The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
3 M onetarism Is Dead; Long Live the Quantity T h eory 19 Prospects fo r International Policy Coordination: Som e Lessons fro m the EMS 30 International Linkages in the Term Structure o f Interest Rates 44 H ow Much Low er Can the U nem ploym ent Rate Go? 58 A Com parison o f Proposals to Restructure the U.S. Financial System TH E FEDERAL A RESERVE X liY N k o l ST. LOUIS 1 F ederal Reserve Bank of St. Louis Review July/August 1988 In This Issue . . . The first article in this Review is the text o f the 1988 Hom er Jones M em o rial lecture presented by W illiam G. Dewald o f the U.S. Department o f State. In 'Monetarism Is Dead; Long Live the Quantity Th eoiy,” Dewald distinguishes between simplistic monetarism, which he believes “was w idely interpreted as providing an alternative to short-run Keynesian m odel forecasts" and “ the Quantity Theoiy, whose focus is on the long run.” After an analysis o f long-run relationships and short-run forecasts, De wald concludes that w e simply don ’t know enough about the magnitude and timing o f m onetary effects to fine-tune the econom y effectively in the short run with m onetary policy actions; consequently, he believes that “monetarism as a short-run forecasting method should be buried.” He recom mends that Federal Reserve policy procedure should focus on the long-run relationship between m oney growth and nominal GNP growth, as em phasized by the Quantity T h eo iy o f Money, to constrain future in flation. Dewald suggests that having the Federal Reserve “target a noninflationary nominal GNP growth path for the five-year federal budget cy cle would be a step in the right direction.” * * * Although econom ic policymakers from the m ajor industrialized econo mies have called increasingly for greater coordination o f policies, little is known about how such coordination might work in practice. To investi gate the effects that econom ic policy coordination might have, Michael T. Belongia, in the second article in this issue, examines the perform ance o f the European M onetaiy System. This group o f countries has agreed, since 1979, to peg bilateral exchange rates within specified bounds, presumably to enhance overall econom ic performance. In “Prospects for International Policy Coordination: Some Lessons from the EMS,” Belongia compares the performance o f this group o f countries against a group that did not (ex plicitly) peg exchange rates. He finds no systematic difference in perfor mance, either between groups o f countries or before and after the EMS was formed. Thus, policy coordination appears to have had no important influence on econom ic performance, one way or another. * * * During the past 10 years, considerable controversy has developed over how interest rates are related across countries. One approach focuses on the relationship o f interest rates across the maturity spectrum, or term structure. In this view, movements in current short-term interest rates influence h ow interest rates w ill change in the future. In the third article in this Review, “International Linkages in the Term Structure o f Interest Rates,” Clemens J. M. Kool and John A. Tatom use this approach to study the relationship o f short- and long-term interest rates for five countries: the United States, Canada, the United Kingdom, West Germany and Japan. JULY/AUGUST 1988 2 Kool and Tatom explain the theoretical basis for expecting interest rates to be connected across countries. They show that long-term rates have been closely related among the five countries, w hile short-term rates have not. This finding raises a serious challenge to the term-structure view. The authors point out that the influence o f changes in foreign short term rates on both domestic short- and long-term rates could show up initially only in the latter rate. They find some evidence that changes in U.S. short-term rates have had a statistically significant influence on long term rates in Canada, Japan and West Germany, but that changes in onemonth interest rates elsewhere are not significant. The relationship from the United States to other countries is not robust either. Kool and Tatom conclude that the term structure is not a reliable mechanism connecting interest rates across countries. Instead, they suggest that changes in inter national long-term rates are related directly and quickly because o f rela tively com m on inflation rates and real interest rate developments. * * * In light o f recent declines in the nation’s unem ploym ent rate to below 5.5 percent, Keith M. Carlson, in the fourth article in this Review, examines the question, H ow much low er can the em ploym ent rate go without accel erating inflation? This critical rate, usually referred to as the “natural rate o f unem ploym ent,” was the subject o f major research in the late 1970s. At that time, estimates o f the natural rate ranged between 5 percent and 7 percent, but generally centered on 6 percent. Carlson reviews the develop ments that have affected the natural rate since W orld War II, focusing on the period since 1979. The author notes that the most obvious structural change in recent years has been the shifting age com position o f the labor force, which has reduced the natural rate about one-half percentage point below its 1979 level. The minimum wage, individual and em ployer tax rates and, possibly, unemployment benefits also have m oved favorably. As a result, the current natural rate appears to be w ell below the 6 percent m idpoint estimated in 1979. Restrictions on the activities o f banking organizations are being relaxed by federal and state authorities, and Congress is debating a more com plete restructuring o f the U.S. financial system. In the final article in this Review, "A Comparison o f Proposals to Restructure the U.S. Financial System,” R. Alton Gilbert describes six o f the m ajor proposals for restructuring the financial system and examines h ow they might affect the returns to share holders o f banking organizations as w ell as the potential losses to federal deposit insurance funds. Examining various methods o f combining a hypothetical bank and non banking firm in the same corporation leads the author to conclude that, under certain conditions, the expected loss to the insurance funds is smaller if banks offer nonbanking services directly, rather than through separate corporate entities. Gilbert also concludes that loans by bank sub sidiaries to their nonbank affilitates generally are not in the interest o f the bank holding company. FEDERAL RESERVE BANK OF ST. LOUIS 3 William G. Dewald William G. Dewald is deputy director, Planning and Economic Analysis Staff, at the Department of State. This paper, the second annual Homer Jones Memorial Lecture, was presented at St. Louis University on May 6, 1988. Views expressed are the author's own and do not represent the U.S. Department of State or the federal government. Monetarism Is Dead; Long Live the Quantity Theory J n OCTOBER 1979, when the Federal Open Market comm ittee adopted new operating procedures purported to be directed at control o f monetary aggregates, newspapers reported that economists at the Federal Reserve Bank o f St. Louis celebrated. Many had been hired and inspired bv Hom er Jones, its form er research director, to whose m em ory this lecture is dedicated. The celebration was premature. Those new procedures w ere a cover for genu inely restrictive policy actions that reversed the upward ratcheting o f inflation begun in the 1960s. It threatened to get out o f hand in 1979. Such a policy reversal was altogether appropriate, but, as in earlier episodes, it represented an abrupt shift in direction made necessary because earlier policy had taken the econom y so far off course. Whether or not the Federal Reserve genuinely attempted to control growth in the m onetary aggregates begin ning in 1979, it no longer does. The reason is not that it could not, but that the relationship between growth in the aggregates and GNP, and in turn inflation, appeared so unpredictable. Conse quently, in recent years the Federal Reserve has reverted to manipulating open market purchases and sales o f securities to hold federal funds rates or free reserves within target ranges as was the practice from the 1920s until 1979. In 1961, soon after leaving the Federal Reserve Bank o f Minneapolis, I gave a talk there in which I criticized Federal Reserve operating procedures for focusing on free reserves or interest rates rather than growth in the monetary aggregates. The Federal Reserve was characterized as a base ball player w ho can't hit a curve. He swings at where the ball was, not where it is. The example I cited was the experience in 1960 when the Federal Reserve persisted in targeting low er and low er interest rates even as monetary growth turned negative and the econom y slipped into recession. The Chairman o f the Board o f Governors o f the Federal Reserve in those days was William McChesnev Martin. He likened the role o f monetary policy to “leaning against the w ind,” the idea be ing that m oney market conditions as measured by interest rates or free reserves w ould tighten during business expansions and ease during contrac tions. In 1988, the Federal Reserve no longer tight ens, it snugs. Whatever the name, there is a prob lem with this approach. Even if the Federal Reserve takes no action, interest rates can change because o f changes in total spending in the econ om y and associated credit demands. The risk is that the Federal Reserve w ill attribute a decline in interest rates, as it did in 1960, to its policies when in fact, by not selecting a low enough interest rate JULY/AUGUST 1988 Chart 1 GNP Deflator Pre-1946 0.12 - 0 .0 8 Q uarterly Log G row th, Seasonally A djusted 1910 Chart 2 RealPre-1946 GNP Q uarterly Log G row th, Seasonally A djusted - 0 .0 8 - 0.12 1910 Chart 3 Nominal GNP n Pre-1946 0.16 0.08 - 0 .0 8 - 0 .1 6 FEDERAL RESERVE BANK OF ST. LOUIS Q uarterly Log G row th, S easonally A djusted 5 Post- 945 Q uarterly Log G row th, Seasonally Adjusted 0.12 0.08 0.04 /V 0 - 0 .0 4 - 0 .0 8 50 60 70 1980 - 0.12 JULY/AUGUST 1988 6 target, it sells open market securities and forces a contraction in m onetary aggregates. As a result, interest rates are prevented from falling as much as if no action w ere taken. There are problems associated with interest rate targets, but what about monetary targets? My pre sentation today addresses w hether the relation ship between m onetary growth and GNP has be come so unpredictable as to justify the abandonment o f monetary targets which seems to have occurred. MONETARISM AND THE QUANTITY THEORY Monetarism, the apparent heir of the Quantity Theory o f Money, was born in the 1960s. Not only was the name changed but also the concept. Un like the Quantity Theory, whose focus is on the long run, monetarism was w idely interpreted as providing an alternative to short run Keynesian m odel forecasts, a view not always shared by its progenitors. The Federal Reserve Bank o f St. Louis equation, which explained quarterly GNP growth largely as a function o f monetary growth, became a major monetarist forecasting tool.' Its simplicity and apparent reliability captured the one-dimensional attention o f Wall Street and Washington. GNP growth was estimated to reflect growth in nar rowly defined money, M l, in the current quarter and the previous year; and it was found to rise about 3 percent a year independently o f m onetaiy growth. With hindsight, w e know that this stable M l velocity trend was peculiar to the period on which the estimates w ere based, initially the 1950s and 1960s but then the unfolding 1970s as well. The Federal Reserve Bank o f St. Louis m odel went beyond the estimated GNP or demand growth equation to incorporate potential supply growth which together determ ined inflation and unemployment, and a credit market which deter m ined interest rates.2By the end o f the 1970s and into the 1980s, the weekly publication o f M l changes became a major news event and market 'Leonall C. Andersen and Jerry L. Jordan, “ Monetary and Fiscal Actions: A Test of Their Relative Importance in Economic Stabi lization,” this Review (November, 1968), pp. 11-24. 2Leonall C. Andersen and Keith M. Carlson, "A Monetarist Model for Economic Stabilization,” this Review (April, 1970), pp. 7-25. 3Meltzer, Allan H. “ On Monetary Stability and Monetary Reform” in Y. Suzuki and M. Okabe, eds., Toward a World of Economic Stability (Tokyo: University of Tokyo Press, 1988), pp. 51-74. http://fraser.stlouisfed.org/ FEDERAL RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis force because these data provided a basis for fore casts o f total dem and growth, inflation and inter est rates. The problem with the simplistic monetarism that afflicted Wall Street and Washington was that it accepted Milton Friedman's dictum that in flation is always and everywhere a monetary p h e nom enon but not his stipulation that lags are long and variable. M y point today builds on this theme. Monetary policy actions are appropriately directed at longrun stability o f the general level o f prices but not at offsetting undesired short-term movements in total demand, unemployment, or, for that matter, prices. I shall argue that w e know enough to keep inflation trends within bounds but not enough to fully stabilize the price level let alone the business cycle. A corollary is that monetarism as a shortrun forecasting m ethod should be buried; but the Quantity Theory, defined as the predictability of GNP growth on the basis o f growth in the m one tary aggregates, should be recognized as the cor rect principle for controlling inflation in the long run; and Federal Reserve operating procedures should be made consistent with that principle. SHORT-TERM FORECASTS Let me make a few remarks about short-term forecasts. None are very good for very long. Based on Federal Reserve “green” books, Allan M eltzer reports that the Federal Reserve's record o f fore casting nominal GNP growth a year ahead over the period 1967 through 1982 had a root mean square error equal to about 60 percent o f average nominal GNP growth.3 Since GNP growth averaged about 10 percent a year, the forecast error is 6 percentage points, indicating that one-third o f the time fore casts w ould miss bv more than 6 percentage points and half the time by more than 4 percent age points. Furthermore, and most important, the Federal Reserve forecasts w ere way off track, miss ing average growth by more than 5 percentage points, the result o f the Federal Reserve persist ently underestimating GNP growth during a pe 7 Chart 4 Federal Debt and Deficit Ratios to GNP riod o f a rising inflationary trend. These striking results are confirm ed in an analysis o f Federal Reserve forecasting that Karamouzis and Lombra presented at the Carnegie-Rochester Conference last month.4They found that the Federal Reserve forecasts systematically underpredicted GNP growth during expansions and overpredicted dur ing contractions. According to Meltzer, private forecasters have had a somewhat better record than the Federal Reserve, but one still is talking about errors o f 4 percentage points a third o f the time and nearly 3 percentage points, half the time. Since inflation is such a lagging factor, changes in nominal GNP growth are initially translated into real growth changes. Hence, errors o f 3 percentage points or more in real GNP growth half the time translate into being unable to distinguish reliably between a boom and a recession in either the current quarter or a year ahead. M eltzer was mainly summarizing the perfor mance o f non-monetarist forecasts, but one can make at least as critical remarks about monetarist short-term forecasts in the 1980s. Like many an other forecaster, M ilton Friedm an’s record is blemished. For example, he forecast a recession that didn’t materialize in 1984 and an equally illu sory inflation in 1986. In 1988, not only Friedman but others o f comparable persuasion are w orried about the consequences o f the contraction in monetary growth in 1987. I too am concerned, though it is worth m ention ing, as Jim Meigs, an early colleague o f Hom er Jones at the Federal Reserve Bank o f St. Louis, has “Karamouzis, Nicholas and Raymond Lombra “ Federal Reserve Policy Making: An Overview and Analysis of the Policy Pro cess," Carnegie-Rochester Conference Series on Public Policy, forthcoming. JULY/AUGUST 1988 8 Chart 5 Money: M1 and M1a Pre-1946 Quarterly Log Growth, Seasonally Adjusted Chart 6 GNP/Money: Various Measures Pre-1946 Quarterly Log Growth, Seasonally Adjusted rem inded me, that Hom er was suspicious about all short-term forecasts, including those based on monetary growth. It was his persistent question ing that created the flurry o f econom etric work about monetary relationships for w hich the Fed eral Reserve Bank of St. Louis becam e famous. HISTORICAL RELATIONSHIPS: THE BROAD PICTURE The historical relationship between monetary growth and spending confirms Jones’ suspicion. Let me present some charts which put the experi ence o f the 1980s in perspective. Chart 1 records inflation in the United States since 1907, with 1907-45 and 1946-87 plotted sep http://fraser.stlouisfed.org/ FEDERAL RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis arately. The blue-shaded areas identify reces sions. Quite clearly inflation was a lot m ore vari able in the initial period, though, because o f deflations during recessions in the earlier period, there was no sustained inflation trend as there was in the second period. Chart 2 plots the real GNP growth rate — a mea sure o f growth in the real supply o f goods and services. Though it averaged about 3 percent a year both before and after the end o f 1945, the magni tude o f the booms and busts was much greater in the earlier period. Since real growth averaged about as much in each period, it follows that the inflation uptrend in the second period was an aggregate demand not an aggregate supply phe nomenon. Chart 3 presents the nominal GNP growth rate 9 Quarterly Log Growth, Seasonally Adjusted Post-1945 0.20 