s excess return, that is, its return in excess of the risk-free rate of interest. Because the portfolios used are market-value weighted index portfolios of stocks or bonds, the variability of excess return is assumed to reflect predominately systematic risk. 4. This article focuses on “buy and hold” strategies, constructed purely as hedges, which are described in a later section. A number of studies on foreign exchange markets claim that foreign exchange movements contain a predictable component. For example, Glen and Jorion (1993) and Levich and Thomas (1993) show that by taking positions in foreign exchange derivatives based on forecasts of exchange rate movements, it is possible to earn “excess returns.” The key unresolved issue regarding these returns is whether they represent compensation for risk exposure. Exploiting apparent foreign exchange market inefficiencies may offer the potential to enhance expected return without increasing risk. 5. In September 1985 the finance ministers and central bank governors of the so-called Group of Six industrial countries (the United States, France, Germany, the United Kingdom, Japan, and Canada) met at the Plaza Hotel in New York City and reached what was later referred to as the Plaza Accord or Agreement. They announced that it would be desirable for most major currencies to appreciate vis-à-vis the U.S. dollar and pledged to intervene in exchange markets to accomplish this objective. The dollar had already started to fall during the spring and summer of 1985. 6. For perfectly correlated returns (p = 1), the standard deviation of portfolio returns is exactly equal to the weighted sum of the standard deviations of the individual returns. Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 35 return would guide the selection of an optimal portfolio along the efficient frontier. International Diversification. The principles of diversification apply regardless of the kind of assets or currencies of their denomination. What is different about investing abroad is the assumption of foreign exchange risk that comes with owning foreign securities. The impact of exchange rate movements on unhedged dollar returns can be understood by considering the return in terms of two sources of risk, volatility of the foreign security returns and volatility of the foreign exchange rate. Movements in both are largely unpredictable, and they are generally correlated with one another. A security’s rate of Foreign exchange rate return, measured from fluctuations induce changes period t – 1 to t, is defined as the rate of in portfolio returns because price appreciation plus uncertain future exchange associated cash flows rates translate returns such as dividend or interest payments. The on foreign-currencyreturn based on prices denominated investments denominated in foreign into dollar returns. or “local” currency is referred to as the local return. 7 The rate of change of the exchange rate, st / s t–1 – 1, is denoted by et, where positive values signify an appreciation of the foreign currency. The rate of return at time t in dollars on an unhedged foreign investment is rt = (1 + rlt)(1 + et) – 1 = rlt + et + rltet.8 The dollar return rt depends on the local security return rlt and the rate of change of exchange rate et. The dollar return can be approximated by rt ' rlt + et because the cross-product term rltet is generally small. For example, if the foreign equity index over a three-month period depreciated by 3 percentage points and paid dividends at the rate of 1 percent of the index level, the local return would be –2 percentage points. If during the same period, the exchange rate appreciated by 4 percentage points, the total dollar rate of return would be approximately 2 percentage points. For a portfolio involving securities denominated in several currencies, the diversification effects can be described by giving a weight wi to each portfolio, where the subscript i indexes the portfolios available. The weights for N index portfolios sum to unity. Based on the approximation rt ' rlt + et for the unhedged dollar return of an individual securities portfolio, Eun and Resnick (1988) derive an approximation for variance of the return on an unhedged multicountry securities portfolio that takes the following form analogous to equation (1): 36 σ u ≈ ∑ i =1 ∑ j =1 wi wj ρijσ iσ j N 2 N l l l (2) 144 42444 3 security return covariances + wwρ σ σ ∑ ∑42 144 444 3 N N i =1 j =1 i j e e e ij i j foreign exchange rate covariances σ , ∑ ∑ w w ρ σ3 14442444 +2 N N i =1 j =1 i j l ,e l e ij i j local return–foreign exchange rate covariances where the first term represents weighted covariances of the security returns on the N index portfolios making up the overall portfolio (with superscript l for local return), the second is for the covariances of the corresponding exchange rates (with superscript e for exchange rate), and the third term is for the crosscovariances between exchange rates and security returns, such as the mark exchange rate with the Japanese equity portfolio return. Chart 1 illustrates the benefits of international diversification using as an example data from 1980 to 1985, a period that will be discussed in detail below. The average annual return and standard deviation of a U.S. stock portfolio is represented by the dot. The efficient frontier generated by combining the U.S. portfolio with stock portfolios from Germany, the United Kingdom, Japan, France, Canada, and Switzerland lies above the U.S. portfolio.9 For the same standard deviation of return, the internationally diversified portfolio offers a higher return. The minimum standard deviation efficient portfolio has a return that is 1 percentage point higher than the standard deviation of the U.S. portfolio and a standard deviation of return that is 2 percentage points lower than the U.S. portfolio’s. In short, investors who evaluate portfolios based on their expected mean returns and standard deviations would choose a portfolio along the efficient frontier. Currency Derivatives and Hedging hile the choice of securities and their degree of diversification fundamentally affects the riskreturn profile of a portfolio, further tailoring of a portfolio’s risk-return characteristics can be achieved through the use of derivative securities. Derivatives are instruments that change the cash flows of a portfolio. This transformation of cash flows alters fluctuations in the market value of a portfolio. Hedging is a transformation of cash flows or market value that the investor regards as reducing the risk of a position. All hedging of securities portfolios considered in this article is implemented using foreign exchange forward contracts. A foreign exchange forward contract is an agreement between two parties to buy (or sell) foreign currency at a future date at an exchange rate determined at the time of the transaction. (In contrast, W Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 CHART 1 Example of an Efficient Frontier for International Equity Portfolios, 1980–85 Unhedged Portfolios Return 20 18 16 U.S. Portfolio 14 13 14 15 16 17 18 19 20 21 22 Standard Deviation a spot contract specifies an immediate exchange of currency at the prevailing exchange rate.) Although other kinds of derivatives can serve the same purpose as forwards, forwards are a simple, cost-effective way to alter the variability of securities portfolio returns.10 Box 1 discusses the costs associated with hedging using forward contracts. Foreign exchange forward contracts are sold by major commercial banks and typically have fixed, short-term maturities of one, six, and nine months. As with many other kinds of derivatives, forward contracts do not involve a net investment upon initiation of a position. Foreign Exchange Forward Contracts. The following example gives a straightforward hedging application, in which all risk is eliminated, and at the same time demonstrates an important arbitrage condition that determines the relationship between forward and spot foreign exchange rates on the one hand and domestic and foreign short-term interest rates on the other. This relationship is useful for understanding the portfolio hedging results. The arbitrage condition known as covered interest parity is given by the following equation: f/s = (1 + rUS)/(1 + rDM), where f and s are forward and spot rates, respectively, expressed in units of dollars per mark, and rUS and rDM are short-term rates of interest in the United States and Germany, respectively (DM for deutsche mark). The meaning of arbitrage condition is clarified as the example is developed. Suppose an investor borrows one dollar and thereby obligates himself to repay 1 + rUS dollars upon maturity of the loan or bond in one month. By converting this borrowing into deutsche marks at the spot exchange rate s, the investor receives 1/s DM per dollar. The investor then buys a one-month German Treasury bill 7. The return on a U.S. domestic securities portfolio is also a local return. 8. This discrete-time formulation is given in Eun and Resnick (1988). Eun and Resnick assume that the investor sells the expected foreign currency proceeds from a foreign investment forward, whereas the examples in the text assume for simplicity that the current value of that investment is sold forward. 9. The efficient frontier is computed by solving the following problem. Suppose there are n securities. Let x be an n 3 1 vector of portfolio weights, m be an n 3 1 vector of mean security returns, and S be an n 3 n covariance matrix of security returns. The efficient portfolio for a target return of µ is determined by finding optimal portfolio weights x* that minimize the varix ′ ∑ x , subject to x ′µ = µ and x ′1 = 1 , where 1 is an n 3 1 vector of ones. An additionance of the portfolio’s returns: min x al constraint is imposed in this study that requires the portfolios’ weights to be nonnegative, that is, the asset portfolios are not permitted to be sold short. Without the nonnegativity constraint, the optimization typically results in improbable or infeasible positions in securities (or portfolios), in particular huge short positions that even most institutional investors cannot assume (Glen and Jorion 1993). The optimal portfolio variance is then given by x * ′ ∑ x *. The efficient frontier is generated by varying the target return µ and solving for the corresponding portfolio variances. 10. One alternative to standard forwards is “quanto” forwards and options. See Rubinstein (1991) and Reiner (1992). With such instruments, the user avoids the quantity risk of forward contract hedges. A less exotic alternative is currency futures contracts traded on futures exchanges. Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 37 B O X 1 The Costs of Hedging he benefits of hedging could potentially be offset by its costs. There are the costs of dealing with a financial intermediary, which provides access to hedging instruments such as forward contracts. These costs depend on the particular instruments. Forward contracts involve commission costs, and transactions at prices or rates reflecting the payment of a bid-ask spread, and possibly the opportunity costs of posting collateral against future losses on the forward position.1 Perold and Schulman (1988) observe that hedging costs are the least significant of the costs associated with international investment. They estimate that rolling over six-month forwards would incur costs reflecting the bid-ask spread and transaction costs of only 0.12 percent per year of the amount invested (Perold and Schulman 1988, 48). A potential cost of the hedge is the forward premium or discount. As discussed in the text, short forward positions are closed out upon maturity of a contract and rolled into new contracts. The new contracts fix the current forward rate for the hedge. If the foreign currency being sold forward is at a discount (because the foreign interest rate exceeds the home country rate), the investor is effectively paying to hedge and the expected return of the investment is reduced by the interest rate differential. The opposite is true of short hedging when the currency is at a forward premium; that is, the interest rate differential can increase the hedged portfolio return.2 The impact on hedged portfolio returns of the cost of carry can be sizable, as will be seen in the next section. Another potential cost is a risk premium implicit in the forward rate. In contrast with the forward discount, this cost is not directly observable. For example, if the forward exchange rate is a downward biased estimate of the T future mark spot rate and an investor wishes to sell marks forward for dollars, the average outcome of entering into such contracts is that fewer dollars per mark would be received through the forward contract than through spot exchanges at the time the forward matures.3 The forward would still give a certain rate of conversion in contrast with the random rate of a future spot transaction, but on average an investor would be paying an implicit premium for a predetermined rate of exchange. Theoretical modeling of the risk premium in forward and futures markets as well as empirical tests of those models have been long-standing research topics (see Hodrick 1987 for a survey of the early literature and Dumas 1996 for more recent studies). Empirically useful characterizations of the risk premium still elude researchers. Most studies find that forwards and futures do not give unbiased estimates of subsequent spot rates; however, linking the estimated bias with variables that measure risk based on theoretical considerations has largely been unsuccessful. The bias fluctuates through time. The literature is substantially in agreement that over long holding periods, typically a few years, the bias and presumably the risk premium are close to zero. This observation is the crux of the argument that currency hedging is a free lunch: hedging delivers a substantial reduction in risk, in the form of a large reduction in the standard deviation of returns, while not entailing the implicit payment of a risk premium. Standard tests for risk premiums applied to the sample used in this article confirm that the average risk premium for each of the six major currencies considered was not statistically significantly different from zero (results available from the authors). 1. Options have similar costs of transacting as well as the payment of the option premium since, unlike forwards, long option positions are net investments. 2. The costs accruing from rolling over hedges can cause serious problems for the hedger if not handled with care. A roll-over hedging strategy used by Metallgesellschaft precipitated a liquidity crisis and eventual bankruptcy of this huge German oil refining and distribution firm. See Culp and Miller (1995). 3. The risk premium could also be collected rather than paid by the forward contract holder. It is not necessarily a cost of using forwards. 38 Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 paying interest of rDM. Such an investment would offer a guaranteed nominal payoff of (1 + rDM)/s DM upon maturity. However, this sum would then have to be translated back into dollars at an unknown future dollarmark exchange rate. The investment would be risky in dollar terms as a result of exchange rate fluctuations. That risk could be eliminated at the outset by a sale of marks for dollars using a forward contract, which specifies a rate of future conversion. The exchange risk is irrelevant with the forward contract because the mark payoff of the investment is translated into dollars at the known forward rate, giving f 3 (1 + rDM)/s dollars toward paying back 1 + rUS dollars of the borrowing. Borrowing at home in dollars and lending abroad in marks must result in equal dollar outcomes on both sides of the transaction; otherwise, investors would seek to exploit or arbitrage the discrepancy because the strategy involves no risk.11 If the dollar receipt of lending exceeds the dollar outlay of borrowing, the foregoing strategy would be undertaken. If the dollar receipt of lending falls short of the dollar outlay of borrowing, the strategy would be reversed, with marks being borrowed and converted into dollar investments. The consequence of this arbitrage pressure is that the following equation must hold: 1 + rUS = f 3 (1 + rDM)/s or f/s = (1 + rUS)/(1 + rDM). The forward exchange rate is determined by the spot exchange rate and the domestic and foreign short-term interest rates for investments with the same maturity as the forward contract. Another way to view this example is from a financial intermediary’s perspective, typically a bank that offers forward contracts to its customers. For example, a bank could enter into a forward contract with a customer who wants to buy marks and sell dollars at the forward rate. The bank could hedge its resulting exposure by borrowing dollars and lending marks to lock in a payment of 1 + rUS dollars and a receipt of f(1 + rDM)/s dollars. In other words, by covered interest parity, no matter what happens to the exchange rate, the bank is guaranteed a later receipt of f(1 + rDM)/s dollars upon expiration of the forward contract and a payment of 1 + rUS dollars. If the mark depreciates against the dollar, the bank gains on its short forward position in marks vis-à-vis its customer but offsets the gain upon translating the mark lending back into dollars. If the mark appreciates, the bank loses on its forward position but recoups the loss by gains on its lending. The bank’s obligation to its customer would be fully covered by these hedging transactions. Portfolio Hedging. In contrast with the covered interest parity example, the investments now under consideration will have maturities or holding periods that are longer than the instruments used to hedge them. For practical reasons, such as reducing the costs of hedging, hedges are adjusted only periodically; consequently, they will be imperfect, leaving an unhedged exposure. As applied in this article, hedging will involve matching a currency hedge with a portfolio in such a way that the full foreign currency exposure of the initial value of the investment position is covered. This type of hedging is someBecause domestic systemtimes called unitary hedging, which has atic risks are likely to proved to be effective differ from country to compared with more country, international sophisticated meth12 ods. A hedged long diversification can further position in foreign secureduce the volatility of rities involves selling portfolio returns by mitithe current foreign currency value of the gating country-specific investment forward for risk. dollars. The investor is said to have a short position in the forward contract—that is, he is obligated to sell foreign currency at the forward rate upon maturity of the contract. As a forward contract matures and is settled with the contract’s counterparty, another forward contract is sold to maintain the hedge for the next, say, three-month period on a continuing underlying foreign asset exposure. This process is called rolling the hedge. The results of single-period hedging can be described using the following notation. The rate of gain (or loss) on the forward contract is fpt–1 – et, where fpt–1 is the forward premium, defined as ft–1/st–1 – 1 (which by covered interest parity is the difference between the domestic and foreign short-term bond yields of the same maturity as the forward). Being short the forward contract implies that a gain accrues to the forward position if the future spot exchange rate at the time the forward matures is below the forward rate. Equivalently, the 11. For simplicity, this example neglects transaction costs and differences in borrowing and lending rates in a given currency. 12. More sophisticated methods that use information derived from joint comovements among forward contract returns and security returns (such as using Japanese yen forwards to hedge mark portfolio exposures) have not yielded better results than the simple full hedging prescription. See Adler and Simon (1986), Eun and Resnick (1988), Kaplanis and Schaefer (1991), and Glen and Jorion (1993); see Anderson and Danthine (1981) and Duffie (1989) for general discussions of hedging predetermined portfolio positions. Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 39 forward position shows a gain if the forward premium exceeds the rate of change in the exchange rate (fpt–1 > et). Note that the forward premium can be negative, which is called a forward discount (a foreign short-term interest rate greater than the domestic). A gain (loss) on the forward position would help to offset a loss (gain) on the unhedged position in the event the foreign exchange rate depreciates (appreciates). The strategy of short hedging the foreign exposure of the initial investment does not perfectly hedge the foreign securities position because the investment result is unhedged. It turns out that the imperfect hedge is not of great consequence, as demonstrated in the empirical discussion, because the magnitude of the hedging error is small. The initial value of the investment is sold forward for gross return of 1 + fpt–1 at time t, and the investment result is converted at the prevailing spot exchange rate, giving a gross dollar return of rlt(1 + et). As noted above, the cross product term rltet, which is the hedging error, is small and is ignored in the discussion to follow. The net dollar return on the hedged portfolio can be interpreted in two equivalent ways: either as the sum of the gross return on the hedged initial foreign investment and the unhedged investment return minus one, (1 + fpt–1) + rlt(1 + et) – 1 ' rlt + fpt–1, or as the sum of the gross return on the unhedged foreign investment and the return to a short forward position minus one, (1 + rlt)(1 + et) + (fpt–1 – et) – 1 ' rlt + fpt–1.13 The hedged dollar return is thus approximated by the local return plus the forward premium. Based on unitary hedges of exposures to each country’s index portfolio, the variance of the hedged diversified portfolio dollar return is σ h ≈ ∑ i =1 ∑ j =1 wi wj ρi jσ iσ j N 2 N l l l (3) 144 42444 3 local return covariances + ∑ ∑ wwρ σ σ 144424443 N N i =1 j =1 fp i j ij fp fp i j forward premium covariances + 2∑ i=1 ∑ j =1 wi wj ρilj, fpσ ilσ jfp , 144424443 N N security return–forward premium covariances where the forward premium standard deviations (with superscript fp for forward premium) and correlation coefficients replace those of the foreign exchange rates that appear in equation (2). The key argument for currency hedging is that the variance reduction by diversifying internationally that may be realized through the first term in equation (2) for local returns may be offset by the contributions of the second two terms for the exchange rate interactions. Foreign exchange rates tend to be more highly correlated than international equity or bond returns.14 40 In contrast, the forward premium has a much lower standard deviation and a lower correlation with local returns than the spot exchange rate. Both of these characteristics may improve the risk-return trade-off for internationally diversified portfolios. Analysis of Unhedged and Hedged Internationally Diversified Portfolios his section evaluates the impact of currency hedging on diversified portfolios of bonds and diversified portfolios of stocks. After a brief overview of the data used to construct internationally diversified portfolios, the effects of currency hedging are assessed by analyzing efficient frontiers for hedged diversified portfolios and unhedged diversified portfolios. Data. Equity and government bond investments in seven countries are considered: Germany, the United Kingdom, Japan, France, Canada, Switzerland, and the United States. Portfolio performance is examined at quarterly intervals, based on portfolio values, spot exchange rates, and three-month forward rates as of the last day of the quarter. The full period runs from first quarter (Q1) 1980 to Q4 1996. The first subperiod, Q1 1980 to Q4 1985, was selected to match or substantially overlap the sample periods in Thomas (1988, 1989), Perold and Schulman (1988), Kaplanis and Schaefer (1991), and Glen and Jorion (1993). The equity returns under consideration are computed from stock indexes compiled by Morgan Stanley Capital International (MSCI) and include the reinvestment of dividends paid during the holding period. The indexes for each country represent portfolios of all listed firms, included in industry proportions that reflect industry composition in the local market. The stocks in the index are weighted by the market capitalization of the included firms, which themselves are drawn from a representative sample of large, medium, and small capitalization firms (MSCI 1995). The government bond index is the government bond subsector of the Salomon Brothers World Bond Index, which is a value-weighted index of bonds with at least one year to maturity. The bond portfolio data cover Q1 1986 to Q4 1996. Coupon payments are reinvested. Three-month forward and spot exchange rates are from Data Resources, Inc. Stock and bond returns are expressed as excess returns by subtracting the three-month Treasury bill yield from U.S. dollar returns and by deducting a foreign country’s three-month risk-free yield from its security returns.15 This adjustment improves the comparability of returns that are computed for multiyear periods and has little effect on the measured standard deviation of return. Equities. Charts 2–5 display the efficient frontiers for internationally diversified equity portfolios for various subperiods. The top panel in each chart shows two T Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 frontiers, one for the unhedged efficient portfolios and the other for the hedged. The unhedged portfolio frontier is derived from the unhedged dollar excess returns during a given period and the hedged portfolio frontier from the hedged dollar excess returns. (All rates shown in the charts and tables are in percent on an annual basis.16) Optimal weights were computed to combine these individual country index portfolios. The left-hand starting point of a frontier represents the minimum standard deviation portfolio’s excess return and standard deviation. The curve moves to the northeast as higher excess return necessitates the addition of greater risk, as approximated by the standard deviation of excess return. The range of excess return and standard deviation of a given frontier reflects all possible efficient outcomes that could be derived from the seven individual country index portfolios that entered the portfolio optimization.17 The dot in this graph is the excess return and standard deviation of the U.S. index portfolio. The optimal portfolio weights appear in the second and third panels of the chart. The second panel gives the weights for the unhedged efficient portfolios, and the third gives them for the hedged efficient portfolios. The country weights are vertical slices of this area plot, which shows how the weights vary continuously from the minimum standard deviation portfolio on the extreme left-hand side to the maximum standard deviation portfolio on the extreme right-hand side. Chart 2 dramatically illustrates what drove currency hedging advocates’ enthusiasm. The hedged portfolio efficient frontier is mostly to the northwest of the unhedged portfolio’s. (This unhedged portfolio efficient frontier, computed using returns rather than excess returns, appeared as Chart 1.) Hedging delivers much higher excess return at substantially lower risk. Note that simply holding the U.S. index was an inefficient choice compared with either type of diversified portfolio. The hedged efficient portfolio frontier does not include the U.S. portfolio. Not surprisingly, the optimization for the portfolio weights mainly selected the high excess return markets of Japan, the United Kingdom, and Germany. The optimization constrained the weights on these portfolios to be nonnegative (that is, selling an index portfolio short was not permitted). However, no constraint was placed on the share of a particular country’s index in the portfolio. In practice, it may not be cost-effective to attempt to take large securities positions in countries with relatively low capitalization equity markets. (Such purchases could raise the cost of shares The impact of exchange if an institutional portfolio manager attemptrate movements on ed to acquire a large unhedged dollar returns position.) can be understood by Table 1 shows the results by country and considering the return in subperiod. Panel A of terms of two sources of Table 1 for unhedged risk, volatility of both forstock index portfolios during 1980–85 indieign security returns and cates that the standard the foreign exchange rate. deviations of the quarterly non-U.S. portfolio unhedged dollar excess returns are all substantially greater than that of the U.S. portfolio. The reason for the volatility of the dollar excess returns is apparent from the rows giving the standard deviations of the foreign exchange returns and the correlation coefficients of the local excess return with the rate of change in the foreign exchange rate. Given the way the variables are measured in this article, the standard deviation of the local excess return is identical to the standard deviation of the hedged dollar excess return.18 The correlation coefficients are between 0.3 and 0.6. The relatively high foreign exchange rate variances and positive local excess return–foreign exchange 13. This second interpretation can also be viewed in terms of an unhedged investment position and a position in domestic and foreign bonds that substitutes for the forward contract. Namely, a short forward position is synthesized by being short foreign bonds (borrowing) and being long domestic bonds (lending), resulting in predetermined payment of foreign currency and receipt of domestic currency. Specifically, from the earlier discussion of covered interest parity, the dollar receipt from lending would be 1 + rUS and the mark payment would be (1 + rDM)/s. 14. During the 1986–96 period, the average correlation coefficient between two countries, excluding Canada, is about 0.6 for equity or bond returns on index portfolios, whereas the average correlation is 0.8 for foreign exchange returns. The average foreign exchange correlation drops to 0.5 when Canada is included. 15. The foreign three-month risk-free rate is estimated by the negative of the difference between the forward premium and the three-month Treasury bill yield. 16. All rates are reported as annualized quarterly logarithmic differences of the variables. Means are annualized and converted to percentages by multiplying by 400; standard deviations by 200 (==43100) 17. The optimization algorithm frequently failed to converge for efficient portfolios that approached the extremes of maximum excess return and maximum standard deviation. These portfolios typically consist of a single index, as seen in the panels for the optimal portfolio weights. 18. The identity occurs when the forward premium is assumed to equal the U.S.–foreign interest rate differential. Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 41 CHART 2 International Equity Portfolios, 1980–85 Efficient Frontiers 16 Excess Return 14 12 U.S. Hedged Unhedged 10 8 6 4 7 9 11 13 15 17 Standard Deviation 19 21 23 Optimal Por tfolio Weights, No Currency Hedging 1.0 U.S. We i g h t 0.8 0.6 0.4 France Japan 0.2 U.K. Germany 0 14.4 15.0 18.0 21.5 Standard Deviation Optimal Por tfolio Weights, Currency Hedging 1.0 France 0.8 We i g h t Japan 0.6 0.4 U.K. 0.2 Germany 0 9.1 9.6 12.2 Standard Deviation Note: Standard deviations on portfolio weight charts are not measured in equal intervals. 42 Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 15.2 CHART 3 International Equity Portfolios, 1986–96 Efficient Frontiers 10 Unhedged Excess Return 9 8 7 U.S. 6 Hedged 5 4 3 2 12.5 13.5 14.5 15.5 16.5 17.5 Standard Deviation Optimal Por tfolio Weights, No Currency Hedging 1.0 U.S. 0.8 We i g h t France 0.6 Switzerland 0.4 Canada U.K. Japan 0.2 Germany 0 12.7 12.8 13.4 17.2 Standard Deviation Optimal Por tfolio Weights, Currency Hedging 1.0 0.8 We i g h t U.S. 0.6 0.4 Canada 0.2 Japan Germany 0 12.8 13.0 13.6 14.4 Standard Deviation Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 43 CHART 4 International Equity Portfolios, 1986–90 Efficient Frontiers 12 Unhedged Excess Return 10 8 6 4 2 U.S. Hedged 0 –2 –4 –6 16 20 24 Standard Deviation 28 32 Optimal Por tfolio Weights, No Currency Hedging 1.0 Switzerland France 0.8 We i g h t Canada 0.6 Japan 0.4 U.K. 0.2 Germany 0 16.9 17.5 19.5 23.7 Standard Deviation Optimal Por tfolio Weights, Currency Hedging 1.0 0.8 We i g h t U.S. 0.6 0.4 Japan Canada 0.2 Germany 0 16.5 17.5 19.2 Standard Deviation Note: Standard deviations on portfolio weight charts are not measured in equal intervals. 