0.15 0.10 0.05 vA,^v-v Wj 0 M1a M1 (Fed. Res.) - 0 .0 5 - 50 60 — a measure o f growth in nominal dem and for goods and services. Though most values are posi tive, there are some big negatives in recessions through 1960. Since then there has been slowed, not negative, GNP growth during recessions be cause w e have had considerable inflation even in recessions. In terms of proximate causes, Chart 3 shows that slowed GNP growth has always been associated with slowed real growth in recessions, and accelerated GNP growth w ith accelerated real growth in expansions. Thus, decreased variability in real growth in the post-World W ar II period is linked to less variability in nominal GNP growth. What about sources o f nominal GNP growth? Conventional wisdom to the contrary, the timing o f government spending and tax changes is not 70 1980 0.10 - 0 .1 5 systematically correlated with GNP growth. The 1980s provide a good example. Fiscal policy by every measure was expansionary, yet nominal GNP growth contracted. Chart 4 plots the ratio o f nominal federal debt held by the public to nominal GNP. There is a nominal deficit if the debt rises, but a real deficit only if the debt rises faster than inflation. An in crease in the debt to GNP ratio reflects the real deficit rising faster than real growth. The historical record shows that real deficits relative to real GNP did not amount to much before W orld War I. Big real deficits occurred in both W orld Wars, the early 1930s, and since 1980. Since nominal GNP growth accelerated in the wars but decelerated in the 1930s and 1980s, there is no consistent rela- JULY/AUGUST 1988 10 Chart 7 M2 Chart 8 GNP/Money: Various Measures 0.20 Pre-1946 Q uarterly Log G rowth, Seasonally Adjusted tionship. Furthermore, in the 40 years since the end o f W orld War II, there have been nine busi ness cycle expansions and in only one — the cur rent one — did real deficits rise significantly rela tive to GNP. Perhaps these official debt figures are the w rong ones to look at because they do not incorporate discounted values o f future entitle ments and tax receipts. Others might find what they are looking for in these data, but I conclude 5M1 includes currency and demand deposits; M1A omits de posits that pay interest; M2 adds small time and savings de posits, overnight repurchase agreements and Eurodollar de posits, and, since 1959, shares in thrifts and money market mutual funds. http://fraser.stlouisfed.org/ FEDERAL RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis that there is no consistent relationship between fiscal deficits and GNP growth. Charts 5 and 6 present the growth rates o f the monetary aggregates: M l, M IA and M2.5The rec ord shows that m ajor increases in GNP growth in W orld Wars I and I I w ere accom panied by both accelerated monetary growth and rising fiscal deficits, and postwar contractions in GNP growth by the reverse movements. Nonetheless, there are 11 many instances w hen fiscal and m onetary actions pushed in opposite directions. This divergence permits identification o f which is the dominating factor affecting GNP growth. In the early 1930s, real federal deficits ballooned but monetary growth collapsed. So did GNP growth. In the 1960s and 1970s, real deficits grew less than real GNP if at all. Monetary growth increased and so did GNP growth and inflation. In particular episodes, such as 1966-67 w hen real deficits went one w ay and total spending growth the other, it was monetary growth that tipped the balance. Charts 7 and 8 present the growth rates in M l and M2 velocities. By definition, velocity growth is GNP growth in excess o f m onetary growth. The charts reveal h ow steadily M l velocity increased in the 1950s through the 1970s, and how ragged its changes in eveiy other period. The charts also show h ow M2 velocity remained largely trendless in comparison with M l velocity which dipped in the 1930s but then rose persistently after 1945 until the 1980s. Note well that in eveiy recession both M l and M2 velocities fell so that to cushion GNP growth w ould require faster monetary growth. In the worst recessions, including 1981— 82, m onetary growth did not accelerate as velocity growth slowed; and in the worst inflations, includ ing the late 1970s, monetary growth did not decel erate as velocity speeded up. Hence, monetary growth has often been an ineffective counterbal ance to moderate excesses in GNP growth. JULY/AUGUST 1988 12 Table 1 Average Forecast Errors and Changes in Economic Trends (weighted least squares) M1 (1924:4-1987:1) R2 Constant Inflation Real Growth Interest Rate Degrees of Freedom 0.30 -1 .0 0.4 -0 .1 0.1 ( - 0 .5 ) (1.3) ( - 0 .3 ) (0.3) 9 0.29 -1 .0 0.5 -0 .2 ( - 0 .5 ) (1.9) ( - 0 .7 ) — 10 0.26 -1 .0 0.5 ( - 0 .5 ) (2.0) — — 11 M2 (1919:2-1987:1) R2 Constant Inflation Real Growth Interest Rate Degrees of Freedom 0.43 -0 .1 0.4 0.1 -0 .1 ( - 0 .1 ) (2.5) (0.7) (- 0 .4 ) 11 0.42 -0 .1 0.4 0.2 ( - 0 .1 ) (2.8) (1.2) — 12 0.35 -0 .1 0.4 ( - 0 .1 ) (2.6) — — 13 T-statistics in parentheses. Independent variables are changes from the last business cycle average in the estimation period to the average for the forecast period. HISTORICAL RELATIONSHIPS: SPECIFIC FORECASTS William Gavin and I have been studying the quality o f quarterly GNP forecasts based on the monetary aggregates.6Though there are many studies that have examined the post-World War I I period, w e were interested in a broader historical experience. Our focus was on out-of-sample fore casts — the kind needed to direct monetary aggre gate changes to achieve a desired GNP growth path. Quarterly GNP growth forecasts for each business cycle w ere based on estimates o f the relationship between GNP growth and four quar terly lags o f monetary growth for the three preced ing cycles, that is, a m odified St. Louis equation. On the average, both M l and M2 changes were estimated to change GNP growth roughly propor tionally w hile velocity trends w ere significant in relating M l but not M2 to GNP. Overall there were 15 forecast intervals for M2 but only 13 for M l because there was no quarterly information about the split between demand and time deposits be fore 1914. The first forecast for M l was the busi ness cycle 1924:4 - 1927:4. 6Gavin, William T., and William G. Dewald, “ Velocity Uncertainty: An Historical Perspective,” U.S. Department of State, Bureau of Economic and Business Affairs, Planning and Economic Analy sis Staff Working Paper 87/4, November 1987. Gavin was an economist at the State Department in 1987 on leave from the Federal Reserve Bank of Cleveland. FEDERAL RESERVE BANK OF ST. LOUIS There are many factors that influence GNP growth. Consequently, in our single equation models that relate GNP growth solely to m onetary growth, w e expected that shifts in the econom y including m onetary policy reactions to econom ic performance w ould lead to biases in the forecasts. For example, w e expected that low er interest rates in a forecast period w ould decrease velocity and reduce GNP growth relative to monetary growth. To measure the effect o f such shifts, w e regressed average forecast errors on changes in inflation, interest rates and real growth from the last busi ness cycle in the estimation interval to the average observed in the forecast cycle. As noted, there was a large decrease in the vari ance o f forecasts from the pre-1946 to the post1945 period. To account for such heteroscedasticity, w e weighted observations by the expected standard deviation o f the mean forecast errors and then used ordinary least squares to estimate ef fects o f shifts in inflation, interest rates and real growth trends on forecast errors. Table 1 presents the results. The only consistent link to forecast 13 errors was change in the inflation trend, not inter est rates, and not real growth. Gavin and I also examined cross-countiy evi dence. The results appear in table 2. W e estimated the relationship between annual GNP growth and current and lagged M l growth for 39 countries for the late 1950s through 1979. GNP growth forecasts for each countiy were made for 1980-84. As in our U.S. time series analysis, these cross-country GNP forecast errors w ere strongly correlated with changes in inflation trends, even excluding out liers such as Bolivia, Brazil, Mexico and Peru that had huge inflation accelerations in the 1980s. Why the consistent link to shifts in inflation trends? Look at chart 9. It is apparent that w ide swings in interest rates over the business cycle w ere not closely related to M l velocity movements. Furthermore, since real growth averaged about the same before as after the end o f 1945, one cannot attribute the persistent rise in M l velocity until 1982 to that source. Rather, the rise in M l velocity after 1945 was associated with a persistent rise in the inflation trend. Table 2 Average Forecast Errors and Changes in Inflation Trends (M1 models only for 39 countries) _________ All countries R2 Constant Inflation Degrees of Freedom 0.7 0.8 0.3 (9.0) 37 Excluding outliers 0.2 0.6 0.3 (2.4) 33 Countries: Australia, Austria, Belgium, Bolivia, Brazil, Canada, Colombia, Denmark, Dominican Republic, Ecuador, El Salvador, Finland, France, Greece, Guatemala, Honduras, Iceland, Ireland, Italy, Japan, Mexico, Netherlands, New Zealand, Norway, Paraguay, Peru, Philippines, Portugal, South Africa, Spain, Sri Lanka, Sweden, Switzerland, Thailand, Turkey, United Kingdom, United States, Venezuela, West Germany. Sample periods vary because of data availability but are approximately 1957-84. Chart 9 Interest Rates and GNP/M1 JULY/AUGUST 1988 14 Chart 10 Interest Rates and GNP/M2 Velocity 4-6 month commercial paper Chart 10 reveals a much weaker association between M2 velocity and interest rates and much less o f a trend. The shift in the series is attribut able to a redefinition o f M2 in 1959 to include a variety o f non-bank liabilities that were not in the Friedman and Schwartz definition. Chart 11, w hich plots only recent data, reveals a close relationship between M2 velocity and the Treasury bill rate less a calculated w eighted aver age own-rate on M2.7Depository institutions re spond to persistent changes in market rates by altering deposit rates, but, even when uncon strained by deposit interest ceilings, adjustments are not that quick or complete. Since the post-war ratcheting up o f interest rates reflected an uptrend in inflation, it follows that lags in setting deposit 7Moore, George, Richard Porter, and Dave Small. “ Modeling the Disaggregated Demands tor M2 and M1 in the 1980s: The U.S. Experience,” Federal Reserve Board Conference on Monetary Aggregates and Financial Sector Behavior in Interdependent Economies (forthcoming). http://fraser.stlouisfed.org/ FEDERAL RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis Interest rates interest rates led to rising opportunity costs of holding M2 balances and to increased M2 velocity when inflation trended up strongly as in 1978—80. In the opposite circumstances when inflation trended dow n strongly as in 1982-87, falling op portunity costs o f holding M2 balances decreased M2 velocity. Something similar was going on in earlier years too. Thus, Gavin and I found that shifts in inflation trends, but not interest rates, w ere consistently tied to errors in GNP growth forecasts based on not only M l but also M2 growth. Table 3 presents the average GNP growth fore cast errors for full cycles based on M l and M2 growth. Neither totally dominates the other though M2 m odels were best on average and in 15 Chart 11 M2 Velocity vs M2 Opportunity Costal 11 Standardized Values the current cycle. Our finding that forecast errors are smaller for M2 than for M l or M IA but not by a large margin suggests robustness to the choice of the monetary aggregate.8W e also looked at the monetary base and found that M2 models pro vided the best forecasts on average for both 1907— 45 and 1946-87. Turning again to table 3, some forecast errors are huge. Root mean square errors average 17 to 18 percent in the pre-1946 period, though only about 6 percent in the post-World War I I period. Bv the standards that M eltzer discussed, such errors are comparable to Federal Reserve forecast errors in the “green" book. An inference is that attempts to fine tune GNP growth by controlling either M l or M2 growth w ould miss GNP growth targets by more than 6 percentage points one-third o f the time and by 4 percentage points half the time. What was said about not being able to distinguish boom from recession holds for our forecasts just as for the Federal Reserve's. However, there is a difference. The average forecast error associated with our simple relationship o f monetary growth to nominal GNP growth appears to be w ell under the reported average errors in Federal Reserve “green” books that M eltzer reported. OPERATIONAL ISSUES The operational question is what to do in the short run to achieve a long-term inflation objec tive. Suffice it to say that the Federal Reserve need not iron out every wrinkle in monetary growth to eliminate inflation trends, but it is necessary to tie 8For a different opinion, see Michael R. Darby, Angelo R. Mas cara, and Michael L. Marlow. ‘‘The Empirical Reliability of Mon etary Aggregates as Indicators,” Research Paper No. 87 (U.S. Department of the Treasury, 1987). JULY/AUGUST 1988 16 Table 3 GNP Growth Forecast Errors Annual Rates of Percent Change Business cycles 1924:4-1927:4 1928:1-1933:1 1933:2-1938:2 1938:3-1945:4 Average Pre-1946 Root mean square errors Average forecast errors M1 M1A 1.8 -5 .9 - 1 .8 - 6 .6 -3 .1 M2 BEST M1 0.1 1.0 5.3 -0 .1 M2 M2 M1 M2 1.5 M1A M2 BEST 10.3 15.1 28.1 17.5 9.6 13.6 26.3 18.5 M2 M2 M2 M1 M2 17.7 17.0 M2 4.7 - 1 .3 - 2 .9 - 3 .0 0.6 -0 .1 1.6 - 1 .7 - 2 .3 M2 M1 M1 M1 M2 M2 M1 M2 M2 11.1 7.0 4.4 5.4 4.4 3.7 3.7 7.6 8.6 8.8 7.8 6.1 5.9 3.0 4.1 4.5 6.5 4.3 M2 M1 M1 M1 M2 M1 M1 M2 M2 1946:1-1949:4 1950:1-1954:2 1954:3-1958:2 1958:3-1961:1 1961:2-1970:4 1971:1-1975:1 1975:2-1980:3 1980:4-1982:4 1983:1-1987:1 8.8 0.2 - 0 .6 0.3 - 2 .0 0.3 1.2 -2 .6 - 6 .6 Average Post-1945 -0 .1 - 0 .5 M1 6.2 5.7 M2 Overall Average - 1 .0 - 0 .3 M2 9.8 9.1 M2 1.9 4.8 -3 .2 4.0 11.1 4.9 monetary growth to real growth over the medium term to avoid the kind o f disturbances that shifts in inflation trends engender. The Federal Reserve needs to adopt systematic operational procedures to shift its policy targets on the basis o f observed deviations o f GNP growth from desired levels. growth even as it supplied funds to support accel erating monetary growth w hich was reflected in accelerating inflation, higher interest rates, an increasing velocity trend and unexpectedly large GNP growth. Could that sad cycle have been avoided? One way w ould have the Federal Reserve set a GNP growth target equal to long-term real growth plus an inflation target, perhaps zero in the long run but not unreasonably only a partial step in that direction in any one period. The point is not to set m onetary targets on the basis o f short-run forecasts o f what real and nominal GNP growth is predicted — I hope I have made clear how errorprone such forecasts are — but rather on the basis o f long-run real growth projections plus an in flation goal, not a current GNP forecast.9 Suppose in 1978, to pick a year, the Federal Re serve had aimed at 3 percent real growth — the long-term average — and an inflation target 2 per centage points below the 6.8 percent inflation in 1977. Target GNP growth for 1978 w ould have been 7.8 percent; for 1979, 5.8 percent; 1980, 3.8 percent; 1981 and thereafter, 3 percent — the long-term average real growth rate. Such a procedure in the 1970s w ould have led to very different results from what w e got. The Federal Reserve persistently underforecast GNP 9A somewhat similar proposal is found in Bennett McCallum. “ Robustness Properties of a Rule for Monetary Policy,” Carnegie-Rochester Conference Series on Public Policy, Vol. 29, forthcoming. FEDERAL RESERVE BANK OF ST. LOUIS Fourth-quarter-over-fourth-quarter GNP growth in 1978 was 14.8 percent, not 7.8 percent. GNP growth stayed high: 9.5 percent in 1979 and again in 1980. Inflation accelerated: 7.7 percent in 1978, 8.5 percent in 1979 and 9.4 percent in 1980. Part of the problem was rising velocity, but the problem 17 was com pounded because the Federal Reserve validated the inflation process by an open market policy that permitted monetary aggregate growth o f no less than 7 percent in any o f those years and by as much as 11 percent. It was not distinguish ing between the w ind it was leaning against and the thrust o f its ow n actions. One cannot be certain about velocity m ove ments in the short run but in the circumstances of the late 1970s with a rising inflation trend, one could have anticipated rising velocities. By what ever means the Federal Reserve might have chosen to control its open market operations — targeting free reserves, federal funds rates, or monetary base injections — over the course o f those years it w ould have had to take actions to restrict m one tary growth to prevent inflation from accelerating. What was required in 1978, if not sooner, was a genuinely restrictive policy such as w e finally got in 1980-81. That policy arrived too late to avoid enormous econom ic destruction. Inflationary expectations had becom e entrenched in market contracts denom inated in dollars. The costs of disinflation: the worst recession since the 1930s, an overhanging burden o f domestic and interna tional debt accumulated on the basis o f mistaken price expectations, and a legacy o f uncertainty about whether it might not happen again. WHY NOT TARGET NOMINAL GNP GROWTH? It is my contention that putting a GNP target up front for the Federal Open Market Committee to aim at w ould allow it to m obilize its staff to design the best w ay to keep monetary growth and GNP growth dow n w hen such a course is obviously right as it was in the late 1970s. There is doubt lessly an element o f discretionary fine-tuning in GNP targeting, but with a twist. Deviations from the target nominal GNP path should induce Fed eral Reserve actions to move monetary growth up or down in order to bring forecast GNP growth back to a long-run non-inflationary path. Perhaps, there should be some limit on how much change in targeted GNP to be permitted in a particular period. In any case, to avoid getting off track as in the 1970s, the Federal Reserve has to direct its considerable powers toward controlling inflation trends by actions that push m onetary growth in the right direction w hen nominal GNP growth is off target. CONCLUSION To eliminate inflation trends, m onetary growth must be kept lo w on average and close to real growth trends. Extraordinary increases as in 197779 or 1985-86 ought to be avoided so that offset ting decreases are not necessitated; but the past is history. What about the future? Certainly w e want to avoid another cycle o f inflation and disinflation. By luck or design the Federal Reserve in 1987 and early 1988 has pursued policies that are not so different from what I have suggested. Monetary aggregates are grow ing at about 4 percent annual rates, close to appropriate rates to bring inflation down gradually toward zero. I w ould hope that the lessons o f history could be applied to stay on such a path. A positive reform to make clear the responsibili ties o f the Federal Reserve regarding long-term inflation w ou ld be to bring it into the federal budget process. Have it announce nominal GNP targets each year on w hich to base Administration budget projections over the ensuing five fiscal years. Both GNP growth and inflation are critical to the budget w ith respect to tax receipts and ex penditures, particularly interest outlays. W hy have the Administration make arbitrary assumptions about GNP growth and inflation as it does now when the Federal Reserve, whose powers are so important in determining nominal magnitudes, could target such values and be held accountable for attaining them? It should take responsibility for what it can control in the medium term — nominal spending growth and inflation — and not play meteorologist by leaning against the uncer tain winds o f the business cycle. Can w e devise ways to create the right incen tives for Federal Reserve officials to pursue poli cies to keep inflation low? The Germans and the Japanese have. In contrast to their success in keeping inflation low, w e have gone through the motions o f having the Federal Reserve announce monetary target ranges to Congressional Oversight Committees beginning in 1975, and since then the worst cycle o f inflation and disinflation since W orld War II. Setting medium-term targets for GNP growth as I have recom m ended w ou ld establish a new responsibility. However, unless the monetary authorities shoulder that responsibility by taking actions to stabilize nominal GNP growth around a medium-term non-inflationary path, nothing w ould be gained. Establishing yet another target range w ould make sense only if deviations from it induced stabilizing policy reactions. JULY/AUGUST 1988 18 Perhaps, the Federal Reserve must be put on a shorter leash? We could specify a legal limit to the monetary base that the Federal Reserve was au thorized to put into circulation in a fiscal year. Budget authority is required for the Treasury to spend, w hy not for the Federal Reserve? Then again, it might be somewhat unrealistic to count on Congress to check the inflationary tendencies o f the Federal Reserve. An even shorter leash has been suggested bv Milton Friedman (and not in jest). He w ould disband the Federal Open Market Committee and hire a federal em ployee to pur chase Treasury securities each week as specified by law to keep some monetary aggregate on a long-term zero inflation course. Despite the budget savings in his proposal, w ide variation in velocities historically suggests that w e might do better than fixing a monetary growth rate in per petuity. The fact is that broadly stabilizing monetary policies have been observed on occasion in his tory. Even during the past decade, some countries have managed their affairs to avoid the worst ex cesses o f inflation and disinflation that w e and many others experienced. W e can’t repeat history, but w e ought to learn from it. In the light o f the contribution o f Federal Reserve actions to instabil ity in m onetary growth, nominal GNP growth and inflation, having it target a non-inflationary nom i nal GNP growth path over a five-year federal budget cycle would be a step in the right direc tion. Responsibility for control o f inflation w ould be assigned to the institution that has the most direct pow er to influence nominal GNP growth and, in turn, inflation. For nominal GNP targeting to succeed in eliminating inflation trends, how ever, Federal Reserve officials must have the un derstanding and courage to support the necessary FEDERAL RESERVE BANK OF ST. LOUIS policy actions to get back to a non-inflationary GNP growth path whenever the target is missed. If they did im plement such a policy, they w ould not likely eliminate all the ups and downs in the econ omy, but they w ould avoid repeating the most egregious mistakes o f m onetary history. DATA SOURCES Data used in preparing the charts and statistical study sum marized in this lecture came from a variety of sources. M1 and M2 for May 1907 to December 1958 from Milton Fried man and Anna Jacobson Schwartz, A Monetary History of the United States: 1867-1960, (Princeton University Press, 1963); and January 1959 to March 1987 from the Board of Gover nors of the Federal Reserve System. Values of M1 were semi-annual until June 1914 and were used in constructing the charts. Monetary base for May 1907 to December 1918 from Friedman and Schwartz; and January 1919 to March 1987 from the Federal Reserve Bank of St. Louis, adjusted for required resen/e ratio changes but not seasonality. The Census X-11 program in SAS was used to seasonally adjust these monthly data from which quarterly averages were calculated. Commercial paper rate for May 1907 to December 1970 from Board of Governors of the Federal Reserve System, Banking and Monetary Statistics, 1976; and January 1971 to March 1987 from the Federal Reserve Bulletin. Quarterly averages were calculated from the monthly series. GNP and GNP deflator for 1907:Q2 to 1947:Q4 from Robert J. Gordon, “ Price Inertia and Policy Ineffectiveness in the United States, 1890-1980,” Journal of Political Economy (December 1982), 1087-1117; and 1948:Q1 to 1987:Q1 from the Depart ment of Commerce, Bureau of Economic Analysis. All computation were performed on an IBM AT using RATS PC version 2.0 or LOTUS version 2.01. Data from different sources were spliced by transforming the early series to growth rates and computing revised level series based on actual levels of the most recent series. The original data used in the Gavin and Dewald study are available from the author on a LOTUS spreadsheet upon re quest with an accompanying 51/4 inch diskette and a stamped, self-addressed disk mailer. 19 Michael T. Belongia Michael T. Belongia is a research officer at the Federal Reserve Bank of St. Louis. Anne M. Grubish provided research assistance. Prospects for International Policy Coordination: Some Lessons from the EMS "Altogether, then, econom ic co-operation is no match for m otherhood.” Clive Crook, The E conom ist H E strong rise in the value o f the dollar in the early 1980s and its sharp decline since February 1985 are alleged to have had wide-ranging effects on the econom ies o f the United States and its ma jor trading partners. In response to concerns about the costs o f adjusting to large exchange rate movements specifically and the effects o f diver gent econom ic policies generally, policymakers have called for greater coordination o f econom ic policies among the w o rld ’s m ajor industrial coun tries.1But, despite the stated official desire for greater policy coordination, little is certain about how it might work in practice. Some theoretical 'At the September 1985 Plaza Accord, for example, the G-5 countries agreed to coordinated intervention policies that would reduce the value of the dollar. Since that meeting, there have been subsequent economic “ summits” to discuss both target values for exchange rates (the Louvre Accord of February 1987) and indicators by which policies could be monitored (the June 1987 Venice Summit). Both the Bank for International Settlements (BIS) and the OECD have called for greater fiscal policy cooperation, with lower budget deficits in the United States and expansionary policies in Japan and Germany. See Bank for International Settlements (1987) and Organization for Economic Cooperation and Development (1987). 2Models using game theory have tended to conclude that policy cooperation will produce lower social welfare losses than non results suggest that there are potential gains from coordinated policy actions; these results, however, are not robust.2 One example o f an explicit agreement for policy coordination is the European Monetary System (EMS). Established in 1979, the EMS was form ed to stabilize bilateral nominal exchange rates among m em ber countries. Because it is difficult to iden tify the direct benefits o f more stable exchange rates per se, analysts typically have discussed the potential benefits o f such coordination in terms of increased trade flows, faster real growth and pol icy convergence among m em ber nations. cooperative policies. Some empirical work has provided evi dence that supports the game theory results; see Currie and Levine, for example. It should be noted, however, that both lines of work are based on arbitrary social welfare functions and the existence of a benevolent policymaker. The public choice litera ture, in contrast, suggests that the wealth of the policymaker dominates social objectives as a criterion for choosing particular policy paths. If true, a quite different loss function would apply to policy choices. More generally, the game-theoretic results depend heavily on the loss function specified. Fischer (1987) and Frankel and Rockett (1987) also have shown that the results depend importantly upon the economic models used to evaluate policy effects. JULY/AUGUST 1988 20 As the one case in which some form o f explicit cooperation has been adopted, the EMS offers an opportunity (and data) to examine its effect on a variety o f econom ic indicators. This article reviews the econom ic experience o f EMS countries relative to non-EMS countries during the 1980s to see whether exchange rate coordination has been associated with differential gains in other m ea sures o f econom ic well-being as well as to draw inferences about the likely effects o f policy coordi nation on a greater scale by the industrial econo mies. THE EMS: AN OVERVIEW The EMS, w hich was established formally on March 13,1979, was first com posed o f the nine European Community (EC) countries: Belgium, Denmark, France, West Germany, Ireland, Italy, Luxembourg, The Netherlands and the United Kingdom. Greece, which subsequently joined the EC, became an EMS partner in 1985 but Spain and Portugal, w hich joined the EC in 1984, have not yet becom e members o f the EMS. Briefly, EMS m em bership requires each nation first to deposit 20 percent o f its gold and gross dollar assets with the European Monetary Cooperation Fund (EMCF). In exchange, each nation receives an equivalent amount o f European Currency Units (ECUs), which serve primarily as a unit o f account for EMS functions (see Appendix). This asset exchange, however, is not so much a separate part o f joining the EMS as it is a preliminary step to pursuing the System s objectives.3The second part o f EMS membership involves the agreement to pursue stable nominal exchange rates, at agreed levels, for each bilateral set o f rates. One rationale for this policy objective is that exchange rate variability is a source o f uncertainty that reduces trade and the traded goods sector is a large portion o f each EMS member econom y.4 Although exchange rate objectives are “set,” the EMS is not strictly a fixed-rate system; adjust 3A detailed summary of the ECU, as well as the evolution of the EMS, is in Lingerer, et al. (1986). Karamouzis (1987) presents a shorter overview of the system and policy coordination. “Both the theoretical and empirical evidence on a link between exchange rate variability and trade are ambiguous. DeGrauwe (1987,1988), for example, provides evidence suggestive of a negative effect. Many others, surveyed in Farrell, et al. (1983), find no significant relationship between measures of exchange rate variability and trade. And, moving in the opposite direction, Franke (1987) provides theoretical reasoning for a positive relationship between exchange rate variability and trade. On balance, however, the predominant result seems to be that there is no important relationship between the two variables. http://fraser.stlouisfed.org/ FEDERAL RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis ments to the exchange rate levels have been made from time to time.5For example, in a major ex change rate realignment in March 1983, the French franc, Italian lira and Irish pound w ere devalued between 2.5 percent and 3.5 percent, while the remaining currencies w ere revalued between 2.5 percent (Danish krone) and 5.5 per cent (German mark). As the IMF explains: Like previous realignments, this realignm ent had becom e necessaiy as a result o f continued differ ences in the underlying strength o f the participat ing countries’ external positions, w hich reflected in turn divergences in econ om ic policies and costprice perform ance. These differences had gen er ated expectations o f exchange rate changes and led to large speculative capital flows." Similarly, in 1985, the lira was devalued 6 percent and other currencies revalued 2 percent when [t]he worsening o f the current account reflected prim arily the m aintenance o f a rate o f grow th in dom estic dem and higher than that o f Italy’s part ners as w ell as the lagged effects o f a significant loss o f com petitiveness vis-a’-vis other EMS coun tries over the previous tw o years .7 Thus, when fundamental differences in econom ic performance require changes in the established exchange rate targets, the EMS has revalued them. Table 1 shows the dates o f these revaluations and their effect on individual currencies. Between revaluations, bilateral rates are allowed to vaiy within margins o f 2.25 percent o f the d e sired values; because Italy historically has had higher rates o f inflation than the other EMS coun tries, the lira has a band o f 6 percent. Should bilat eral rates violate these margins, however, the cen tral banks in control o f the two currencies are expected to intervene in foreign exchange markets in amounts necessaiy to bring the rates back into the agreed-upon ranges.8 The foregoing discussion represents a simple characterization o f EMS policy coordination. The most important exception to this characterization for this study is that, although the U.K. exchanged 5Ungerer, et al., table 10. 6lbid, p. 12. 7lbid, p. 13. 8See Ungerer, et al., pp. 4-8, for a discussion of how interven tions are conducted by the central banks of nations that partici pate in the exchange rate mechanism (ERM). 21 Table 1 EMS Realignments: Percentage Changes in Bilateral Central Rates German mark 1979 9/24 11/30 1981 3/23 10/5 1982 2/22 6/14 1983 3/21 1985 7/22 1986 4/7 8/4 1987 1/12 Belgian franc Danish krone French franc Irish pound Italian lira Dutch guilder -6 -3 + 5.5 -2 .8 6 -4 .7 6 +2 -3 + 5.5 - 8 .5 -3 + 4.25 + 5.5 +2 +3 + 1.5 +2 +1 + 2.5 +2 +1 +3 +2 - 5 .7 5 - 2 .5 +2 -3 -8 - 3 .5 +2 -2 .7 5 - 2 .5 -6 + 4.25 + 3.5 +2 +3 +2 SOURCE: Deutsche Bundesbank, Intereconomics (September/October 1987). gold and dollar reserves for ECUs, it did not agree to participate in the cooperative effort to stabilize exchange rates.9Thus, while the U.K. is an EMS member, its exchange rate is not specifically tied to those o f the other EMS nations. To make this distinction, the EMS countries that participate in the exchange rate mechanism (ERM) often are referred to as the ERM countries. The ERM Has Reduced Exchange Rate Variability Various studies have concluded that the ERM has significantly reduced the variability o f ex change rate movements among the m em ber coun tries. Table 2, reproduced from an IMF study by Ungerer, et al. (1986) provides one indication o f how much the variability o f m onthly average nom inal exchange rates, as measured by the coefficient o f variation, declined after the EMS was form ed; a similar pattern emerges if one examines data for real exchange rates (nominal exchange rates ad justed by CPIs) or other measures o f variability, such as standard deviations; these reductions in bilateral exchange rate variability between ERM 9Greece, Portugal and Spain also do not participate in the exchange rate mechanism. 