44 Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 30.6 CHART 5 International Equity Portfolios, 1991–96 Efficient Frontiers 15 14 Excess Return Unhedged 13 12 11 10 Hedged U.S. 9 8 7 6 5.5 6.5 7.5 8.5 9.5 10.5 11.5 12.5 Standard Deviation Optimal Por tfolio Weights, No Currency Hedging 1.0 0.8 We i g h t U.S. 0.6 Canada 0.4 Switzerland Japan 0.2 Germany 0 5.9 6.0 7.0 11.0 Standard Deviation Optimal Por tfolio Weights, Currency Hedging 1.0 0.8 We i g h t U.S. 0.6 Canada 0.4 Switzerland Japan 0.2 Germany 0 6.5 6.7 7.6 12.5 Standard Deviation Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 45 46 TA B L E 1 S t o c k I n d e x P o r t f o l i o s Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 Panel A: 1980–85 Germany U.K. Japan France Canada Switzerland U.S. Unhedged Stock Index Portfolio Dollar Excess Returns Mean excess return Standard deviation 7.30 4.46 9.95 –0.59 –3.46 3.90 5.42 23.55 20.26 22.12 29.13 24.83 25.06 15.14 — Hedged Stock Index Portfolio Dollar Excess Returns Mean excess return 15.82 11.54 10.69 5.18 –1.28 12.30 Standard deviation 15.74 12.85 13.86 21.20 22.20 16.02 — Percent change in standard deviation of hedged returns to standard deviation of unhedged returns –33.2 –36.6 –37.4 –27.2 –10.6 –36.1 — Perfect Foresight Hedge Dollar Excess Returns Mean excess return 15.65 11.51 10.19 4.54 –1.73 11.75 — Standard deviation 15.39 13.11 14.16 22.08 22.29 15.59 — Forward Premium Mean 4.56 0.06 4.59 –3.36 –0.59 6.54 — Standard deviation 1.04 1.42 1.62 2.59 0.71 1.33 — –0.28 –0.12 –0.12 –0.15 0.12 –0.30 — Correlation between local excess return and forward premium Foreign Exchange Returns Mean return –3.96 –7.02 3.85 –9.12 –2.78 –1.86 — Standard deviation 13.14 12.14 13.18 13.61 4.42 13.68 — 0.30 0.24 0.26 0.34 0.55 0.37 — Correlation between local excess return and foreign exchange return TA B L E 1 S t o c k I n d e x P o r t f o l i o s (cont.) Panel B: 1986–96 Germany U.K. Japan France Canada Switzerland U.S. Unhedged Stock Index Portfolio Dollar Excess Returns Mean excess return Standard deviation 4.21 9.29 3.67 8.64 3.33 8.19 7.96 19.34 18.00 28.35 20.70 14.68 16.32 14.41 Hedged Stock Index Portfolio Dollar Excess Returns Mean excess return Standard deviation Percent change in standard deviation of hedged returns to standard deviation of unhedged returns 0.04 4.49 0.77 2.85 1.17 5.22 — 22.66 17.97 25.70 22.23 13.06 20.99 — 17.2 –0.1 –9.3 7.4 –11.0 28.6 — Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 Perfect Foresight Hedge Dollar Excess Returns Mean excess return Standard deviation 1.30 4.92 0.81 3.58 0.96 6.90 — 21.38 17.19 25.17 21.41 12.81 20.01 — Forward Premium Mean 0.04 –3.26 2.07 –2.35 –1.98 0.95 — Standard deviation 1.58 1.08 1.10 1.43 0.74 1.45 — Correlation between local excess return and forward premium 0.03 0.00 0.13 0.00 0.33 0.00 — 0.18 3.93 — Foreign Exchange Returns Mean return 4.21 1.54 4.97 3.45 Standard deviation 12.83 11.92 13.19 11.82 4.26 14.43 — Correlation between local excess return and foreign exchange return –0.52 –0.33 –0.05 –0.38 0.29 –0.64 — 47 (Continued on page 48) 48 TA B L E 1 S t o c k I n d e x P o r t f o l i o s (cont.) Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 Panel C: 1986–90 Germany U.K. Japan France Canada Switzerland U.S. Unhedged Stock Index Portfolio Dollar Excess Returns Mean excess return Standard deviation 2.67 10.86 11.37 11.34 2.22 1.26 3.96 25.75 22.54 35.57 27.83 18.40 20.45 19.70 — Hedged Stock Index Portfolio Dollar Excess Returns Mean excess return –4.90 1.23 6.17 1.80 –3.90 –5.77 Standard deviation 29.88 23.83 31.95 30.04 16.55 27.05 — 16.0 5.8 –10.2 7.9 –10.0 32.2 — Percent change in standard deviation of hedged returns to standard deviation of unhedged returns Perfect Foresight Hedge Dollar Excess Returns Mean excess return –3.46 1.70 5.64 2.46 –4.26 –3.72 — Standard deviation 27.90 22.21 30.68 28.46 16.09 24.89 — Forward Premium Mean 2.24 –3.85 2.55 –1.60 –2.40 2.52 — Standard deviation 0.72 0.85 0.73 0.92 0.65 0.96 — Correlation between local excess return and forward premium 0.14 –0.03 0.12 –0.06 0.32 0.27 — Foreign Exchange Returns 9.81 5.77 7.75 7.93 3.72 9.55 — Standard deviation Mean return 11.82 11.79 14.33 10.71 3.66 13.43 — Correlation between local excess return and foreign exchange return –0.52 –0.37 0.04 –0.39 0.48 –0.67 — TA B L E 1 S t o c k I n d e x P o r t f o l i o s (cont.) Panel D: 1991–96 Germany U.K. Japan France Canada Switzerland U.S. Unhedged Stock Index Portfolio Dollar Excess Returns Mean excess return 5.50 7.98 –2.75 6.40 4.26 13.97 11.30 Standard deviation 12.28 13.60 20.83 12.59 11.09 11.53 7.81 5.40 14.38 — Hedged Stock Index Portfolio Dollar Excess Returns Mean excess return Standard deviation Percent change in standard deviation of hedged returns to standard deviation of unhedged returns 4.16 7.20 –3.73 3.72 14.59 11.42 19.50 13.34 9.08 13.08 — 18.8 –16.0 –6.4 5.9 –18.1 13.5 — Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 Perfect Foresight Hedge Dollar Excess Returns Mean excess return Standard deviation 5.27 7.61 –3.22 4.51 5.31 15.75 — 14.25 11.86 19.93 13.68 9.06 13.85 — Forward Premium –1.79 –2.77 1.68 –2.97 –1.64 –0.35 — Standard deviation Mean 1.52 1.20 1.32 1.71 0.78 1.48 — Correlation between local excess return and forward premium 0.17 –0.01 0.14 0.07 0.32 0.07 — Foreign Exchange Returns Mean return –0.45 –1.98 2.66 –0.29 –2.78 –0.76 — Standard deviation 13.42 11.99 12.35 12.59 4.23 15.09 — Correlation between local excess return and foreign exchange return –0.60 –0.29 –0.22 –0.50 0.34 –0.68 — 49 rate change covariances contribute to the relatively high standard deviations of many of the unhedged efficient portfolios in Chart 2.19 Nevertheless, the effects of international diversification still result in some efficient unhedged portfolios having lower standard deviations than that of the U.S. domestic portfolio. During the 1980–85 period, the average across all countries for the standard deviation of hedged dollar excess returns was 30.2 percent smaller than the average standard deviation of unhedged dollar excess returns. The hedged dollar excess return standard deviations for Japan and the United Kingdom are lower than the U.S. standard deviation, and none of the remaining standard deviations is as large as the corresponding unhedged dollar excess return The strategy of short hedgvalues. This finding is consistent with those ing the foreign exposure of reported in Thomas the initial investment does (1988), Perold and not perfectly hedge the Schulman (1988), and Kaplanis and Schaefer foreign securities position (1991), including data because the investment that extended back to result is unhedged. 1978. Based on this substantial variance reduction, Perold and Schulman offered this advice: “Our prescription does not say the prescient investor should not selectively lift a hedge, just that hedging should be the policy, and lifting the hedge an active investment decision” (1988, 45). The effectiveness of currency hedging using forwards is apparent from the results under the heading Perfect Foresight Hedge Dollar Excess Returns. These measures of the mean excess return and standard deviation reflect a currency hedge that was scaled to match the ex post quarterly security return. That is, rather than matching the hedge to the initial beginning-ofquarter portfolio value, the hedge was constructed to match the end-of-quarter portfolio value. In this way, the perfect foresight hedge covers the investment return, which is not known with certainty in practice. In this panel as well as in the others to follow, the unitary hedge results are close to those of the perfect foresight hedge, especially relative to the results derived from unhedged positions. Another point to notice is that, from equation (3), the hedged portfolio standard deviation depends on the variance of the forward premium as well as on the covariance of the forward premium with the local excess return. The forward premium standard deviations are an order of magnitude smaller than the foreign 50 exchange return standard deviations. The correlation coefficients for the forward premium with the local excess return are negative, a fact that also contributes to reducing the hedged portfolio variance. In Chart 2, most hedged efficient portfolios have a lower standard deviation than the minimum standard deviation unhedged portfolio. The mean annual rates of dollar excess return for the unhedged country portfolios are determined approximately by the sum of the average local return for the period and average rate of foreign exchange appreciation less the risk-free rate of interest. The dollar’s appreciation depressed unhedged relative to hedged dollar excess returns. As discussed above, the hedged dollar return is approximately the local return plus the forward premium, implying, by covered interest parity, that the hedged dollar excess return is approximately the local excess return—for example, rl + (rUS – rDM) – rUS = rl – rDM. All mean unhedged dollar excess returns are less than the corresponding mean hedged dollar excess returns. (The same is true of Japan, whose currency appreciated against the dollar during this period, because the Japanese risk-free rate was much less than the U.S. risk-free rate.) Chart 3 for 1986 to 1996 presents an entirely different picture of the currency hedging argument. The unhedged efficient frontier dominates the hedged frontier. The case for currency hedging of internationally diversified equity portfolios has not held up because of the instability of the covariance structure, that is, the variability through time of standard deviations and correlation coefficients of excess returns. Most striking in Panel B of Table 1 is the standard deviations of hedged dollar excess returns across countries. Only Canada has a lower standard deviation of excess return compared with the United States, and consequently the hedged portfolio frontier was generated mainly by positions in the Canadian and U.S. portfolios. (In the 1980–85 subperiod, only two countries, France and Canada, have substantially higher volatility than the United States.) At the same time, the United States has the highest mean hedged dollar excess return during 1986–96, making it the endpoint of the hedged portfolio frontier. Another related point to note in Panel B for the 1986–96 subperiod is that the correlation coefficients of the unhedged dollar excess returns and foreign exchange returns show a reversal of signs for all coefficients except Canada’s compared with the corresponding values in Panel A for 1980–85. The negative correlation and relative increase in foreign market volatility translate into unhedged dollar excess return standard deviations and hedged dollar excess return standard deviations that are much closer in size relative to the values in Panel A. The combined effects of the Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 apparent changes in the covariance structure between these two periods account for the reversed positions of the unhedged and hedged efficient portfolio frontiers in Charts 2 and 3. During the 1986–96 subperiod, foreign currencies generally appreciated against the dollar. Unhedged foreign portfolio investments did well ex post because investment proceeds converted into more and more dollars over time. By limiting the dollar excess return to the local excess return, hedging stripped out the positive foreign exchange return while largely exposing the investor to higher local excess return volatility. Charts 4 and 5 display the subperiod efficient frontiers for 1986–90 and 1991–96, respectively. Qualitatively, the results are similar for the combined period. The same is true of efficient portfolios for either 1986–96 or 1986–90 that exclude the quarter containing the 1987 crash (these graphs are not reported). The large portfolio weight on Canada, with its –3.9 percent hedged dollar excess return, pushes the minimum variance hedged portfolio’s standard deviation close to –4 percent in Chart 4 for 1986–90. (Of course, if a negative excess return were expected by investors ex ante, no one would hold the portfolio.) Corresponding to Charts 4 and 5, Panels C and D, respectively, of