10lbid, pp. 4 -5 and pp. 18-21. Also see related evidence, pro vided by Rogoff (1985a), who found that bilateral exchange rates between EMS members have become more predictable. "S ee Ungerer, et al., tables 16-21. The coefficient of variation is the standard deviation of a series divided by its mean. 12A contrary view is presented by Fels (1987). He argues that, because only n-1 bilateral rates in an n-exchange rate system are freely determined, the ERM really is nothing more than a participants are statistically significant." Finally, as depicted in the bottom portion o f table 2, the IMF analysis indicates that exchange rates for nonERM countries, such as the United Kingdom, the United States and Japan, generally experienced increased variability in the post-1979 period. Thus, relative to the exchange rate behavior o f non-ERM industrial countries, the ERM has significantly reduced fluctuations in the real and nominal bilat eral exchange rates among its members.12 ECONOMIC POLICY COORDINATION: A MORE GENERAL ANALYSIS The ERM has achieved greater exchange rate stability. The usefulness o f such policy coordina tion, however, must be judged ultimately on the basis o f relative econom ic performance. This more general criterion for judging the efficacy o f such coordination is important because econom ic the ory does not suggest that stable exchange rates, per se, guarantee generally desirable econom ic outcomes. dollar/Dmark system that pulls other exchange rates with it. More important, he argues that the ERM appears to have succeeded in the early 1980s only because the dollar’s real value had risen sharply and stimulated export sales from ERM countries to the United States. As a consequence, member nations did not feel the need to pressure Germany to lead a currency devaluation through expansionary measures. Fels also conjectures — and is supported by recent developments — that realignments or other pressures on the ERM will occur as the dollar weakens. JULY/AUGUST 1988 22 Exchange Rate Stability, Economic Policies and Economic Performance Are Not Necessarily Related! The ERM does not specify explicitly that m em ber nations must coordinate policy actions. In other words, although the ERM members may agree to specific ranges on bilateral exchange rates, maintaining those ranges may be achieved, in principle, by a w ide variety of policy actions. To illustrate this point, consider a simple m odel o f the nominal exchange rate: (1) e = (m* - m) - h(i*-i) - k(y*-y) m onetary financial real policy market output measure conditions conditions where: e = the exchange rate foreign $ domestic $ m = the nominal m oney supply; i = the nominal interest rate; y = real GNP; k = the incom e elasticity o f real m oney demand; h = the interest response o f real money balances; and indicates values in a foreign country. All variables in equation 1, except the interest rate, are expressed as natural logarithms.13The equa tion implies that a countiy's currency w ill depreci ate (one unit o f domestic currency w ill purchase few er units o f the foreign currency) if domestic m oney growth accelerates, domestic nominal in terest rates decline or domestic real econom ic growth slows relative to changes in the equivalent measures in a foreign economy. Once one recognizes, as in equation 1, that differences between domestic and foreign eco nomic values determine the level o f exchange rates, one can see clearly that a stable value for the nominal exchange rate is consistent with many different econom ic and policy environments and outcomes. For example, two countries could ex hibit individually real growth o f plus or minus 3 percent; as long as the difference between their real growth rates remained unchanged, however, the exchange rate, ceteris paribus, w ould be sta ble. Similarly, inflation in each countiy could be 20 percent or zero; other things the same, however, l3This model, taken from Dornbusch (1980), is based on the standard monetary approach to the balance of payments. http://fraser.stlouisfed.org/ FEDERAL RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis Table 2 Some Representative Comparisons of Monthly Average Variability in Nominal Exchange Rates1 Country 1974-78 1979-85 ERM Belgium Denmark France Germany Ireland Italy Netherlands 20.3 25.0 31.6 29.2 36.0 36.0 21.1 13.6 14.8 15.9 16.3 12.2 19.3 13.2 Non-ERM Austria Canada Japan Norway Sweden Switzerland United Kingdom United States 20.3 44.1 44.5 25.3 30.2 44.0 32.7 34.7 12.3 52.0 48.1 24.2 31.6 25.9 37.8 55.7 'Adapted from table 16, Ungerer, et al. (1986). Figures are average values for the coefficient of variation, based on bilateral nominal exchange rates weighted by MERM weights. the exchange rate w ould be stable so long as the inflation differential w ere stable. Thus, stable exchange rates can be observed under a w ide range o f econom ic policies and conditions. Equation 1 also points out that the exchange rate can be affected by policy actions in either the domestic or foreign country. If, for example, e were the French franc/DM exchange rate and the DM were rising (e, measured as French francs per DM, w ould be rising), e could be decreased (the DM made to decline) by increasing the German m oney stock relative to the French m oney stock. One w ay in which this might be accomplished w ould involve the Bundesbank and/or the Bank of France selling DM-denominated assets and buying franc-denominated assets, thus increasing the supply of marks and reducing the supply of francs. These changes in the markets for the franc and mark effectively w ould change the relative franc/DM price, that is, the exchange rate. 23 Notice, however, the effects o f such an action. The m oney supply w ou ld expand in Germany and decline in France. First, if the Bundesbank were pursuing m oney growth w ithin specified target ranges, the need for intervention o f the sort d e scribed could well lead to money growth above the announced target path. Moreover, depending upon the magnitude and duration o f intervention, the pursuit o f a stable exchange rate (and its effects on the German m oney stock) could cause a rising price level in Germany; other short-run ef fects on output, unem ploym ent and interest rates could be observed as well. Thus, in this one illus tration, the two countries could achieve one objec tive at the expense o f failing to attain others 14 ECONOMIC PERFORMANCE BEFORE AND AFTER THE ERM W hether exchange rate stability has im proved econom ic performance or brought about greater policy convergence among ERM countries is an empirical issue. In this section, this issue is as sessed in two com plem entary ways. ERM vs. N on-E R M Economic Performance: Another Look at the Evidence To compare econom ic conditions before and after the ERM agreement, a set o f monthly data for m ajor indicators o f policy actions and econom ic performance in the ERM countries and selected large non-ERM econom ies was assembled. The test consists o f comparing the average growth rates and variances o f the narrow (M l) m oney stock, CPI and index o f industrial production and the average levels and variances o f short-term interest rates between two periods: February 1975-Februaiy 1979 (before ERM), and April 1983Decem ber 1987 (the “stable” ERM period). The interval between March 1979—March 1983, which IMF analysis has characterized as “frequent peri ods o f exchange market strain and numerous con sequent realignments o f central rates,” was not examined.15The transition period was om itted to focus on the comparison between the presumably less stable pre-ERM period and the relatively sta ble ERM period. 14For more general treatments of how policies and economies are linked, see Frenkel (1986) or Kahn (1987). ,5Ungerer, etal., p. 11. Specific hypotheses to be investigated with these comparisons include the following: If greater exchange rate stability brought about higher out put growth and low er inflation, a comparison of period 1 versus period 3 should reveal significantly higher output growth (as measured by industrial production) and significantly low er inflation rates (as measured bv CPIs) in the later period than in the earlier one. If these conditions are produced by the ERM, the same indicators for the non-ERM countries should exhibit significantly different, less beneficial output and price performance. Equation 1 implies that stability in nominal exchange rate levels may be associated with greater volatility in m oney growth, interest rates or output, the equation’s right-hand-side arguments.1” If this is the case, measures o f variability for these variables may have increased significantly in the ERM countries since 1979. Conversely, equation 1 w ou ld im ply no change in the variability o f these variables since 1979 in the non-ERM countries that did not attempt (at least explicitly) to reduce bilat eral exchange rate variability. Some caution in making these comparisons is necessary because they rest on a ceteris paribus assumption. The simple tests used here do not control for the effects o f events that are unique to some countries (for example, a crop failure in Europe) or the differential effects across countries o f a com m on phenom enon (for example, the en ergy price decline o f the 1980s). Thus, rather than attributing a specific result — for example, a change in average m oney growth rates or the vari ance o f interest rates — to the ERM, the com pari sons are intended solely to reveal consistent pat terns o f change in the ERM and non-ERM countries. If there are consistent differences in the econom ic or policy perform ance between the ERM and non-ERM nations, it may be an initial indication o f the possible effects o f exchange rate coordination. Differences in the Average Values o f Selected Economic Indicators The results in table 3 examine the econom ic measures that the simple theoretical m odel sug gested as important in achieving greater exchange rate stability. The table 3 entries compare the ,6Wood (1983), examining data for all EMS countries, found greater nominal exhange rate stability to be associated with greater variation in unanticipated interest rate changes in all cases except Ireland. JULY/AUGUST 1988 24 Table 3 Mean Values of Major Economic Indicators Country ERM Belgium Denmark France Germany Italy Netherlands Ireland Non-ERM United Kingdom United States Canada Japan Austria Nora/ay Sweden Switzerland Period Money growth (M l) Inflation (CPI) Short-term interest rates 2/75-2/79 4/83-12/87 2/75-2/79 4/83-12/87 2/75-2/79 4/83-12/87 2/75-2/79 4/83-12/87 2/75-2/79 4/83-12/87 2/75-2/79 4/83-12/87 2/75-2/79 4/83-12/87 6.72’ 5.792 13.70 17.132 10.96 8.652 10.75 6.54 18.00* 11.64*3 10.35 6.00 19.98* 7.50* 6.77* 3.36* 8.72* 4.70* 9.08* 4.75* 3.72* 1.24* 13.95* 7.35* 6.21* 1.43* N.A. N.A. 5.93* 7.51* 11.67* 10.11* 8.25’ 9.84* 3.85* 4.84* 12.84* 14.95* 5.45 5.72 N.A. N.A. 2/75-2/79 4/83-12/87 2/75-2/79 4/83-12/87 2/75-2/79 4/83-12/87 2/75-2/79 4/83-12/87 2/75-2/79 4/83-12/87 2/75-2/79 4/83-12/87 2/75-2/79 4/83-12/87 2/75-2/79 4/83-12/87 15.71 17.222 6.79 8.85 7.76* 17.35*2 10.01 5.39 7.87 5.81 10.01 14.395 12.24 4.092 10.39* 1.61 *2 13.69* 4.59* 6.96* 3.46* 8.16* 4.11 6.35* 1.27* 5.30* 2.71* 7.92 6.50 9.50* 5.88* 1.84 2.022 9.46 9.98 6.23* 8.15* 8.53“ 9.44 6.70* 5.39* N.A. 3.72 8.40* 12.89*5 8.10* 11.23* 1,62*6 3.00* Growth of industrial production 6.18 2.822 4.19 5.192 3.79 1.42 3.11 2.99 14.22 3.132 2.57 2.85 7.18 8.55 2.75 3.20 6.61 5.23 4.09 7.162 6.23 5.88 3.53 3.322 6.20 12.24 -2 .6 9 4.73 N.A. N.A. All data are monthly. Asterisks denote that values are statistically different at the 0.05 level. 'Data begin in 1976.02. 2Data end in 1987.09. 3Data end in 1987.06. ■'Data begin in 1978.05. 5Data end in 1986.12. 6Data begin in 1975.09. SOURCE: International Financial Statistics, International Monetary Fund. mean values for major econom ic indicators prior to 1979 and since 1983; entries designated with an asterisk are values that differ significantly between the tw o periods shown. The data show that the inflation rate o f each ERM country has been reduced significantly since 1983. Some observers expected this result from an ' 7DeGrauwe and Verfaille, pp. 29-30, also show that the uncoordinated policy actions of non-ERM industrialized economies achieved lower average rates of inflation, and did so more FEDERAL RESERVE BANK OF ST. LOUIS exchange rate agreement, arguing that the policies o f low inflation countries, such as Germany, could dominate those o f the high inflation countries, such as Italy. The bottom portion o f table 3, h ow ever, indicates that inflation rates in the United Kingdom and other non-ERM countries — despite the absence o f any explicit exchange rate agree ment — also were significantly reduced.17This quickly, than the coordinated ERM actions. This result is consistent with the theoretical reasoning in Rogoff (1985b). 25 Table 4 Variances of Major Economic Indicators Country ERM Belgium Denmark France Germany Italy Netherlands Ireland Non-ERM United Kingdom United States Canada Japan Austria Norway Sweden Switzerland Period Money growth (M1) Inflation (CPI) Short-term interest rates 2/75-2/79 4/83-12/87 2/75-2/79 4/83-12/87 2/75-2/79 4/83-12/87 2/75-2/79 4/83-12/87 2/75-2/79 4/83-12/87 2/75-2/79 4/83-12/87 2/75-2/79 4/83-12/87 188.80' 181.012 591.03* 2.406.07*2 176.02* 707.15*2 146.27 177.64 172.73* 91.25*3 507.71* 153.99* 358.50 409.71 15.35 15.03 130.37* 37.27* 7.85 10.85 11.50 7.37 54.18* 15.66* 42.69* 19.64* N.A. N.A. 5.06* 2.87* 27.22* 2.56* 1.57* 3.83* 0.45 0.53 8.56 7.03 11.65* 0.29* N.A. N.A. 2/75-2/79 4/83-12/87 2/75-2/79 4/83-12/87 2/75-2/79 4/83-12/87 2/75-2/79 4/83-12/87 2/75-2/79 4/83-12/87 2/75-2/79 4/83-12/87 2/75-2/79 4/83-12/87 2/75-2/79 4/83-12/87 298.06 417.332 17.89* 48.71* 96.47* 518.58*2 289.58* 712.13* 311.31 205.05 1,334.30* 580.04*5 815.42* 1,489.48*2 618.98* 146.38*2 100.73* 26.00* 9.17 6.86 18.88* 7.99*2 97.78* 42.50* 25.82 29.51 45.21* 25.83* 34.73 31.72 10.36 14.402 5.50* 1.40* 2.36 2.32 2.68< 2.05 6.06* 1.57* N.A. 0.18 6.50* 1.32*5 2.90 4.27 1.036 1.28 Growth of industrial production 12,371.77* 2,826.57*2 2,904.63 2,982.492 645.52 439.13 302.35* 1,116.63* 29,548.77* 749.01 *2 508.84* 1,808.63* 2,018.50 2,266.872 715.70* 158.17* 75.98 63.41 214.42* 443.98*2 162.64* 287.42* 529.33* 1,135.33*2 1,610.54* 17,020.64* 408.92* 1,455.67* N.A. N.A. All data are monthly. Asterisks denote that variances are statistically different at the 0.05 level. 'Data begin in 1976.02. 2Data end in 1987.09. 3Data end in 1987.06. “Data begin in 1978.05. 5Data end in 1986.12. 6Data begin in 1975.09. SOURCE: International Financial Statistics, International Monetary Fund. result suggests that some common, worldw ide phenom enon is a m ore likely source o f low er in flation rates observed among the industrialized countries than the policy coordination associated with the ERM nations. The remainder of the data in table 3 fail to iden tify any unique econom ic circumstances associ ated with the ERM group alone. M oney growth declined significantly for two ERM countries, in terest rates rose in four (and fell in one) and indus trial production was statistically unchanged in all seven. The non-ERM group also displayed gener ally higher interest rates and unchanged indus trial production growth. Thus, there is no change in the average value o f a particular econom ic indi cator that can be identified uniquely w ith the ERM countries. Variation in Economic and Policy Indicators The results in table 4 show a mixed pattern of performance with respect to the variances o f the JULY/AUGUST 1988 26 Table 5 Growth in Real Trade Flows (Exports plus Imports) Period ERM country Belgium Denmark France Germany Ireland Italy Netherlands Non-ERM country Austria Canada Japan Norway Sweden Switzerland United Kingdom United States With ERM countries With non-ERM countries 1973-78 1979-86 1973-78 1979-86 1973-78 1979-86 1973-78 1979-86 1973-78 1979-86 1973-78 1979-86 1973-78 1979-86 7.7% 2.2 12.8 3.8 6.8 2.9 7.0 2.9 20.0 9.4 4.6 8.1 6.2 2.4 12.5% 4.5 5.2 3.2 8.5 5.7 7.6 6.8 8.2 4.8 5.4 10.1 10.5 5.3 1973-78 1979-86 1973-78 1979-86 1973-78 1979-86 1973-78 1979-86 1973-78 1979-86 1973-78 1979-86 1973-78 1979-86 1973-78 1979-86 11.3 4.8 4.4 6.5 11.6 9.5 8.2 9.3 6.0 4.6 8.0 4.5 11.4 5.4 4.4 7.3 4.2 4.0 4.2 7.2 7.2 10.0 9.7 6.6 3.8 4.9 6.8 3.8 5.1 4.2 4.2 8.5 Data are nominal trade values, reported in dollars by the IMF, adjusted by the U.S. GNP deflator and the Federal Reserve Board's trade-weighted exchange rate index (TWEX). assorted econom ic and policy indicators. Short term interest rates w ere significantly less variable in the third period for three o f the ERM countries but significantly more variable in France. Similarly, industrial production became significantly less variable in Italy, but more volatile in two other countries, especially in Germany, where the vari ance o f industrial production increased by a factor of four. In contrast, for the United Kingdom, which does not participate in the ERM, the variances o f the inflation rate, interest rates and industrial produc tion all declined significantly. While the variances o f m oney growth increased significantly in four non-ERM countries, two o f these countries and four non-ERM nations overall achieved less vari FEDERAL RESERVE BANK OF ST. LOUIS able inflation rates. Three non-ERM countries experienced significantly less interest rate volatil ity, w hile none experienced greater variability. Five o f the eight non-ERM countries saw greater output variability in the more recent period. Overall, as in table 3, this mixed picture does not yield any uniquely beneficial results associated with ERM membership. Exchange Rate Variability and Trade Flows A somewhat different result emerges when data on trade flows are examined as in table 5. One possible result o f reducing exchange rate variabil ity is that the greater exchange rate certainty might increase trade flows. Since exchange rate 27 variability did decline among ERM countries but increased both among the non-ERM nations and between the ERM and non-ERM countries, it is interesting to see how trade flows changed after 1979 both within the ERM group and between the ERM and non-ERM nations. Table 5 shows that the growth o f intra-ERM trade declined in ERM econom ies (except Italy) during the period o f greater exchange rate stabil ity. In contrast, trade by non-ERM members both with each other and the ERM group often rose, even though these exchange rates became more variable. Canada, Japan and the United States are the notable cases o f this result. On the basis of these results, again holding other things constant, greater ERM exchange rate stability was not asso ciated with relatively larger intra-ERM trade. SUMMARY Proposals for policy coordination among the major industrial econom ies have been discussed more frequently in recent years. Initially such proposals were intended to correct what were perceived as problems created by a “ high” value o f the U.S. dollar; subsequently, they w ere intended to mitigate the adverse consequences o f variable exchange rates and the falling value o f the dollar. One attempt to coordinate domestic policies in recent years in pursuit o f stable bilateral nominal exchange rates is found in the EMS. Evidence based on data before and after the establishment o f the EMS suggests that, w hile bilateral exchange rates have becom e more stable, other measures of econom ic performance and policy actions fail to show the effects o f such coordination. Lower in flation rates in ERM countries have been matched by low er inflation rates in major non-ERM econo mies. Other variables, such as money growth, inter est rates and real output measures also show no consistent differential response in ERM and nonERM countries in recent years. The data do not even show that intra-ERM trade has increased any more than trade with non-ERM countries, despite the reductions in exchange rate variability among ERM nations. Overall, the only experience w e have with conceited policy coordination does not indi cate that general econom ic or policy measures have been much affected — one way or another — by such coordination. REFERENCES Bank for International Settlements. 