Table 1 split the 1986–96 period into subperiods as a check on the stability of the portfolio excess returns, standard deviations, and correlations. The 1986–90 subperiod includes the crash of 1987, which was a global phenomenon. The greatest negative equity returns occur in each country in the fourth quarter of 1987. As documented in Panel C, the United States actually had the second least volatile equity market during this subperiod. This subperiod was also the time of the most rapid depreciation of the dollar against the currencies of the six countries, as shown in Panel C. In Panel D for 1991–96, equity market volatility subsided but foreign exchange market volatility increased compared with the earlier subperiod. The United States had the least volatile equity market. The correlation of local excess returns and foreign exchange returns is generally negative across these subperiods. Although there is some variation in the correlation coefficients in these subperiods, the correlation structure is distinctly different from what it was during the 1980–85 period with the exception of Canada, which has a relatively large positive correlation in each period. Bonds. The bond portfolio results differ markedly from the equity portfolio results. Although the Salomon Brothers bond indexes are only available as of 1985, the results here are consistent with those from other stud- ies that relied on different indexes from earlier periods. The full period ran from 1986 to 1996 to match the same period used for equities. The standard deviations of the hedged dollar excess returns in Panel A of Table 2 for the 1986–96 period are much smaller than those for the equity index dollar excess returns, whether hedged or unhedged, or those for foreign exchange returns. The correlation coefficients for foreign exchange returns and local excess returns are positive for five of six countries and all except Japan’s are small in magnitude. A comparison of the hedged dollar excess returns with the perfect foresight hedge dollar excess returns reveals that quarterly hedging closely matches the perfect foresight case, especially for the standard deviation of Putting the practice of curexcess return. rency hedging on a firmer The standard deviations of unhedged dolfoundation requires better lar excess returns are models and techniques for always substantially predicting the correlation larger than those for hedged dollar excess structure. returns. Exposure to foreign exchange rate fluctuations contributes disproportionately to the risk of holding foreign bonds. The average reduction across countries in the standard deviation of hedged dollar excess returns relative to the unhedged excess returns standard deviation is 55.6 percent. This decline is about the same as or somewhat greater than the magnitudes for similar measures reported in Perold and Schulman (1988), Thomas (1989), and Kaplanis and Schaefer (1991) for the years from 1975 to 1988 or subperiods within that span of years. Panels B and C show that the lower variability of hedged excess returns also occurs for the 1986–90 and 1991–96 subperiods. The efficient bond frontiers generated by the excess returns for hedged and unhedged bond positions are shown in Charts 6–8. The full-period results appear in Chart 6. The hedged portfolio frontier lies to the southwest of the unhedged portfolio frontier and intersects the unhedged portfolio frontier just above the point for the U.S. bond portfolio, which is the minimum variance portfolio for the unhedged portfolio frontier. Unitary hedging of foreign exchange exposures has a pronounced impact on the standard deviation of dollar excess return. The configuration of these 19. The contribution of these elements of the unhedged efficient portfolio variance is apparent from equation (2). Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 51 52 TA B L E 2 B o n d I n d e x P o r t f o l i o s Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 Panel A: 1986–96 Germany U.K. Japan France Canada Switzerland U.S. Unhedged Bond Index Portfolio Dollar Excess Returns Mean excess return Standard deviation 5.78 6.46 5.95 7.85 4.92 3.65 2.78 13.38 15.16 16.27 12.31 8.35 14.57 5.34 Hedged Bond Index Portfolio Dollar Excess Returns Mean excess return 1.61 1.66 3.05 2.05 2.76 0.68 — Standard deviation 3.72 7.84 5.97 5.37 6.46 4.27 — –72.2 –48.3 –63.3 –56.4 –22.6 –70.7 — Percent change in standard deviation of hedged returns to standard deviation of unhedged returns Perfect Foresight Hedge Dollar Excess Returns Mean excess return 1.51 1.37 2.73 1.96 2.63 0.69 — Standard deviation 3.69 7.70 5.78 5.25 6.46 4.33 — 2.07 –2.35 –1.98 0.95 — Forward Premium Mean 0.04 –3.26 Standard deviation 1.58 1.08 1.10 1.43 0.74 1.45 — Correlation between local excess return and forward premium 0.01 –0.03 –0.08 –0.10 0.09 0.02 — 0.18 3.93 — Foreign Exchange Returns Mean return Standard deviation Correlation between local excess return and foreign exchange return 4.21 1.54 4.97 3.45 12.83 11.92 13.19 11.82 4.26 14.43 — 0.02 0.14 0.29 –0.12 0.17 –0.14 — TA B L E 2 B o n d I n d e x P o r t f o l i o s (cont.) Panel B: 1986–90 Germany U.K. Japan France Canada Switzerland U.S. Unhedged Bond Index Portfolio Dollar Excess Returns Mean excess return Standard deviation 7.24 8.48 5.23 10.09 5.86 4.92 2.03 13.70 17.76 19.19 13.01 8.19 15.03 5.72 Hedged Bond Index Portfolio Dollar Excess Returns Mean excess return –0.33 –1.15 0.03 0.56 –0.26 –2.11 — Standard deviation 3.79 9.11 7.29 6.20 6.47 3.57 — –72.3 –48.7 –62.0 –52.4 –20.9 –76.2 — Percent change in standard deviation of hedged returns to standard deviation of unhedged returns Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 Perfect Foresight Hedge Dollar Excess Returns Mean excess return –0.58 –1.85 –0.55 0.24 –0.46 –2.32 — Standard deviation 3.69 8.75 7.02 5.91 6.40 3.49 — Forward Premium Mean 2.24 –3.85 2.55 –1.60 –2.40 2.52 — Standard deviation 0.72 0.85 0.73 0.92 0.65 0.96 — Correlation between local excess return and forward premium 0.33 –0.03 –0.16 –0.34 0.04 0.33 — 3.72 9.55 — Foreign Exchange Returns Mean return Standard deviation Correlation between local excess return and foreign exchange return 9.81 5.77 7.75 7.93 11.82 11.79 14.33 10.71 3.66 13.43 — 0.33 0.40 0.48 0.06 0.26 0.26 — 53 (Continued on page 54) 54 TA B L E 2 B o n d I n d e x P o r t f o l i o s (cont.) Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 Panel C: 1991–96 Germany U.K. Japan France Canada Switzerland U.S. Unhedged Bond Index Portfolio Dollar Excess Returns Mean excess return Standard deviation 4.56 4.77 6.55 5.98 4.14 2.59 3.40 13.37 12.94 13.80 11.90 8.65 14.48 5.11 Hedged Bond Index Portfolio Dollar Excess Returns Mean excess return 3.22 3.99 5.57 3.30 5.28 3.00 — Standard deviation 3.53 6.58 4.37 4.61 6.32 4.52 — –73.6 –49.1 –68.3 –61.3 –27.0 –68.8 — Percent change in standard deviation of hedged returns to standard deviation of unhedged returns Perfect Foresight Hedge Dollar Excess Returns Mean excess return 3.24 4.05 5.46 3.40 5.21 3.19 — Standard deviation 3.53 6.59 4.19 4.63 6.35 4.62 — 1.68 –2.97 –1.64 –0.35 — Forward Premium Mean –1.79 –2.77 Standard deviation 1.52 1.20 1.32 1.71 0.78 1.48 — Correlation between local excess return and forward premium 0.19 –0.11 0.06 0.07 0.03 0.15 — –2.78 –0.76 — Foreign Exchange Returns Mean return –0.45 –1.98 2.66 –0.29 Standard deviation 13.42 11.99 12.35 12.59 4.23 15.09 — Correlation between local excess return and foreign exchange return –0.13 –0.07 0.07 –0.26 0.30 –0.30 — CHART 6 International Bond Portfolios, 1986–96 Efficient Frontiers 8 Excess Return 7 6 Unhedged 5 4 3 2 Hedged U.S. 1 0 3 5 7 9 Standard Deviation 11 13 Optimal Por tfolio Weights, No Currency Hedging 1.0 We i g h t 0.8 0.6 U.S. France Canada 0.4 0.2 0 5.3 6.1 8.1 12.1 Standard Deviation Optimal Por tfolio Weights, Currency Hedging 1.0 0.8 U.S. Switzerland France We i g h t 0.6 0.4 Japan Germany 0.2 0 3.7 3.8 4.5 5.5 Standard Deviation Note: Standard deviations on portfolio weight charts are not measured in equal intervals. Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 55 CHART 7 International Bond Portfolios, 1986–90 Efficient Frontiers 10 Hedged U.S. Unhedged Excess Return 8 6 4 2 0 –2 3 5 7 11 9 13 Standard Deviation Optimal Por tfolio Weights, No Currency Hedging 1.0 U.S. We i g h t 0.8 0.6 Canada 0.4 Switzerland 0.2 France 0 5.7 6.2 8.5 12.7 Standard Deviation Optimal Por tfolio Weights, Currency Hedging 1.0 U.S. 0.8 We i g h t Switzerland 0.6 France 0.4 Germany 0.2 0 3.5 4.6 3.7 Standard Deviation Note: Standard deviations on portfolio weight charts are not measured in equal intervals. 56 Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 5.6 CHART 8 International Bond Portfolios, 1991–96 Efficient Frontiers Excess Return 7 6 Unhedged Hedged 5 U.S. 4 3 2 4 6 8 10 12 14 Standard Deviation Optimal Por tfolio Weights, No Currency Hedging 1.0 0.8 We i g h t U.S. Canada France 0.6 0.4 Japan 0.2 0 5.1 6.1 13.0 8.0 Standard Deviation Optimal Por tfolio Weights, Currency Hedging 1.0 U.S. Switzerland Canada We i g h t 0.8 0.6 0.4 Japan Germany 0.2 0 3.4 3.5 3.9 4.4 Standard Deviation Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 57 frontiers implies that the decision to hedge depends on an investor’s preference for risk and return; high returns are attainable only through unhedged positions and conversely low-risk returns only through hedged positions. The comparatively more volatile 1986–90 subperiod predominates in the full sample as the efficient frontiers and the optimal weights that compose them are similar. As is evident from Chart 7 and Table 2, Panel B, Switzerland’s heavy weight in the minimum variance hedged portfolio is mainly responsible for the negative excess return. In contrast, for the 1991–96 period in Chart 8, the hedged efficient frontier dominates the unhedged, except for the high levels of excess return above about 5.6 percent. In all three cases, the U.S. bond portfolio is, or is very nearly, the minimum variance unhedged efficient portfolio, which reflects the high volatility of unhedged foreign bond portfolios.20 Conclusion his analysis of efficient portfolios of stocks and bonds only partially confirms the claims of the proponents of currency hedging. Simple unitary hedging consistently yields a low standard deviation of excess return on efficient, internationally diversified bond portfolios. This finding agrees with those of other studies of internationally diversified bond portfolios. Whether this result is optimal depends on investor preferences. For hedged and unhedged portfolios from 1986 T to 1996, the efficient frontiers corresponding to hedged and unhedged positions partition the excess return and standard deviation outcomes, with low excess return, low standard deviation results for hedged portfolios and high excess return, high standard deviation results for unhedged portfolios. Internationally diversified equity portfolios do not show the same gains or consistency of results. The risk reduction achieved through currency hedging found in several earlier studies is confirmed for the 1980–85 subperiod. The efficient frontier for hedged portfolios lies far to the northwest of the frontier for unhedged portfolios. However, this hedged portfolio performance is reversed in the 1986–96 period. Another way to state this finding is that the covariance structure of international equity excess returns was unstable—it was subject to a large shift that drastically altered hedging outcomes. Earlier articles on currency hedging, and especially the unitary hedging prescription, were predicated on an implied confidence in the stability of the covariance structure of security and foreign exchange returns. Putting the practice of currency hedging on a firmer foundation requires better models and techniques for predicting the correlation structure (see especially King, Sentana, and Wadhwani 1994 and Solnik, Boucrelle, and Le Fur 1996). Developing these tools is clearly a challenge that calls for continuing research. 20. The low standard deviations of hedged efficient bond portfolios relative to the standard deviation of the U.S. bond portfolio was confirmed by an alternative procedure. As discussed in note 9, the optimal portfolio weights and the efficient frontier itself depend on the ex post security return. Instead of computing weights by optimization, weights were randomly drawn from a uniform distribution and their sum normalized to one. Hedged diversified portfolios were then formed using these weights. Of 10,000 such randomly weighted hedged portfolios, fewer than 2 percent had standard deviation greater than that of the U.S. portfolio in the 1986–96 period. A similar procedure confirmed the results for stock portfolios shown in Charts 2 and 3. Fewer than 1 percent of randomly weighted hedged stock portfolios had standard deviation greater than that of the U.S. portfolio in 1980–85. In contrast, more than 99 percent of randomly weighted hedged stock portfolios in 1986–96 had standard deviation greater than the U.S. portfolio’s. The remaining fraction had large weights on the U.S. and Canadian stock portfolios and very small weights on the other countries’ portfolios, consistent with the third panel of Chart 3. 58 Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 REFERENCES ADLER, MICHAEL, AND DAVID SIMON. 1986. “Exchange Risk Surprises in International Portfolios.” Journal of Portfolio Management 12 (Winter): 44–53. KAPLANIS, EVI, AND STEPHEN M. SCHAEFER. 1991. “Exchange Risk and International Diversification in Bond and Equity Portfolios.” Journal of Economics and Business 43:287–308. ANDERSON, RONALD, AND J.P. DANTHINE. 1981. Journal of Political Economy 89 (December): 1182–96. KING, MERVYN, ENRIQUE SENTANA, AND SUSHIL WADHWANI. 1994. “Volatility and Links between National Stock Markets.” Econometrica 62 (July): 901–33. CHOW, EDWARD H., WAYNE Y. LEE, AND MICHAEL E. SOLT. 1997. “The Exchange-Rate Risk Exposure of Asset Returns.” Journal of Business 70:105–23. LEVICH, RICHARD M., AND LEE R. THOMAS III. 1993. “Internationally Diversified Bond Portfolios: The Merits of Active Currency Management.” NBER Working Paper No. 4340, April. CULP, CHRISTOPHER L., AND MERTON H. MILLER. 1995. “Metallgesellschaft and the Economics of Synthetic Storage.” Journal of Applied Corporate Finance 7 (Winter): 62-76. MARKOWITZ, HARRY. 1952. “Portfolio Selection.” Journal of Finance 7 (March): 77–91. DUFFIE, DARRELL. 1989. Future Markets. Englewood Cliffs, N.J.: Prentice Hall. DUMAS, BERNARD. 1996. “Partial Equilibrium versus General Equilibrium Models of the International Capital Market.” In The Handbook of International Macroeconomics, edited by Frederick Van der Ploeg. Cambridge, Mass.: Blackwell. EAKER, MARK R., AND DWIGHT M. GRANT. 1991. “Currency Risk Management in International Fixed-Income Portfolios.” Journal of Fixed Income (December): 31-37. EUN, CHEOL S., AND BRUCE G. RESNICK. 1988. “Exchange Rate Uncertainty, Forward Contracts, and International Portfolio Selection.” Journal of Finance 43 (March): 197–215. ———. 1994. “International Diversification of Investment Portfolios: U.S. and Japanese Perspectives.” Management Science 40 (January): 140–61. GLEN, JACK, AND PHILIPPE JORION. 1993. “Currency Hedging for International Portfolios.” Journal of Finance 48 (December): 1865–86. HODRICK, ROBERT J. 1987. The Empirical Evidence on the Efficiency of Forward and Futures Foreign Exchange Markets. Chur, Switz.: Harwood Academic Publishers. MORGAN STANLEY CAPITAL INTERNATIONAL. 1995. Index Data Dictionary. New York: DRI/McGraw-Hill. PEROLD, ANDRÉ F., AND EVAN C. SCHULMAN. 1988. “The Free Lunch in Currency Hedging: Implications for Investment Policy and Performance Standards.” Financial Analysts Journal (May/June): 45–50. REINER, ERIC. 1992. “Quanto Mechanics.” Risk 5 (March): 59–62. RUBINSTEIN, MARK. 1991. “Two into One.” Risk 49 (May): 49. SOLNIK, BRUNO, CYRIL BOUCRELLE, AND YANN LE FUR. 1996. “International Market Correlation and Volatility.” Financial Analysts Journal 52 (September/October): 17–34. THOMAS, LEE R. 1988. “Technical Notes: Currency Risks in International Equity Portfolios.” Financial Analysts Journal (March/April): 68–71. ———. 1989. “The Performance of Currency-Hedged Foreign Bonds.” Financial Analysts Journal (May/June): 25–31. Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 59 Common Trends and Cycles and the Structure of Florida’s Economy E D G A R PA R K E R The author is an analyst in the regional section of the Atlanta Fed’s research department. He thanks David Avery, Zsolt Becsi, Tom Cunningham, Robert Eisenbeis, Frank King, Whitney Mancuso, William Roberds, Gus Uceda, and Tao Zha for helpful conversations and comments on earlier drafts. F LORIDA, LIKE THE REST OF THE NATION, HAS UNDERGONE MANY ECONOMIC CHANGES IN THE LAST QUARTER-CENTURY. SOME OBVIOUS EXAMPLES OF THIS ONGOING EVOLUTION ARE THE DECLINE OF THE MANUFACTURING BASE AND THE GROWTH OF INTERNATIONAL TRADE. IN THE CASE OF FLORIDA SPECIFICALLY, THE GROWTH OF THE IMPORTANCE OF TOURISM HAS ALSO FIG- URED SIGNIFICANTLY. IN ADDITION TO CHANGES AT THE STATE LEVEL, FLORIDA’S CITIES HAVE BECOME LESS SIMILAR OVER TIME. AS MIGHT BE EXPECTED, THESE GRADUAL ECONOMIC CHANGES COULD AT SOME POINT CAUSE THE STATE’S METRO ECONOMIES TO INTERACT IN NEW WAYS. FOR EXAMPLE, LABOR MARKETS THAT MAY HAVE BEEN VERY SIMILAR IN STRUCTURE AND BEHAVIOR IN ONE PERIOD MAY HAVE BECOME MORE HETEROGENEOUS IN A LATER PERIOD. Such changes in the structure of a regional economy have implications for economic forecasters, policymakers, businesses, and the general public. The ultimate effects of economic shocks on a region depend on the ways different parts of that region are linked to each other and to external areas, as well as the region’s relative degree of homogeneity. A particular economic policy or shock may have a completely different effect on a highly homogeneous region than it would on a more heterogenous one. This article uses multiple cointegration and common cycles analysis to study the evolution of the relationships among some major Florida cities’ labor 60 markets. (See the glossary on page 66 for short discussions of the technical terms.) Cointegration analysis is used to examine the degree and type of long-run relationships that exist in these labor markets. This analysis is extended with the introduction of the common cycle methodology (of Vahid and Engle 1993) to illustrate the short-run dynamics of the labor markets studied. Cointegration ointegration analysis deals with long-run equilibrium relationships among economic variables. When a group of variables move together in a common way over time they may be cointegrated––that C Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 is, influenced by a common (random) trend. This comovement can be caused by economic links that tie the variables together in a long-term bond. Economic theory is used to suggest variables to test for cointegration. Examples may include strongly linked variables such as consumption and income or the levels of total payroll employment among metropolitan statistical areas in a homogeneous, well-integrated state economy. Once economic theory suggests a list of variables that may be cointegrated, statistical tests such as the EngleGranger (1987) test and the Johansen (1995) test can be used to determine formally if a group of variables is cointegrated. This article suggests that the labor markets within six Florida metropolitan statistical areas (MSAs) may share a cointegrating relationship. Cointegration analysis can also reveal the response of particular labor markets to shocks in other labor markets. For example, changes in labor demand and supply in one MSA can be transmitted to another. Such information, by helping determine which MSAs are more independent of one another (weakly exogenous) and which react strongly to disturbances in surrounding markets (endogenous), can be valuable for clarifying how the effects of state level policy changes as well as economic shocks are transmitted among individual MSAs. The Florida MSAs studied are the six largest: Fort Lauderdale, Jacksonville, Miami, Orlando, Tampa, and West Palm Beach. The study of their labor markets began with collecting the seasonally adjusted monthly levels of total nonagricultural payroll employment from January 1970 to June 1996. The data were tested over the entire time period for a cointegrating (or long-run equilibrium) relationship among the MSAs. The hypothesis of a cointegrating relationship over this time period is not rejected. Even when they are governed by the same basic factors, however, economic relationships change over time. For this reason, the stability of the relationships over the entire sample period was examined using a rolling regression. The results are presented in the first panel of Chart 1. The number 1 on the vertical axis represents the 5 percent level of significance. At points above this line the hypothesis that the equilibrium relationship of the entire time period studied is the same as the subperiods (or the cointegrating vectors of the full sample are the same as those of the subsample) is rejected. The first panel of Chart 1 shows that the full sample can be divided into three subperiods. The first, 1970 to 1980:06, is a period of rejection of the hypothesis that the full-sample cointegrating vectors are those of the subsample. Next appears a subsample that suggests increasing acceptance of the stability of the coefficients of the cointegrating vectors over the period from 1980:07 to 1987:12. Finally, there is a period of high acceptance of the null hypothesis, from 1988:1 to 1996:06. The stability tests suggest that the relationships among the labor markets of the MSAs change over time. Next, tests are applied to these subperiods. First, the sample of the period from 1970:01 to 1980:06 is studied to determine which MSAs are included in the longrun equilibrium and which are weakly exogenous. Then an observation is dropped and the cointegrating relationship is examined again. This process was continued for a three-year period. It was found that a stable period in the cointegrating relationships Florida’s cities have from 1970:01 to 1978:08 become less similar (with all cities includover time. As might be ed in the cointegrating relationship test and expected, these gradual with Miami, Tampa, economic changes could and West Palm Beach at some point cause the found to be weakly exogenous) was interstate’s metro economies rupted by a period of to interact in new ways. transition beginning around 1978:09. The data show that the