1987). Crook, Clive. “ Living With Uncertainty," supplement to The Economist (September 26, 1987). Currie, David, and Paul Levine. “ International Cooperation and Reputation in an Empirical Two-Bloc Model,” paper presented at the Konstanzer Seminar on Monetary Theory and Mone tary Policy, Konstanz, West Germany (June 17, 1987). DeGrauwe, Paul. “ International Trade and Economic Growth in the European Monetary System,” European Economic Review (1987), pp. 389-98. _________“ Exchange Rate Variability and the Slowdown in Growth of International Trade," IMF Staff Papers (March 1988), pp. 63-84. DeGrauwe, Paul, and Guy Verfaille. “The European Monetary System: An Evaluation,” mimeo (University of Louvain, Janu ary 1986). Dornbusch, Rudiger. “ Exchange Rate Economics: Where Do We Stand?” Brookings Papers on Economic Activity (1:1980), pp. 143-85. Farrell, Victoria, et al. “ Effects of Exchange Rate Variability on International Trade and Other Economic Variables," Staff Study No. 130 (Board of Governors of the Federal Reserve System, December 1983). Fels, Joachim. "The European Monetary System 1979-87: Why Has It Worked?" Intereconomics (September/October 1987), pp. 216-22. Fischer, Stanley. “ International Macroeconomic Policy Coordi nation,” Working Paper No. 2244 (National Bureau of Eco nomic Research, May 1987). Franke, Gunter. “ Exchange Rate Volatility and International Trade: The Option Approach,” Working Paper (revised), (Universitat Konstanz, August 1987). Frankel, Jeffrey, and Katherine Rockett. “ International Macroec onomic Policy Coordination When Policymakers Do Not Agree on the True Model," American Economic Review (June 1988), pp. 318-40. Frenkel, Jacob A. “ International Interdependence and the Constraints on Macroeconomic Policies,” Weltwirtschaftliches Archiv (Heft 4, 1986), pp. 615-45. Kahn, George. “ International Policy Coordination in an Interde pendent World,” Federal Reserve Bank of Kansas City Eco nomic Review (March 1987), pp. 14-32. Karamouzis, Nicholas V. “ Lessons from the European Mone tary System,” Federal Reserve Bank of Cleveland Economic Commentary (August 15, 1987). Organization for Economic Cooperation and Development. “ Cooperative Policy Action to Restore Satisfactory Growth,” Economic Outlook No. 41 (OECD, June 1987). Rogoff, Kenneth. “ Can Exchange Rate Predictability Be Achieved Without Monetary Convergence? Evidence from the EMS," European Economic Review (1985a), pp. 93-115. _________“ Can International Monetary Cooperation Be Coun terproductive?” Journal of International Economics (May 1985b), pp. 199-217. Ungerer, Horst, Owen Evans, Thomas Mayer and Philip Young. The European Monetary System: Recent Develop ments, Occasional Paper No. 48 (Washington, D.C.: Interna tional Monetary Fund, December 1986). Wood, Geoffrey E. “The European Monetary System — Past Developments, Future Prospects, and Economic Rationale," in Roy Jenkins, ed., Britain and the EEC (London: Macmillan, 1983). Annual Report, (Basle, JULY/AUGUST 1988 28 Appendix The European Currency Unit The European Currency Unit (ECUl seives primarily as a unit o f account for a variety of functions within the European Community (EC). For example, the value o f the ECU is a reference point from which to judge the divergence of individual currency values from desired values. More generally, the ECU is a unit o f account for the EC's budget, its Com m on Agricultural Policy and its other finance and credit activities. The ECU originally had been intended to seive also as a means o f settlement and a reserve asset. In both cases, however, its use has been small. It is rarely used as a means o f settlement and, as a reseive asset, is largely a substitute for the gold and dollar deposits a m em ber countiy gave up to join the EMS. The ECU itself is simply a weighted-basket of EMS member currencies. As shown in the table, as o f September 17, 1984, one ECU was equal to the market value o f 3.71 Belgian francs, 0.219 Danish krones and so on across the 10 EMS currencies. Over time, both the weights attached to member currencies and their market values relative to nonEMS currencies have changed so that the value o f the ECU has varied (see chart on opposite page). The private use o f ECUs, however, is a different matter. Because it represents a basket o f EC cur rencies and because a formal agreement exists to keep constituent currencies within specified bounds, investors have viewed financial instru ments denom inated in ECUs to be less risky than similar instruments denom inated in a specific currency. For this reason, sight and time deposits, loans and bonds all have been offered denom i nated in ECUs. Thus, the ECU may be view ed best as a currency index unit o f account that varies less than its constituent currencies. Representative Composition of the ECU Currency National currency units September 17,1984 Percentage weights September 17,1984 3.71 0.219 1.31 0.719 0.00871 140.00 0.14 0.256 0.0878 1.15 8.2 2.7 19.0 32.0 1.2 10.2 0.3 10.1 15.0 1.3 Belgian franc Danish krone French franc Deutsche mark Irish pound Italian lira Luxembourg franc Netherlands guilder Pound sterling Greek drachma 100.0 SOURCE: Ungerer, et. al. (1986), table 4. FEDERAL RESERVE BANK OF ST. LOUIS 29 Chart 1 Value of the ECU in U.S. Dollars Dollars 1979 80 81 NOTE: Data are period averages. Dollars 82 83 84 85 86 87 1988 JULY/AUGUST 1988 30 Clemens J.M. Kool and John A. Tatom Clemens J. M. Kool is an assistant professor of monetary eco nomics at Erasmus University, Rotterdam, The Netherlands. John A. Tatom is an assistant vice president at the Federal Reserve Bank of St. Louis. Anne M. Grubish provided research assistance. International Linkages in the Term Structure of Interest Rates I NTF.RF.ST rales usually differ for assets w illi different terms to maturity.1The term structure of interest rates shows the relationship among the interest rates, or yields to maturity, o f differentlived assets and their terms to maturity.2Analysts often view the term structure as the link between current and future short- and long-term interest rates. This link is important because o f the w idely held belief that monetary authorities are able to influence only short-term m oney market rates, w hile long-term rates are more relevant in making investment and consumption decisions. An un derstanding o f the transmission mechanism from current short-term rates to future interest rates is crucial, according to this view, in implementing and evaluating monetary policy. This article extends the analysis o f the term structure by examining whether movements in the domestic term structure are influenced by foreign interest rate developments. In the term structure 'A given yield curve implicitly assumes that other characteristics of the short- and long-term assets are identical. Yields on financial assets differ for many reasons, including differences in default risk, marketability and tax treatment. In term structure research, it is typical to examine short- and long-term govern ment securities and to assume that differences in their maturity are the main determinant of the differences in their yields. See Wood (1983). 2ln this article, the terms interest rate and yield to maturity are used interchangeably. Both measures reflect the average expected rate of return over the remaining life of the underlying financial asset. These measures usually will differ from the FEDERAL RESERVE BANK OF ST. LOUIS view, if interest rates are related across countries, then foreign short-term interest rate movements are transmitted both directly to changes in dom es tic short-term rates, and indirectly, via the respec tive domestic term structures, to changes in long term interest rates. Thus, short- and long-term interest rate changes are closely correlated across countries and long-term interest rates changes in one country are related to foreign short-term rates as well. This hypothesis is com m only thought to hold for the United States, whose policies are blamed for adverse interest rate developments abroad. This article examines the relationships among the term structures o f interest rates in the United States, Canada, the United Kingdom, West Germany and Japan. Our ch ief focus is on the extent to which movements in short- and long term interest rates are related internationally and w hether changes in foreign interest rates in fluence domestic interest rates. holding-period return, which equals the return on an asset over a fixed period. For example, the one-year holding-period return on a 10-year bond is the annual coupon payment plus the capital gain over the year, while the yield to maturity on the same bond is the average of all current and expected future one-year holding-period returns until the end of the bond’s life. Only when the holding period and the remaining maturity of the bond are equal will these two measures of return coincide. An approximately linear negative relation holds between the change in the long-term interest rate and the holding period return. For details, see Shillerefa/. (1983). 31 TERM STRUCTURE THEORY The expectations theory is the principal theoret ical approach to the term structure o f interest rates. This theory assumes that investors view short- and long-term government bonds as perfect substitutes, that is, investors are indifferent to the maturity o f holdings o f government securities.3 This assumption implies that eveiy investment strategy in government securities has the same expected return over any given future holding period. For example, suppose that the current one-year rate is 6 percent, w hile the interest rate on a 10-year bond is 8 percent. If investors expect future one-vear rates to remain constant at 6 per cent, it is m ore attractive n ow to buy the 10-year bond. Over a 10-year holding period, the 10-year bond vields an 8 percent annual rate o f return; rolling over a sequence o f one-year assets for the same 10-year period yields an expected rate o f only 6 percent. Increased purchases o f 10-vear bonds bid up their price, thereby depressing their yield; similarly, sales o f existing one-year securities (to switch into higher-yielding 10-year bonds) low er their price and raise their expected yield. This process continues until the long rate equals an average o f current and expected future short rates. a positive (negative) spread, or difference, between today’s long- and short-term interest rates will reflect it. DOMESTIC TERM STRUCTURE EQUATIONS According to theories o f the domestic term structure o f interest rates, the current spread be tween the level o f long- and short-term interest rates is directly related to the expectation that future short-term interest rates and, therefore, long-term rates w ill be higher. Formally, this rela tionship for the long-term interest rate, R„ can be written as: 111 AR, = P0 + P,S, , + e„ where S, , is the lagged spread, (R, , — i, ,), and i, , is the lagged short-term interest rate. In theory, the parameter (30is expected to be negative and (3, is expected to be positive.4The intercept (3„ is the negative o f a term or risk premium. Long-term interest rates, on average, will not change (AR,+i = 0) when the spread, S„ equals the ratio o f ( _ (3„/f3,|; this ratio is positive when there is a positive term premium and zero w hen (30is zero. Long-term rates are expected to rise (fall) when the current spread exceeds (is less than) ( —(3„/p,). Thus, the interest rates on a three-month bill, a one-year bill and a 30-year bond w ill differ accord ing to the market’s assessment of expected inter est rates on short-term assets beyond the life of each instrument. The one-year bill w ill have a yield that reflects not only the expectation for the next three months em bodied in the three-month bill rate, but also the expectations for the subse quent nine months. Similarly, a 30-year bond will have a current vield to maturity that is influenced by the same expectations as the one-year bill for the first year, but also by expectations for the re maining 29 years o f its life. If future short-term rates are expected to be higher (lower) than today, A similar equation can express the same relation for short-term rates (Ai,). Term structure theory does not provide a theoretical value for the spread coefficient in such an equation, since expected changes in future short rates are not necessarily distributed uniformly over each future period's short rate. Thus, a rise in the spread need not indicate that the short rate w ill be higher next month, although it does indicate that some unspecified future short rates (and long rates for assets that span the period) w ill be higher than they are now. W hen expected future short-term interest rate changes are distributed uniformly or w eighted more heavily to the near future, which 3There are several competing theories of the term structure of interest rates. For a detailed description of these theories, see the discussion in a textbook such as Wood and Wood (1985), chapter 19. They all, however, have a common foundation in what sometimes is called the traditional expectations theory. The purpose of this article is not to test competing theories of the domestic term structure; instead, it uses a general specifi cation as a point of departure to study international linkages among interest rates. holding-period return on a one-month T-bill is its interest rate, it; the one-month holding-period return on a 10-year bond is the interest rate R, plus the annualized expected capital gain; this capital gain is inversely proportional to the expected one-month change in the 10-year bond rate, AR,,,. The equality results in a (3, measure that is positive, but very small in a monthly analy sis, and proportional to the average in-sample long rate; the expression for 0Oequals ( - TP)p, where TP is the average term premium. See Mankiw (1986) for an example of this derivation. 4Theoretical expressions for the values of |30 and p, are found by equating holding-period returns, adjusted for any term premium. Consider a monthly analysis of one-month Treasury bills and 10-year bonds, like that below. The one-month JULY/AUGUST 1988 32 generally is the case in theoretical or empirical investigations, the spread coefficient in a short term rate equation should also be positive.'1 A variant o f the standard term structure equa tion used in many macroeconom ic models is: (2) S, = a„ + a, S,., + a, Ai, + a., Ai, , + e,. In equation 2, the long-term interest rate is a long distributed lag o f past short-term rates." Equation 2 can be rewritten as: (3) AR, = p0 + p,S,_, + (3, Ai, + Ai, _, + e, Equation 3 adds information on the current change in the short rate and a lagged value o f this change to equation l.7One rationale for adding “new s” about the short rate is that short-rate changes reflect new information about expected future short rates beyond the information con tained in the recent spread." If markets are ef ficient and adjust within one period, (3a should equal zero and (3, should be positive. Equations like this are w idely used to study the term struc ture o f interest rates empirically. Accordingly, we use it as a point o f departure in investigating inter national term structure linkages. 5Estimates of long- and short-rate equations like equation 1 often lack predictive content and are systematically at odds with the theory. See, for example, Shiller (1979), Shiller et al. (1983), Mankiw and Summers (1984) and Mankiw (1986). Most studies of the term structure find support for the inclusion of a term, or risk, premium in a term-structure equation: it is com mon to include a constant and nonzero term premium in char acterizations of the expectations theory. See Wood and Wood (1985) or Clarida and Campbell (1987), for examples. More over, financial theory indicates that risk premia also are related to returns in other financial markets, like stocks, and to expec tations about general economic conditions. For example, Cox Ingersoll and Ross (1981) modify the expectations theory to account for a negative effect on the term premium that is pro portionate in magnitude to economic uncertainty. The possibil ity of a nonzero average, or constant, term premium is included in the estimates below. Specifications like equation 1 implicitly presume that all other influences on interest rates in the next period, beyond the current spread information and the term premium, have zero mean and are uncorrelated with the cur rent spread. Such restrictions are relaxed in empirical models like those below. 6See Mankiw (1986). Modigliani and Sutch (1966) used a more famous variation of such an equation; it had a 16-quarter distributed lag on past short rates, instead of the lagged spread term, and so was explicitly backward-looking, rather than forward-looking as the expectations theory emphasizes. 