point of division indicated by the stability test is a time period of relatively dramatic change that begins one to two years before the actual dividing date of 1980:06. An appropriate end date to use in sampling the first period should therefore be shortly before this transition period. August 1978 was chosen because it is the month just before changes in weak exogeneity among the MSAs occur. The same rolling regression technique used in the original sample was used to test this subperiod. The second panel of Chart 1 shows that the hypothesis that the cointegrating relationship for the period from January 1970 to August 1978 is the same over subperiods of this sample is accepted over most of the time period. The results of tests of long-run exclusion and weak exogeneity for this subperiod are shown in Table 1; all MSAs are included in the long-run equilibrium relation, as the hypothesis of exclusion is rejected. The table also shows that Miami, Tampa, and West Palm Beach are weakly exogenous. Next, moving past the unstable 1980:06–1987:12 transition period indicated in the first panel of Chart 1, the months spanning the last time period are examined. As indicated in the chart, this is the region of high acceptance of the original cointegrating relationship. The dividing date appears to be early 1988, and thus the sample period is from January 1988 to June 1996. As before, tests of the robustness of the cointegrating relationships Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 61 CHART 1 Tests of the Long-Run Relationships Among Labor Markets in Six Florida MSAsa January 1970–June 1996 Significance Level 5 3 1 1973b 1978 1983 1988 1993 January 1970–August 1978 Significance Level 5 3 1 Jan-73 b Sep-74 May-76 Jan-78 January 1988–June 1996 Significance Level 5 3 1 Jan-91c Sep-92 May-94 Jan-96 Note: On the y axis, “1” indicates a 5 percent significance level. Data are seasonally adjusted (by the Federal Reserve Bank of Atlanta) monthly levels of total nonagricultural payroll employment. a b c Miami, Orlando, West Palm Beach, Fort Lauderdale, Tampa, and Jacksonville First observation is the result of the initial sample period, January 1970–January 1973. First observation is the result of the initial sample period, January 1988–January 1991. Source: Bureau of Labor Statistics 62 Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 TA B L E 1 Chi-Squar e Tests, Labor Market Data for Six Florida MSAs, January 1970–August 1978 Critical Value Miami Orlando West Palm Beach Fort Lauderdale Tampa Jacksonville 12.21 19.53 20.95 9.30 1.48 8.09 Long-Run Exclusion 5.99 9.16 14.39 17.74 Weak Exogeneity 5.99 0.70 17.54 1.32 TA B L E 2 Chi-Squar e Test, Labor Market Data for Six Florida MSAs, January 1988–June 1996 Critical Value Miami Orlando West Palm Beach Fort Lauderdale Tampa Jacksonville 10.45 22.78 14.10 Long-Run Exclusion 9.49 8.48 8.59 17.72 are performed by sampling around the transition point. The results of this period are much less robust than in the first era, perhaps because of increased linkages of all Florida MSAs to regions outside the state and less homogeneity (more specialization) among the MSAs. Miami is always excluded from the cointegrating relationship. Orlando is excluded in most of the time periods around the transition. These findings support the thesis that Florida’s economy has become less integrated, apparently beginning around 1988. There is strong evidence that Miami and Orlando are excluded from the long-run equilibrium relationship, as shown in Table 2. There is also strong evidence that these two MSAs and Fort Lauderdale are weakly exogenous. This condition indicates that these three MSAs not only do not move with the others over the 1988:01–1996:06 period but they are also insulated from short-run shocks in the rest of the state. Just as before, the stability of the cointegrating relationship is tested over this period using the rolling regression and chisquare tests. The third panel of Chart 1 shows that the observed long-run relationship is stable over most of the last sample period. The above analysis suggests that for some reason the relationship between the MSAs changed over the sample period. Initially the levels of total payroll employment in the cities grew together in a cointegrated relationship. The nature of this relationship then changed, and the MSAs became less bound by the longrun equilibrium relationship. What could have caused this apparent change in behavior? The concepts of temporary cointegration and sudden change as introduced by Siklos and Granger (1996) and Krugman (1991, 26), respectively, may help shed light on the observed relationships. Siklos and Granger use the concept of temporary cointegration to describe data for which the underlying series need not be cointegrated at all times. The relationship shown over one time span may be different from that of another period. This change in the long-run equilibrium relationship might be expected if there are changes in the makeup of particular MSAs over time, leading to possible differences in the demand for and supply of labor in each MSA. The concept of sudden change offers another possible explanation for why the relationships between the MSAs became less cointegrated. Krugman (1991, 26) describes sudden change as the result of a gradual and unnoticed change in the underlying conditions that leads to an explosive apparent change. A likely explanation is that the gradual transitions of Florida MSAs, Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 63 Miami and Orlando in particular, as they became increasingly linked to economic regions outside Florida and internally more heterogeneous, fragmented the state’s economic integration. The growth of tourism in Orlando and foreign trade in Miami have driven the significant changes in these labor markets. In testing for the gradual changes in the MSAs that may have created the new relationships, location quotients are useful. Location quotients indicate the relative concentration of a particular industry in a region. In this study location quotients are constructed using total payroll earnings. They are computed by dividing the percentage of total payroll earnings generated by a particular industry in an MSA by the percentage of the industry’s total payroll earnings at the state level. A location quotient equal to 1 indicates that total payroll earnings in this industry are as concentrated in the studied MSA as they are in the state as a whole. If greater than 1 the location quotient shows greater concentration in the MSA than at the state level, and if less than 1, less relative concentration. The location quotients are consistent with the hypothesis that increased specialization in tourism in Orlando and trade in Miami have led to the breakup of the cointegrating relationship that held the MSAs together. The location quotients identify some gradual changes in the underlying economic structure that may have resulted in sudden change. Growth in import and export activity through the port of Miami are taken to reflect growth in international trade links. For the Orlando area (Orange County) the hotel and service sector is a proxy for tourism-related activities. The water transportation location quotient for the Miami area (Dade County) from 1969 to 1994 depicted in Chart 2, shows the rise from an above-average concentration of water transportation in 1969 to the extremely high level of about three times that of the state at the end of the period. The increasing concentration of water transportation in Miami’s economy clearly shows Miami’s emerging trade links with the rest of the world gradually growing and helping pull Miami out of its cointegrating relationship with the rest of the state. Miami is now the seventh-busiest container port in the United States as well as the number-one cruise port in the world. The location quotients of Orlando’s service sector, measured by sector payroll earnings, tell a similar story about tourism-related growth in that area in Chart 3. In 1969 Orlando was similar to the state in concentration of its service sector. This situation changes over the sample period as this concentration gradually grows to nearly twice the level in the state. Nationally, Orlando is second only to Las Vegas when ranked by the relative percentage of service-sector employment in its economy. Location quotients of hotel total payroll earnings were calculated to further examine the emergence of tourism-related activities in the Orlando area. Although these data are incomplete (the data for hotel payroll earnings exist only from 1985 to 1987 and 1993 to 1994), in Chart 4 it can be seen that the Orlando area already had a high concentration of hotel payroll earnings in 1985 relative to the rest of the state. This concentration continued to grow to more than five times the state’s level by 1994. It seems reasonable to assume that the CHART 2 Dade County Water Transpor tation Location Quotient, 1969–94 3.5 Location Quotient 3.0 2.5 2.0 1.5 1.0 1970 1975 1980 1985 Source: Bureau of Economic Analysis, provided by Regional Financial Associates 64 Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 1990 concentration of hotel earnings in Orlando was much lower in 1970. These dramatic rises in service and hotel-related total payroll earnings indicate the growing importance of tourism in the Orlando area, linking its economy to areas outside of Florida as well as differentiating it from the rest of the state. This emerging link helped remove Orlando from the cointegrating relationship of the early time period. The changing level of stability in cointegrating relationships reveals periods of economic structural change in the labor forces of the Florida MSAs studied. What began as a high degree of cointegration began to lessen by the last period as Orlando and Miami became excluded. As Siklos and Granger state, “It seems realistic to assume that some series are cointegrated only during some periods and not at others. The reason is that events or important changes in some of the CHART 3 Orange County Service-Sector Location Quotient, 1969–94 Location Quotient 2.2 1.8 1.4 1.0 1970 1975 1980 1985 1990 CHART 4 Orange County Hotel Payroll Earnings Location Quotient, 1985–94 Location Quotient 6 4 2 0 1985 1986 1987 1993 1994 Note: Data are unavailable for 1988–92. Source for Charts 3 and 4: Bureau of Economic Analysis, provided by Regional Financial Associates Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 65 Glossary Cointegration between economic variables may exist if these variables tend to move together in a common way over time. Economic theory may suggest which variables to test for cointegration––for example, strongly linked variables such as consumption and income or the levels of total payroll employment among MSAs in a homogenous, well-integrated state economy. Common cycles refers to the short-run dynamics of the time series. In this article the decomposition of the levels of total nonagricultural payroll employment reveals the effects of short-run shocks to the group of MSAs. Common trends refers to the long-run behavior of the levels of total nonagricultural employment in the MSAs. This long-run behavior is revealed by the Vahid-Engle decomposition, which removes the short-run effects of