7This transformation is found by subtracting S, , from both sides of equation 2 to obtain AS„ then adding Ai, to both sides to obtain equation 3. Equation 3 also can be derived from equa tion 1 and its short-rate variant. This form, however, suggests that one source of a negative coefficient on the lagged spread is that short rates are more sensitive to recent spread changes, which is likely if movements in the current spread are more informative about near-term prospective short rates than about all future short rates. Testing this alternative is beyond the scope or purpose of this paper. FEDERAL RESERVE BANK OF ST. LOUIS Domestic Macroeconomic Determinants o f the Level o f Interest Rates Term structure and asset price theoiy explain differences in yields over time or among assets, but do not explain the general level o f interest rates. Closer scrutiny o f the factors influencing both short and long rates might indicate addi tional domestic determinants o f the term struc ture that w ould m odify equation 3. The central factors influencing the general level o f nominal interest rates, according to Fisher (1930), are the expected real rate o f return on capi tal and the expected inflation rate.11Economic theory indicates that the expected real rate o f interest is determined by the marginal productiv ity of capital and the marginal utility o f consum p tion. Numerous econom ic factors, however, can impinge, at least temporarily, on these rather ab stract determinants and, hence, on the real rate o f interest.10 One approach to analyzing the term structure of interest rates m odels the effect o f domestic macro8This view implies that an observationally equivalent view of equation 2 is that the lagged spread incorporates unbiased forecasts of future rates as in equation 1, but news reflected in current short-rate movements is informative about revisions of expected future rates. The Modigliani and Sutch variant of equation 2 has been criticized by numerous analysts, including Phillips and Pippenger (1976). The latter show that a forwardlooking, efficient markets model rejects the Modigliani-Sutch model without including the lagged spread. The results below suggest that their specifications can be improved, however. 9These considerations also suggest that the term structure of nominal interest rates is a combination of a term structure of expected inflation and a term structure of expected real rates. When inflation temporarily accelerates (slows) due to a supply shock, the spread shrinks (widens) because the short-term rate rises (falls) more than the long-term rate. Garner (1987) presents evidence for the United States on the close relation ship between the term structure of interest rates and the term structure of inflationary expectations. 10The standard laundry list of other macroeconomic factors includes the money stock, the price level, tax rates, govern ment expenditures and other fiscal variables, and other domes tic real variables such as private sector aggregate demand for goods and services, the business cycle, the mix between consumption and investment, and risk. Both current values and expectations of future values of these variables and their growth rates affect current and expected future real rates. 33 econom ic changes on the two com ponents o f both long- and short-term interest rates." If financial markets are efficient, however, investors will have used the available relevant domestic information to price assets, including government bonds o f all maturities. Thus, no additional domestic informa tion exists that can improve on the im plicit fore cast o f future interest rates reflected in the current term structure. International Term Structure Linkages W hether additional information on foreign in terest rates influences the domestic term structure depends on the exchange rate regime. Under a fixed-exchange-rate system, domestic interest rates and monetary policy are not independent o f foreign developments. Inflation rates tend to be equal across the countries that have a fixed-rate commitment; they equal the rate o f depreciation o f the purchasing pow er o f the com m odity or the m oney against which the exchange value o f the currencies are fixed. In addition, if capital markets are integrated internationally, a change in the real rate o f return in anv one country is transmitted to all nominal rates both domestically and abroad as investors attempt to maximize real rates o f re turn.12Since in this case the expected inflation rate and real rate are closely linked across countries, the nominal interest rate, at all maturities, is also closely linked. Econom ic developments at hom e or abroad could influence the interest rates com mon to all countries, but foreign factors w ould not have an independent influence on a domestic term structure like that shown in equation 3; news o f such developments w ould be fully captured in the Ai, term for the domestic economy. "W ood and Wood (1985) have noted an interpretation problem with such a procedure; do such variables enter as determi nants of a term premium, via a “segmented markets” argu ment, for example, or do they provide additional information on the time path of expected future real interest rates? This prob lem mediates against the arbitrary introduction of current information for such variables. Moreover, the long list of poten tially relevant macroeconomic factors and the dynamics of their effects operating through lags indicate that this approach is difficult and invariably controversial to implement. Even if undertaken successfully, however, the effort could be quite misleading. ,2For a more detailed description of interest rate relations across countries, see Bisignano (1983) and Kirchgassner and Wolters (1987). Glick (1987) provides evidence on the real-interest-rate linkage between the United States and the Pacific Basin coun tries. The relation o f interest rates across countries for a given maturity, called the covered interest parity condition, is (4! i = i* + (f —e ln, e where i and i* = the domestic and foreign interest rates, respectively, for comparable assets with respect to maturity and risk, e = the current or spot exchange rate expressed as the number o f domestic currency units per unit o f foreign currency, f = the corresponding forward rate one period in the future, and n = the annualizing factor for the term o f assets being compared, which equals 12 divided by the number o f months to maturity. Under a credible fixed rate regime, the expected forward rate w ould equal the spot rate at all matu rities, so that countries w ou ld have the same term structure o f interest rates. Domestic news in one country that affects domestic rates and the term structure w ould be im m ediately transmitted abroad, so that (i = i*) w ou ld hold for all maturi ties. Even in the absence o f a credible fixed-rate com mitment, m onetary authorities may still have a long-run exchange-rate objective and may periodi cally intervene in the exchange market or conduct policy to further that goal.'3 If they do, interna tional rates could still be related, although the relationship w ou ld be looser, and changes in short rates especially w ould not be systematically coin cident. rate of appreciation of the foreign currency. Even if risk premia exist in foreign exchange markets, the forward rate will be a close approximation of the future spot rate in most situations, so the second term will approximate the expected rate of appreciation of foreign currency. Meese and Rogoff (1983) argue that this approximation is often unsatisfactory. Covered interest parity has been tested widely and successfully for short-term rates; such tests are typically restricted to short-term assets to ensure that there is an active market in forward contracts for foreign exchange for a comparable period. See Frenkel and Levich (1975,1977) or, more recently, papers that reject the stronger variant called the “ Fisher open" hypothesis or uncovered interest parity, such as Cumby and Obstfeld (1984). The strength of international linkages also depends on the extent to which assets of the same maturity across coun tries are substitutes; different tax regimes, transactions costs or other factors can impair the international interest rate linkage. 13ln the absence of risk premia in foreign exchange markets, the forward rate would equal the expected future spot rate, so the second term on the right-hand side would equal the expected JULY/AUGUST 1988 34 In a “pure float,” or regime with no exchangerate commitments, interest rates across countries can be independent if countries pursue indepen dent inflation rate objectives and if the real inter est rate is constant. Movements in foreign interest rates can be reflected in the prospective change in the exchange value o f the domestic currency, rather than in domestic interest rates. Even in this case, however, the implicit exchange rate change can have undesirable effects on other policy objec tives, such as the price level, so that interest rates will still not be independent across countries. changes o f these interest rates and spreads in each country. The levels o f rates show consider able variability, but the mean long rate exceeds the mean short rate in each country. The mean level o f short rates in Canada and the United Kingdom are not significantly different from each other.16 The same is true for West Germany and Japan, but their mean interest rate levels are low er than those in the other three countries. The mean U.S. short rate is significantly higher than in Japan and West Germany but low er than in Canada and the United Kingdom. In particular, a rise (fall) in a foreign interest rate need not spill over to the domestic rate if the d o mestic currency is free to appreciate (depreciate) relative to the foreign currency. But the apprecia tion (depreciation) o f the domestic currency can depress (raise) the domestic price level as w ell as, temporarily, the inflation rate. While there may not be an explicit exchange rate objective, an in flation or other objective can be at odds with such exchange rate effects and, therefore, require policy actions to raise (lower) interest rates. Hence, for eign interest rate changes can result in equal d o mestic rate changes despite the absence o f an exchange-rate comm itm ent.14Finally, w hen do mestic short-term interest rate movements are only temporarily insulated from foreign m ove ments by central bank intervention, long rates will still reflect these foreign changes immediately. In this case, foreign short-term rates w ould exert an independent effect on domestic long-term interest rates, given domestic short-term rates. The rank ordering o f the mean long rates is the same as for short rates, but the mean levels o f the long rate are significantly different for each pairwise comparison o f countries. The mean spread is not significantly different for four of the country pairs: the United Kingdom and Canada, the United Kingdom and Japan, the United King dom and Germany, and Japan and West Germany; in the other six pairwise comparisons (four o f which are for the United States), the mean spreads are significantly different. The mean o f changes in interest rates is approximately zero for each coun try and maturity class. In each country, the stand ard deviation o f changes in the short rate far ex ceeds the standard deviation o f changes in the long rate, indicating the greater volatility o f short rates in all five countries.17 THE EMPIRICAL RELATIONSHIPS To analyze domestic and international interest rate relationships, end-of-month observations of representative short- and long-term interest rates for the United States, Canada, the United King dom, West Germany and Japan w ere selected for the period from April 1977 to June 1987. This pe riod was chosen on the basis o f data availability.” The Data: Some Simple Statistics The top panel in table 1 shows the means and standard deviations o f the levels and monthly 14The theoretical and empirical basis of this absence of indepen dence under floating exchange rates has been developed extensively by Mussa (1979) and Swoboda (1983). 15A description of the data is contained in the appendix to this article. l6The tests of differences in the means in table 1 use a “ pooled ttest" with a 5 percent significance level. FEDERAL RESERVE BANK OF ST. LOUIS The bottom panel o f table 1 shows correlation coefficients for both levels and changes o f short rates, long rates and the spread for each country. A correlation o f 0.18 or larger in absolute value is statistically different from zero at the 5 percent significance level. The evidence suggests that the short and long rates are highly correlated within each country. Similarly, monthly changes in short and long rates are highly correlated in each coun try except Japan. The spread is dom inated by the short rate; this is indicated by the significant negative correlation between the spread and the short rate in all five countries and the absence o f a significant positive correlation for the spread and the long rate in any country. The level o f the long rate and the spread are insignificantly correlated for the United States 17This smaller long-rate variability reflects the notion that, if the long rate is a weighted average of current and expected future short rates, some short-rate variability over time is expected to average out. 35 Table 1 The Term Structure Data For Five Countries (April 1977 to June 1987) United States Standard deviation Mean Levels Short-Rate (i) Long-Rate (R) Spread (R - i) Changes (A)' Short-Rate (Ai) Long-Rate (AR) Spread (AR - Ai) 8.75% 10.51 1.76 0.01 0.01 - 0 .0 0 2.77% 2.32 1.59 1.18 0.54 1.01 Japan United Kingdom Standard deviation Mean 11.67% 12.39 0.72 0.01 -0 .0 0 -0 .0 1 2.95% 1.88 1.92 0.92 0.59 0.83 Standard deviation Mean 6.30% 7.23 0.93 -0 .0 2 -0 .0 3 -0 .0 1 1.66% 1.27 1.16 0.56 0.28 0.59 Germany Mean 6.44% 7.65 1.21 -0 .0 1 -0 .0 0 0.01 Canada Standard deviation Mean Standard deviation 2.71% 1.34 1.62 10.98% 11.49 0.52 3.20% 2.20 1.76 0.63 0.31 0.57 0.01 0.01 0.00 0.84 0.55 0.75 Correlation of Interest Rates Within Each Country2 0.77 0.82 0, R) -0 .7 8 -0 .5 5 (i, Spread) -0 .2 1 0.03 (R, Spread) 0.71 - 0.65 0.07 0.90 - 0 .9 3 -0 .6 8 0.85 0.76 0.31 0.47 -0 .7 8 0.19 0.13 0.89 0.35 0.43 - 0 .8 7 0.07 0.49 0.76 0.19 (Ai, AR) (Ai, A Spread) (AR, A Spread) 0.52 -0 .8 9 -0 .0 7 Data sources listed in the appendix to this article. 'Data tor changes are from May 1977 to June 1987. C ritical value for 95 percent confidence level is 0.18; for 99 percent confidence level, it is 0.23. and Japan. For the other three countries, this cor relation is even negative. In first differences, the correlations between the long rate and the spread are always positive, though significantly so only in the United Kingdom, Japan and Canada, w hile the changes in the short rates remain significantly negatively correlated with the changes in the spread. Table 2 presents correlations o f rates across countries that allow a preliminary assessment of w hether and how rates are linked internationally; the critical level for 5 percent significance is the same as in table 1, 0.18. All long rates appear to be highly correlated in levels. This correlation is strongest among the United States, Canada and West Germany. The same is true for the level of short rates. W hile changes in long rates are signi ficantly positively correlated across all five coun tries, only the United States and Canada exhibit a '“Monthly rates of increase in consumer prices also support this suggestion. The smallest pairwise correlation over this period is 0.17 for Germany and Japan, and for Canada and the United Kingdom; these are significant at a 6.5 percent significance level. The other eight pairwise correlations exceed 0.33 and are significant at a 1 percent level. On the same basis (and like the long-rate results above), the mean inflation rate is not strong significant positive correlation between changes in short-term interest rates. Changes in short-term rates in West Germany and Canada, as w ell as in West Germany and Japan, are marginally significantly related but the other seven country pair correlations are not. The strong and significant correlations for both long-term interest rate levels and changes across all five countries suggest that these countries ex perienced strongly similar inflationary develop ments between 1977 and 1987.18The fact that there is generally an absence o f a significant correlation between contem poraneous short-rate changes is important for at least two reasons. First, it suggests that the period was characterized by a free float. Second, it suggests that the strong positive corre lation of changes in long rates does not arise through a term structure transmission mechanism that runs from foreign short rates (and associated significantly different for Germany and Japan, for the United States and Canada, or for Canada and the United Kingdom. The other seven pairwise comparisons are significantly differ ent from zero and the rank ordering is the same as for long rates. JULY/AUGUST 1988 36 Table 2 Correlations of Levels and Changes of Interest Rates in Five Countries (April 1977 to June 1987) United States United Kingdom Japan Germany Canada Short-term rate1 United States United Kingdom Japan Germany Canada 1.00 0.58 0.43 0.80 0.88 1.00 0.57 0.68 0.63 1.00 0.63 0.43 1.00 0.90 1.00 Long-term rate United States United Kingdom Japan Germany Canada 1.00 0.71 0.68 0.91 0.97 1.00 0.71 0.80 0.73 1.00 0.80 0.64 1.00 0.91 1.00 1.0 -0 .0 3 0.05 0 .1 7 ' 1.0 0.19 0.09 1.0 0.18 1.0 1.0 0.35 0.32 0.32 1.0 0.53 0.30 1.0 0.48 1.0 Changes in the short-term rate2 United States United Kingdom Japan Germany Canada 1.0 -0 .0 1 0.01 0.13 0.34 Changes in the long-term rate2 United States United Kingdom Japan Germany Canada 1.0 0.33 0.35 0.52 0.80 'Critical value for 95 percent confidence level is 0.18; for 99 percent confidence level, it is 0.23. 2Data for changes are from May 1977 to June 1987. foreign long-rate movements) to domestic short rates and, again via a term structure, to domestic long rates. These implications can be tested more directly using domestic term structure equations. in the equation. The unconstrained version o f equation 3 used here is: Domestic Term Structure Estimates If the constraint in equation 3 holds, (34equals ip3—(3,). Insignificant terms generally are omitted in the estimates o f equation 5 that are reported in table 3. In particular, the i, , term is generally in significant and omitted. In this case, (3, w ould equal —(3, if the spread constraint holds: (34can be compared directly with the coefficient on R, ,. Additional lags o f long-term and o f short-term rates up to four months earlier w ere checked for significance, but their addition to the table 3 equa tions was uniformly rejected. The domestic term structure equation 3 is used to examine international linkages.19The coefficient P, involves the constraint that the effect o f R, is equal and opposite to that o f i, „ given Ai, and Ai, Viewed another way, the term i, , enters equation 3 through three right-hand-side terms (S, „ Ai„ Ai, ,). Therefore, unconstrained estimates o f the equation are used to avoid any bias im posed by the constraint and to examine each term l9Estimates (not reported) of equation 1 for AR, and a short-rate equation for Ai, have little or no explanatory power; only four of the 20 intercept and slope terms are significantly different from zero. These are the positive lagged spread coefficients in the short-rate equations for the United States, United Kingdom and Japan, and a significant negative constant in the Japanese FEDERAL RESERVE BANK OF ST. LOUIS (5) AR, = (3„ + p, R, , + (3, Ai, + 3, i, , - p3i, + £,. equation. Often (three of five cases), the insignificant laggedspread coefficient flj,) in the long-rate equation is negative, contrary to the theory. 37 Table 3 Domestic Long Interest Rate Regressions (September 1977 to June 1987) United States AR, = 0.30 + (1.49) Canada AR, = 0.50 + (1.93) United Kingdom AR, = -0 .0 1 + (-0 .1 2 ) West Germany AR, = 0.91 + (2.32) Japan AR, = -0 .0 2 + (-0 .9 0 ) r2 DW h SE P 0.26Ai, + 0.09i, , (3.42) (7.28) - 0.1 OR, , (-3 .3 4 ) 0.33 2.04 -0 .2 3 0.45 - 0.33Ai, + 0.07i, , (6.20) (2.66) - 0.11R, , (-2 .8 5 ) 0.27 2.00 0.00 0.48 - 0.24 1.87 - 0.52 - 0.25 1.86 1.40 0.27 0.28 (2.17) 0.01 1.92 - 0.27 - 0.30 Ai, - 0.12Ai, 1 (5.81) (-2 .3 3 ) 0.23Ai, + 0.08i, , (6.14) (2.17) - 0.18R,_, (-2 .3 6 ) 0.06Ai, (1.39) For the United States, Canada and West Ger many, the coefficient on the lagged spread, indi cated by that on R, , in table 3, is negative.20The observation o f a negative and significant coef ficient for p, is a rejection o f the expectations hy pothesis, but, as noted above, there may be sound reasons for this com m on empirical result. For our pur poses, all that is important is that R, , has a statistically significant effect on AR, in three o f the countries and, therefore, should be controlled for in testing international linkages. In Japan, neither the lagged long rate nor short rate have any significant effect; in fact, the change in the long rate is essentially uncorrelated with anv domestic interest rate information. No lagged short rate changes enter significantly in equation 5, except in the United Kingdom, where (1, and (3, are equal in magnitude but (3, is zero. The current change in the short rate is the most powerful ex planatory variable for the change in the long rate; for all countries except Japan, a 1 percent change in the short-term interest rate raises the long rate ^These three equations in table 3, and in table 6 below, include lagged dependent variables so the Durbin-Watson d-statistic (labeled DW in the tables) is not the appropriate test for auto correlation. This problem arises in tables 4 and 5 below, as well. The Durbin h-statistic is computed whenever the number of observations is not too large to prevent this calculation. In tables 3 and 6, h-statistics can be computed and they reject the presence of significant autocorrelation. The critical value is 1.65. For the equations for West Germany, a correction for first-order serial correlation is necessary and its estimated coefficient, p, is indicated in the tables. There is no indication of further significant autocorrelation in the equations, however. by about a quarter of a percent in the same month. Table 4 contains similar domestic regression results for the change in the short rate in each country. Past values o f both short-term and long term interest rates for up to four periods were examined sequentially, both individually and jointly, to see if they provided statistically signi ficant explanatory pow er for the change in the short-term interest rates. For all countries, there is a significant positive relation between the current change in the short rate and the first or second month's lagged change in the long rate.2' This is broadly consistent with the expectations theory that indicates the change in the current long rate reflects changes in expected future short rates. If these expectations are realized, the change in the long rate presages these future short-rate changes. In the Canadian case, the short rate is a threemonth rate instead o f the one-month rate that is available for the other countries; the use o f a three-month rate imparts a natural second-order moving average process in the residuals o f this 21The computed h-statistics indicate the absence of significant autocorrelation. The statistic cannot be computed for the short term interest rate equations in table 4 for Canada or the United Kingdom. In these countries, Durbin's alternative test that regresses errors on the lagged error and all right-hand-side variables is used. The coefficient on the lagged error term provides the test statistic for autocorrelation. This coefficient is not statistically significant in either country, so no correction for autocorrelation was computed. JULY/AUGUST 1988 38 Table 4 Domestic Short-Rate Regressions (September 1977 to June 1987) United States Ai, = 1.15 - 0.13i,_, (3.22) (-3 .3 6 ) Canada Ai, = 0.61 + 0.24Ai,_, (2.38) (2.49) United Kingdom Ai, = -0 .3 1 - 0.25Ai,_, (-0 .5 5 ) ( -2 .3 9 ) + 0.15R,_2 (2.09) West Germany Ai, = - 0 .0 0 - 0.16Ai,„, (-0 .0 2 ) (-1 .8 2 ) Japan r2 DW h SE 0.14 1.96 0.21 1.11 - 0.05i,^2 + 0.35AR,_, (-2 .4 5 ) (2.41) 0.20 2.01 N.C. 0.76 - 0.13i, a + 0.57AR,., (-2 .8 0 ) (3.69) 0.13 2.11 N.C. 0.87 + 0.46AR, 2 (2.48) 0.06 2.03 -0 .5 5 0.62 0.35 1.73 1.60 0.46 + 0.73AR, , (3.83) Ai, = - 0 .0 6 0.14i, , + 0.50Ai,_3 (-0 .2 4 ) (-3 .6 5 ) (6.71) + 0.45AR,_, + 0.13R, 2 (2.80) (2.46) N.C. means the h-statistic cannot be computed. equation.2- Correction for this simply affects the standard error o f the reported coefficients; except for the constant and lagged dependent variables, no standard error is reduced (or t-statistic raised I by more than 5 percent, so that the variable selec tion process is unaffected.23 Foreign Influences on Domestic Rates Although in most countries the dynamics ap pear to be m ore com plicated for the short rates than the long rates, the explanatory pow er o f the estimated equation is rather low, except in Japan. In Japan, most o f the explanation comes from the three-period lagged change in the short rate. Out liers are not the source o f this curious dynamic relation whose explanation is unknown to us. In general, only a small fraction o f future short-rate changes is explained with such domestic informa The estimates in table 3 generally show that typical determinants o f the domestic termstructure, like lagged spread information and cur rent short-rate changes, provide significant and similar information across countries. These esti mates can be used to examine w hether foreign short-term interest-rate changes exert an indepen dent influence on the domestic term structure. The correlation evidence above indicates that long rates are systematically linked across countries. “ Hansen and Hodrick (1980) point out this problem for equa tions such as this. Note that this problem could also arise for the United States for data after April 1984 because of data problems described in the appendix, but we could find no evidence of bias due to this in the U.S. equations in tables 4 or 5. 24Bisignano (1983) specifies a long-rate term structure equation that includes either the realized change in the short rate or the news in the short rate; the difference is marginal. He also concludes that current short-rate changes are unpredictable. Krol (1986) examines the impact of current and lagged domes tic short-term interest rate changes on Eurodollar bond rates and doesn’t find a significant effect for lagged changes; only current, U.S. short-term interest rate changes appear to be relevant in explaining Eurodollar bond rate changes. 23The differences in the information content apparently shows up in the positive coefficient on Ai,_,, unlike the negative coeffi cients for this variable in other countries. For one-month rates in other countries, a rise in the rate is systematically related to a subsequent decline; a rise in the three-month rate in Canada is related systematically to a rise in the next month’s threemonth rate. http://fraser.stlouisfed.org/ FEDERAL RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis tion; current changes in the short rates are largely unexplained.24 39 Table 5 Short-Rate Regressions with Changes in Foreign Short Rates as Explanatory Variables (September 1977 to June 1987) United States Ai,us = 1.22 0.14i{?, (3.53) (-3 .6 9 ) Canada + 0.27Ai« AipA = 0.42 (2.97) (1.70) + 0.21A i^ + 0.15Ai,UK (3.57) (2.06) United Kingdom 0.25AW Ai,UK = -0 .3 1 (-0 .5 5 ) (-2 .3 9 ) + 0.1 5R,uk2 (2.09) West Germany Ai«G = 0.00 0.43Ai™°, (-3 .5 2 ) (0.05) Japan 0.16ijlp1 Aiilp = -0 .0 3 ( —0.08) (-3 .0 9 ) + 0.15Ri?2 + 0.17Ai"G (2.16) (2.75) + 0.44ARS (2.16) + R2 DW h SE 0.41 A ifA (3.17) 0.21 2.10 -0 .5 0 1.07 - P - 0.04ipA2 (-1 .7 7 ) + 0.26ARf_A, (1.89) 0.29 2.23 - 0 .2 0 0.72 - - 0.13i,UK2 (-2 .8 0 ) + 0.57AR,UK1 (3.69) 0.13 2.11 N.C. 0.87 - + 0.32Ai/p (3.41) 0.11 1.95 N.C. 0.60 0.32 (2.35) + 0.41 AR;,P1 (2.64) 0.40 1.79 0.44 0.24 (2.09) + 0.35ARJ* (1.82) + 0.53iAi/p3 (7.39) 1.36 N.C. means the h-statistic cannot be computed. The absence o f systematic significant positive cor relations o f changes in short rates, however, raises the question o f w hether and how short or long rates are linked internationally through a term structure relation. The question examined is w hether current interest-rate changes abroad exert an independent influence on domestic inter est rates beyond the influence o f domestic infor mation.25 Table 5 shows the international linkages b e tween short rates. For each country, current changes in all foreign short rates are added to the domestic equation from table 4; insignificant addi tions (individually or as a group) are omitted.28The data show a strong two-way relationship between “ Lagged information from foreign markets should not be impor tant for current domestic changes. Even if such lagged vari ables were significant, those patterns should be unstable over time, reflecting specific occurrences without having anything to do with stable transmission mechanisms. The significance of lagged information was examined; it is significant in some cases, but not stably so. Thus, the results are omitted. 26As in table 4, correction for a second-order moving-average process in the Canadian equation has no effect on the coeffi cients, summary statistics or variable selection. The h-statistic reported for Canada is that found from the moving-averagecorrected standard error of the coefficient for the lagged de changes in short rates in the United States and Canada. A similar relationship exists between Ja pan and West Germany. No foreign influence is significant for the United Kingdom, suggesting that the British authorities follow ed relatively in dependent policies, especially with regard to the exchange rate.27The interaction between the U.S. and Canadian financial markets reflects the large degree o f integration o f these econom ies and their geographical relation. The lack o f a relationship between changes in short rates in the United States and those in West Germany and Japan could be surprising to many analysts. Monetary authorities in West Germany and Japan generally are assumed to have at least pendent variable. The h-statistic indicates that first-order auto correlation is rejected for the United States, Canada and, after correction, Japan. In the United Kingdom and West Germany, the alternative test discussed in footnote 21 above rejects autocorrelation. 270ve r the sample period, the United Kingdom was not part of any exchange-rate system nor was it the focus of international cooperation arrangements. Most of the discussion of interna tional policy coordination focused on the United States vs. West Germany and Japan. JULY/AUGUST 1988 40 Table 6 Long-Rate Regressions with Changes in Foreign Short Rates As Explanatory Variables (September 1977 to June 1987)_______________ United States ARtus = 0.28 + (1.39) + 0.11 A ifA (1.89) Canada ARfA = 0.34 + (1.45) + 0.17Ai,us + (4.72) United Kingdom ARuk = -0 .0 1 + (-0 .1 5 ) West Germany AR»G = 0.57 + (2.10) + + 0.07 A i ^ (3.72) Japan + A R f = - 0 .0 2 (-1 -0 2 ) R2 DW h SE 0.23Aips (5.82) 0.07i,us1 (2.39) - 0.08 R,us, (-2 .5 5 ) 0.34 2.04 -0 .2 0 0.45 0.23AifA (4.49) 0.10AitUK (2.30) 0.07i“ (3.06) - 0.10RtCA, (-2 .8 4 ) 0.40 1.93 0.35 0.44 + 017AipA 0.29 (3.07) 1.96 - 0.50 0.27Ai,UK (5.43) 0.13Ai,UK, -2 .5 3 ) 0.18Ai,WG (5.53) 0.13AipA (4.52) 0.06i,WG, (2.26) 0.04 AipK ( 1.88 ) - 0.12R"0, 0.48 (-2 .2 8 ) 1.88 0.78 0.23 0 .0 5A if (1.18) 0.05Aips (2.19) + 0.08Ai,CA 0.14 (2.82) 1.98 - 0.26 implicit exchange-rate objectives w ith respect to the U.S. dollar, even if they otherwise try to remain as independent as possible from the United States. Nevertheless, no significant linkages between news in the United States, as reflected in its change in the short-term interest rate, and shortrate changes in West Germany and Japan were found. The significant relationship between changes in short-term interest rates in West Germany and Japan may also surprise analysts. Yet this positive relationship, and the absence o f one for the United States and either West Germany or Japan, are sta ble results; both characteristics are found in esti mates for only the first or the last half o f the sam ple period. Central bank exchange-rate policies may not be so sufficiently rigid and automatic that foreign developments are incorporated instantaneously in domestic short rates. If there is a longer-run 28At a point in time, the same information is used to determine both the one-month and the three-month interest rate. The latter, however, reflects expectations for the two months be yond the current one and is influenced not only by current information influencing this month’s one-month rate, buf also by current information that is specific to the subsequent two months. In the simplest expectations model, the three-month FEDERAL RESERVE BANK OF ST. LOUIS 0.23 (2.05) exchange-rate objective, however, foreign changes in short rates w ill contain information about fu ture changes in domestic short rates; therefore, they should produce immediate revisions o f d o mestic long rates. Table 6 displays evidence exam ining this hypothesis. All foreign short-rate changes w ere added to the preferred equations from table 3; only the significant terms are re ported in table 6. An interesting result is that the Canadian shortrate change affects most countries. This phenom e non probably arises because o f the use o f a threemonth rate for Canada. Changes in this yield are more forward-looking, reflecting the expected yield for the month and the subsequent two months.28Thus, the Canadian yield used here contains more information than the other short term rates, so its significance may arise because of this difference rather than unusual properties o f the Canadian financial market. rate is approximately equal to the arithmetic average of the current and two prospective expected one-month rates, or the one-month rate plus two-thirds of the expected change in the one-month rate, one month from now and one-third of the change in the one-month rate, two months from now. 41 The change in the long rate in the United States, given the influence o f the current change o f the U.S. short rate, is independent o f all foreign shortrate movements. The change in the Canadian rate is included in the U.S. equation, despite the fact that it is marginally insignificant, because it is strongly significant in the other three countries. On the other hand, the change in the U.S. short rate enters significantly in the long-term rate equa tions for Canada, Japan and West Germany, sug gesting that these three countries follow implicit exchange-rate policies that involve infrequent and variable interventions in m oney and currency markets, with lags beyond one month. The longrate equations for Canada and West Germany im prove considerably with the inclusion o f foreign short rates. The U.K. long rate is not significantly affected by the U.S. short rate. The results, with the exception o f the U.K. equation, are consistent with a view of the w orld in which foreign financial mar kets react to movements in U.S. short-term interest rates. Except for the problematical Canadian in fluence, the international linkages shown in table 6 are sensitive to the period chosen. One o f the simplest ways to test for temporal stability of re gression estimates is to break the sample period in half to investigate w hether the estimates are signi ficantly different across the periods. Based on such a consideration, the equations in table 6 w ere re-estimated for each half o f the sample p e riod. The significance o f foreign influences virtu ally vanishes when only the last half o f the sample period (1982:7-1987:6) is used. Only the Canadian rate change remains significant in the U.S. and U.K. equations, and even this variable disappears in the equations for West Germany and Japan. All the other significant foreign influences shown in table 6 drop out; the remaining estimates in the table are virtually unaffected. Thus, although for eign changes in short-term rates sometimes in fluence domestic long rates, this influence is not robust.