shocks and leaves the long-run trends associated with the time series. concentrated in the MSA as in the state as a whole. The relative concentration of the industry in the MSA is greater than that of the state if the location quotient is greater than 1 and the reverse if less than 1. Rolling regressions, in this article, make use of a statistical test (chi-square) to determine whether the cointegrating relationship of the full sample is the same as that of subsamples of the full time period. Starting with a subsample that begins at the start of the original sample, the Chisquare test is performed over and over again adding one more month of data after each test until all the data are included and tested. Sudden change is introduced by Krugman as the result of “a gradual change in the underlying (economic) conditions (that) can at times lead to explosive . . . change” (1991, 26). Endogenous metropolitan statistical areas, in the context of this article, are the cities whose labor markets are dependent on and react to demand and supply shocks from other metropolitan areas. Temporary cointegration is described by Siklos and Granger (1996) as a change in the long-run relationships between variables that could lead to the underlying series not being cointegrated at all times. Location quotients are used to determine the relative concentration of a particular industry in a region. If the location quotient is equal to 1 then the particular industry is as Weakly exogenous metropolitan statistical areas transmit internal supply and demand shocks to other less independent metropolitan areas. institutional features of an economy can interrupt an underlying equilibrium-type relationship possibly for an extended period of time” (1996, 8). Examining cointegrating relationships over different periods of time helps illuminate the evolution of those relationships. short-run behavior of an economic series. If one can demonstrate that a specific set of mathematical conditions is met, then it is possible to decompose data series such as employment in Florida MSAs into their trend (long-run) and cyclical (short-run) components. As the appendix shows, the prerequisites of the Vahid-Engle decomposition are met in data for the Florida MSAs, so the series can be decomposed into their long-term and short-term components. Chart 5 depicts the actual series and estimated employment trends (which incorporate other macroeconomic effects and are therefore not straight lines) for the six MSAs from 1970 to 1996. In Chart 6 the cyclical components of the trends are plotted by themselves. These lines correspond to the distance between the actual series and the estimated trend in Chart 5. These charts show that for each MSA there are several periods when the actual series is either above or below the estimated trend. These deviations from the Common Cycles he remaining discussion explores the short-run dynamics of the Florida MSAs’ labor markets. This analysis will reveal some of the similarities and differences in the reactions of the MSAs to shortrun economic shocks. The short-run behavior of the MSAs can be strikingly different. One MSA may be able to expand employment above its long-run trend while another may be left below its long-run trend. The concepts of common trends and common cycles, as introduced in Vahid and Engle (1993), extend the previous cointegration analysis. Their technique can in some cases be used to separate the long- and T 66 Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 CHART 5 Actual Series and Estimated Tr ends of Total Nonagricultural Payroll Employment for Six Florida MSAs, 1970–96 Employment (thousands) Miami Fort Lauderdale 640 960 900 560 840 Series 480 780 Series 720 400 660 320 Trend Trend 600 240 540 480 160 1970 1975 1980 1985 1990 1995 1970 1975 Employment (thousands) 1985 1990 1995 Jacksonville Orlando 700 450 500 350 Trend Trend Series Series 250 300 150 100 1970 1975 1980 1985 1990 1995 1970 1975 We s t P a l m B e a c h Employment (thousands) 1980 1980 1985 1990 1995 Tampa 1100 400 Series Series 900 300 Trend 700 Trend 200 500 100 300 1970 1975 1980 1985 1990 1995 1970 1975 1980 1985 1990 1995 Source: Series from the Bureau of Labor Statistics Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 67 CHART 6 Estimated Cycles of Total Nonagricultural Payroll Employment for Six Florida MSAs, 1970–96 Employment (thousands) Miami Fort Lauderdale 40 16 20 8 0 0 –20 –8 –40 1970 1975 1980 1985 1990 1995 1970 1975 1985 1990 1995 1990 1995 1990 1995 Jacksonville Orlando Employment (thousands) 1980 40 20 20 10 0 0 –20 –10 –40 –20 –60 1970 1975 1980 1985 1990 1995 1970 1975 Employment (thousands) We s t P a l m B e a c h 1985 Tampa 20 10 10 0 0 –10 –10 1970 68 1980 1975 1980 1985 1990 1995 Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 1970 1975 1980 1985 long-term trend are generated by short-run economic shocks to the growth of total payroll employment. Positive shocks such as a temporary increase in demand for a locally produced product (for example, a defense contract for a local firm) would lead local businesses temporarily to hire more workers than they otherwise would have. While they were employing more workers, the charts of the actual series and trend would show the actual number of employees exceeding the long-term trend. On the cyclical graphs this gap would correspond to an upswing above the horizontal axis. Shocks in one area may also spill over into others through demand for or supply of labor. Further, two or more areas may be subject to the same outside shocks or to shocks propagating across areas. Comparing cycles shown by this series of charts reveals both common and differential effects of shortterm shocks on the MSA’s employment. For example, Miami’s and Orlando’s deviations from their long-run trends appear in the Charts 5 and 6. Over the sample period the short-run behavior of these two MSAs is very different. In fact, they appear to be on opposite paths, with Miami hitting the height of its cycle in 1980 at a time when Orlando is near its lowest point. Looking at all of the MSAs, it can be seen that during most of the expansion of the 1980s, Miami, Fort Lauderdale, Tampa, and West Palm Beach are all above their long-run trend. However, Jacksonville’s level of total payroll employment, similar to Orlando’s, is below its trend. Viewing Miami and Orlando as the driving forces behind Florida’s economy could help explain the apparent division of the state into a countercyclical northern half and a procyclical southern one in terms of total payroll employment during this time period. Further examination of Chart 6 reveals that the MSAs can be grouped into three pairs of similar dynam- ics—Miami and Fort Lauderdale, Orlando and Jacksonville, and West Palm Beach and Tampa. Miami and Fort Lauderdale are the first to rise above their long-run trends in the 1980s’ expansion. They are followed by West Palm Beach and Tampa. Orlando and Jacksonville remained below their long-run trend during most of this period. It is interesting to note that West Palm Beach, although geographically closer to Miami, displays short-run dynamics more similar to Tampa’s in terms of the timing of its cyclical upswing. Conclusion ointegration techniques developed by Johansen (1995) and the common trends and common cycles analysis developed by Vahid and Engle (1993) have aided in studying the long- and short-run interrelationships in the behavior of total payroll employment in six Florida MSAs over the past quartercentury. The analysis showed that these MSAs have shared a long-run comovement in their labor markets. However, there are indications that these relationships have changed as the economic structures of the MSAs have evolved. Further, the cyclical dynamics displayed by these cities suggest that the labor markets of the northern half of the state behave differently from those in the southern half in response to short-run economic shocks. This analysis helps underline the growing diversity of influences on the growth trends of Florida MSAs. It also suggests that these MSAs react differently to shortrun shocks. Both of these dynamics are important in gauging the differing effects of policy or economic shocks on the state in parts and as a whole. C Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 69 A P P E N D I X Decomposing the Series into Trend and Cyclical Components Given r cointegrating vectors defined as the n ´ r matrix [a] and s cofeature vectors defined as the n ´ s matrix [b], stack the vectors in one matrix A: the condition of the Vahid-Engle decomposition that the sum of the two groups of vectors add up to the number of variables in the system. éb ù A = ê ú. ëa û Calculate A-inverse = [a- b-]. Partition A-inverse into the s ´ n matrix [b-] and r ´ n matrix [a-]. This calculation allows the decomposition into permanent (P) and cyclical (C) components such that Y(t) = P + C. It follows, then, that P = b-b Y(t) eliminates the cycles and leaves the trend or permanent component; C = a-a Y(t) eliminates the trend and leaves the cyclical or temporary component. Using the maximum eigenvalue test results presented in Table A, it was found that the time series has four cointegrating vectors. Next, to find the number of cofeature vectors, a test of canonical correlations between the series and certain other variables as explained in Vahid and Engle (1993) was used. This test (see Table B) shows that Florida’s MSAs share two cofeature vectors, satisfying Table A Test of the Number of Cointegrating Vectors Test Statistic Critical Value r Test Statistic 50.30 0 24.63 31.77 1 20.90 27.45 2 17.15 15.65 3 13.39 6.79 4 10.60 0.13 5 2.71 The test of the null hypothesis that the number of the cointegrating vectors is equal to r results in four cointegrating vectors. Table B Test of the Number of Cofeature Vectors Row Appox F Numerator DF Denominator DF Pr > F 1 2.7092 168 1644.062 0.0001 2 1.9561 135 1381.116 0.0001 3 1.7113 104 1113.348 0.0001 4 1.4652 75 840.884 0.0080 5 1.2902 48 564 0.0968 6 1.0946 23 283 0.3502 The F-test of the null hypothesis that the canonical correlations in the current row and all that follow are zero results in two cofeature vectors. The number of cofeature vectors is equal to the statistically zero canonical correlations (see Vahid and Engle 1993 for detailed explanations). The sum of the number of cointegrating vectors and cofeature vectors equals the number of variables in the system, and the Vahid-Engle decomposition can be used. 70 Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 REFERENCES ENGLE, ROBERT F., AND CLIVE W.J. GRANGER. 1987. “Co-integration and Error Correction: Representation, Estimation, and Testing.” Econometrica 55:251–76. SIKLOS, PIERRE L., AND CLIVE W.J. GRANGER. 1996. “Temporary Cointegration with an Application Interest Rate Parity.” University of California at San Diego, Discussion Paper 96-11. JOHANSEN, SOREN. 1995. Likelihood-Based Inference in Cointegrated Vector Autoregressive Models. Oxford: Oxford University Press. VAHID, FARSHID, AND ROBERT F. ENGLE. 1993. “Common Trends and Common Cycles.” Journal of Applied Econometrics 8:341–60. KRUGMAN, PAUL. 1991. Geography and Trade. Cambridge, Mass.: MIT Press. Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Third Quarter 1997 71