-” The results in table 2 show that correlations between changes in long rates internationally are pairwise significant in all cases; this result persists even when the sample is split into approximately equal subperiods. While long rates are not linked strongly through currency and m oney markets, there are significant and stable relationships be tween them. Apparently, the integration o f inter national capital markets assures that nominal long-term rates move together, despite the fact that this integration usually is quite direct and does not arise from the short-term considerations in currency and m oney market em phasized by term structure explanations. SUMMARY AND CONCLUSION This article explores international linkages among interest rates. The framework used for this purpose is a conventional m odel o f the domestic term structure. In a term-structure framework, a change in a foreign short-term rate w ould be ex pected to alter the foreign long-term rate and, if interest rates are linked internationally, to alter domestic short-term rates as well. The latter change, again via a domestic term-structure rela tion, w ould change domestic long-term rates. This article tests these relationships. It also tests w hether foreign short-term rate changes exert an independent term structure influence, given the current change in the domestic short-term rate. When foreign interest rates are added to the domestic, short-term interest rate equations, there is some marginal, though segmented, connection between rates across countries. Changes in short term rates in either Canada or the United States affect short-term rates in the other. In addition, changes in the U.K. short-term rate directly in fluence interest rates in Canada. There is a similar bidirectional connection between short-term rates in Japan and West Germany. There is no signi ficant linkage, however, between changes in U.S. short-term interest rates and changes in short term rates in the United Kingdom, Japan or West Germany, over the full period examined here. The evidence suggests that long-term nominal interest rates are related closely and directly across countries. The addition o f changes in for eign short-term rates to the domestic long-term ^The foreign influences in table 5 are somewhat more robust in a similar test. For the latter half of the sample period, the sig nificant influences of Canadian short rates on U.S. short rates, U.K. short rates on Canadian short rates and short rates in Japan on those in Germany remain significant. The bidirec tional elements from the United States to Canada and from Germany to Japan disappear. Thus, no short-run influence is left running from U.S. short rates to those in any of the coun tries. JULY/AUGUST 1988 42 rate equations, however, generally provides no significant information. Also, short-term interest rate changes are not contem poraneously correl ated across countries. Thus, the relationship be tween long-term nominal interest rates does not arise indirectly through an international termstructure transmission or through com m on shortterm-rate movements that are transmitted through the domestic-term structures. Neither of these channels is found to be significant. REFERENCES Bisignano, Joseph. “ Monetary Policy Regimes and Interna tional Term Structures of Interest Rates” , Federal Reserve Bank of San Francisco Economic Review (Fall 1983), pp. 7 26. Clarida, Richard H., and John Y. Campbell. “The Term Struc ture of Euromarket Interest Rates: An Empirical Investiga tion,” Journal o f Monetary Economics (1987, vol. 19, no. 1), pp. 25-44. Cox, John C., Jonathan E. Ingersoll, and Stephen C. Ross. “A Reexamination of Traditional Hypotheses about the Term Structure of Interest Rates,” Journal of Finance (September 1981), pp. 769-99. Cumby, Robert E., and Maurice Obstfeld. "International Inter est Rates and Price Level Linkages under Flexible Exchange Rates: A Review of Recent Evidence,” in John F. O. Bilson and Richard C. Marston, eds., Exchange Rate Theory and Practice (University of Chicago Press, 1984), pp. 121-51. Fisher, Irving. The Theory of Interest (Macmillan Inc., 1930). Hansen, Lars Peter, and Robert J. Hodrick. “ Forward Ex change Rates as Optimal Predictors of Future Spot Rates: An Econometric Analysis,” Journal of Political Economy (October 1980), pp. 829-53. Kirchgassner, Gebhardt, and Jurgen Wolters. “ U.S.-European Interest Rate Linkage, A Time Series Analysis for West Ger many, Switzerland and the United States,” The Review of Economics and Statistics (November 1987), pp. 675-84. Krol, Robert. “The Interdependence of the Term Structure of Eurocurrency Interest Rates,” Journal o f International Money and Finance (June 1986), pp. 245-53. Mankiw, N. Gregory. “The Term Structure of Interest Rates Revisited,” Brookings Papers on Economic Activity (1:1986), pp. 61-96. Mankiw, N. Gregory, and Lawrence H. Summers. “ Do LongTerm Interest Rates Overreact to Short-Term Interest Rates?” Brookings Papers on Economic Activity (1:1984), pp. 223-42. Meese, Richard A., and Kenneth Rogoff. “ Empirical Exchange Rate Models of the Seventies: Do They Fit out of Sample?” Journal of International Economics (February 1983), pp. 3-24. Modigliani, Franco, and Richard C. Sutch. “ Innovations in Interest Rate Policy,” American Economic Review (May 1966), pp. 178-97. Mussa, Michael. "Macroeconomic Interdependence and the Exchange Rate Regime," in Rudiger Dornbusch and Jacob A. Frenkel, eds., International Economic Policy: Theory and Evidence (John Hopkins University Press, 1979), pp. 160204. Phillips, Llad, and John Pippenger. “ Preferred Habitat vs. Efficient Market: A Test of Alternative Hypotheses,” this Review (May 1976), pp. 11-19. Shiller, Robert J. “ The Volatility of Long-Term Interest Rates and Expectations Models of the Term Structure,” Journal of Political Economy (December 1979), pp. 1190-219. Frenkel, Jacob A., and Richard M. Levich. “ Covered Interest Arbitrage: Unexploited Profits?” Journal o f Political Economy (April 1975), pp. 325-38. Shiller, Robert J., John Y. Campbell, and Kermit L. Schoenholtz. "Forward Rates and Future Policy: Interpreting the Term Structure of Interest Rates," Brookings Papers on Economic Activity (1:1983), pp. 173-217. _________“ Transaction Costs and Interest Arbitrage: Tranquil versus Turbulent Periods," Journal of Political Economy (De cember 1977), pp. 1209-26. Swoboda, Alexander K. “ Exchange Rate Regimes and U.S.European Policy Interdependence," International Monetary Fund Staff Papers (March 1983), pp. 75-102. Garner, C. Alan. “The Yield Curve and Inflation Expectations” , Federal Reserve Bank of Kansas City Economic Review (September/October 1987) pp. 3-15. Wood, John H. “ Do Yield Curves Normally Slope Up? The Term Structure of Interest Rates, 1862-1982,” Federal Re serve Bank of Chicago Economic Perspectives (July/August 1983), pp. 17-23. Glick, Reuven. “ Interest Rates Linkages in the Pacific Basin,” Federal Reserve Bank of San Francisco Economic Review (Summer 1987), pp. 31-42. FEDERAL RESERVE BANK OF ST. LOUIS Wood, John H., and Norma L. Wood. Financial Markets (Harcourt Brace Jovanovich, Inc., 1985). 43 Appendix The Interest Rate Data All data are end-of-month values. Canada S hort rate: interest rate on three-month Treasury bill rates from the database o f the Federal Reserve Board. Long rate: interest rate on government bonds with a remaining maturity o f 10 years from the database o f the Federal Reserve Board. Japan S hort rate: one-month Gensaki rate provided by the Bank o f Japan. Long rate: average yield to maturity on a number o f government bonds with a constant remaining maturity o f nine years, provided by the Bank of Japan. United States Short rate: Until April 1984, the yield on onemonth T-bill rates was available. From May 1984 to June 1987, the series was updated using the inter est rate on three-month T-bill rates. A test o f the adequacy o f this approximation was perform ed by regressing the one-month T-bill rate on a constant and the three-month T-bill rate over the period w hen both w ere available, January 1978 to April 1984. The constant is not significantly different from zero, w hile the coefficient on the threemonth rate is not significantly different from one. The other statistics also justified the approxima tion. The one-month data w ere provided by Pro fessor Alex Kane. The three-month data came from the database o f the Federal Reserve Board. Long rate: the series is the yield to maturity of government securities bonds with remaining maturity o f 10 years from the database o f the Federal Reserve Board. United K ingdom Short rate: one-month interbank deposit rate from the Financial Times. Long rate: average yield to maturity on a number of government bonds with remaining maturity between eight and 12 years from the Financial Times. West Germ any Short rate: one-month interbank deposit rate from the Frankfurter Allgemeine. Long rate: average yield to maturity on a number o f government bonds with remaining maturity over eight years from the Frankfurter Allgemeine. JULY/AUGUST 1988 44 Keith M. Carlson Keith M. Carlson is an assistant vice president at the Federal Reserve Bank of St. Louis. Thomas A. Pollmann provided re search assistance. How Much Low er Can the Unemployment Rate Go? I n JUNE 1988, the civilian unemployment rate dipped to 5.3 percent, its lowest rate since May 1974. The Council o f Econom ic Advisers (CEA), in its 1988 Annual Report, forecast a continuing drop in the unem ploym ent rate, accompanied by a decline in the inflation rate from 4.6 percent in 1987 to 3 percent in 1991.' These developm ents raise an interesting ques tion: how low can the unemployment rate be driven without accelerating inflation? In the late 1970s, considerable research was devoted to the discussion o f such a critical rate, usually referred to as the “natural rate o f unem ploym ent.” This research produced estimates o f the natural rate in the late 1970s ranging between 5 percent and 7 percent, but generally w ere centered on 6 per cent.2With the unemployment rate well above 6 percent for most o f the 1980-87 period, the debate about the level o f the natural rate had subsided; with the unemployment rate moving w ell below 6 percent in early 1988, however, the debate has now resurfaced. 'The CEA report was prepared in February. See Council of Economic Advisers (1988), p. 50. For further detail on the Administration’s forecast, see Office of Management and Budget (1988), pp. 3 b -7 -8 . The annual inflation rate is that for consumer prices measured from fourth quarter to fourth quarter. 2A representative estimate is that of Cagan (1979), p. 215. For a more exhaustive survey of alternative estimates, see Weiner (1986). This article reviews the factors that determine the natural rate o f unemployment, focusing speci fically on developm ents since 1979. First, it dis cusses the concept o f unem ploym ent and summa rizes h ow the government measures unem ploy ment. Second, it reviews the choice o f benchmark years as an aid in the analysis. Finally, it examines the underlying determinants o f the natural rate of unemployment in detail. Though no attempt has been made to derive precise estimates of the natu ral rate, the direction o f its movement in recent years has been detailed.3 BACKGROUND: CONCEPTS AND MEASUREMENT To analyze recent unem ploym ent trends, it is useful to summarize the reasons for unem ploy ment. Since the focus is on unemployment as measured by the U.S. government, some detail about how unem ploym ent statistics are gathered is also useful (see opposite page). 3Most of the studies were done in the late 1970s and have not been updated since then. See Weiner (1986). The major excep tions are Rissman (1986) and Gordon (1987), in which he “ assumes” continuation of the natural rate at 6 percent through 1985. He offers statistical evidence in support of this contention in Gordon (1988). JULY/AUGUST 1988 45 A Prim er on U.S. Unemployment Statistics Each month, during the week containing the 12th day of the month, the U.S. Bureau o f the Census suiveys 65,000 households for the Bu reau o f Labor Statistics.1One-fourth o f the households is replaced each month so that no household is interviewed more than four months in a row. This procedure allows for continuity of the data and reduces the report ing burden on families. In response to a series o f questions, the inter viewer determines w hether each household member 16 years and older is em ployed, unem ployed or not in the labor force. Unemployed persons are those w ho had no em ployment during the survey week but w ere available for work. The interviewer also establishes whether each unem ployed person (1) made specific efforts to find work in the preceding four weeks, (2) was waiting to be recalled to a job from which he or she had been laid off or (3) was waiting to report to a new job within 30 days. Further information is obtained about the reason for unemployment. The unem ployed are 'For a useful summary of how the government collects information for its unemployment report, see Bureau of Labor Statistics (1987). Types o f Unemployment Unemployment can be categorized as frictional, cyclical and structural. A lth o u g h th e g o v e rn m e n t does n ot present its statistics in this wav, such a categorization is still helpful in understanding w hy unemployment occur s. Cyclical unemployment can be most readily understood as representing movements o f the unemployment rate that result from fluctuations o f aggregate demand for goods and services. These fluctuations, in turn, can be traced to monetary and fiscal policy or anything else that affects ag gregate demand. Frictional unem ploym ent results from relative shifts in the supply or dem and for goods and ser vices between industries or occupations. Because information about jobs is costly to obtain, people can be “caught between jobs," resulting in terrrpo classified as (1) job losers (involuntary quits), (2) job leaver's (voluntary quits), (31 reentrants (those w ho previously worked, but w ere out of the labor force before looking for work), and (4) new entr ants (those w ho never worked before but w ere now looking for work). Employed persons are (1) persons who, dur ing the survey week, either worked as paid em ployees in their ow n business or as unpaid workers in a family-operated enterprise, and (2) those w ho did not work but had jobs from which they w ere temporarily absent because o f illness, weather, vacation, strikes or personal reasons.2 The labor force is the sum o f persons aged 16 and over w ho are classified as em ployed or unemployed. The unem ploym ent rate is calcu lated as the total number unem ployed as a percent o f the labor force. Unemployment rates according to sex, age, race and reason for un em ployment are also calculated. 2The government also provides information on the distinction between part-time and full-time employment. raiv unem ploym ent w hile information about other jobs is sought. Sometimes, to emphasize its shortrun transitional nature, this type o f unem ploy ment is called turnover unemployment and is consider ed a vital aspect o f the operation o f a freeenterprise economy. Structural unem ploym ent occurs when there is a mismatch o f workers and job vacancies either by reason o f skill or location. It is only artificially dis tinguishable from frictional unemployment in that it is considered longer in duration and involves, in addition to the costs o f job-information search, training or relocation costs. Categorizing unem ploym ent into three types is a useful w ay to analyze it. The three types o f un em ployment involve costs in obtaining informa tion about the availability o f other jobs. Because labor markets are characterized by heterogeneity of skills and job requirements, it takes time and JULY/AUGUST 1988 46 resources to get such information. In general, this process continues until the expected benefit of the search (present value o f expected future in come) equals the cost o f the continued search (again, in present value terms). Defining the Natural Rate o f Unemployment To analyze unem ploym ent further, one must clarify the meaning o f the term "natural rate of unem ploym ent.” This can be done by explaining the concept using a Phillips curve diagram. Phillips-curve analysis was popularized in the 1960s and is still useful today as an expository device.4In essence, the Phillips curve summarizes the relationship between inflation and the unem ployment rate. When first introduced, it was thought to be a relationship that policymakers could exploit. Over the years, however, this inter pretation has changed. Analysts now generally accept that there is a whole family of short-run Phillips curves, corresponding to different ex pected rates o f inflation (Pl in figure 1). Figu re 1 Phillips Curve Diagram Percent note an increase in sales and w ill interpret the increase as a shift in dem and for their product and attempt to expand employment. Attractive wage offers will induce many workers to cut short their job search and accept employment. Higher prices and low er unemployment w ill result, m ov ing the econom y to point B. This movement is temporary, however, as both employers and workers com e to expect inflation. When the shift in dem and is perceived as general, workers w ill return to their normal job search patterns and employers' dem and for labor w ill be reduced to previous levels. The econom y w ill move to point C, where unemployment is equal to its natural rate, Un, once again, but inflation is higher than it was at point A. The vertical line is called a long-run Phillips curve because it reflects a period long enough for inflationary expectations to adjust fully. The revised interpretation o f the Phillips curve yields a definition o f the natural rate o f unem ploy ment: when the actual rate o f inflation equals the expected rate, the unem ploym ent rate that corres ponds is the natural rate (shown as U„ in figure 1). This does not mean that there is anything “natu ral” about such a rate. For example, it is not con stant over time, but rather is influenced by dem o graphic changes as w ell as government policies. The Phillips curve diagram also allows a more precise definition o f cyclical unemployment. W hen unem ploym ent departs from its natural rate, w e have cyclical unemployment. Or, in other words, cyclical unem ploym ent results when ac tual inflation and expected inflation are unequal. The governm ent’s unem ploym ent statistics provide little help in estimating the natural rate. Conceptually, however, the sum o f frictional and structural unem ploym ent is equal to natural un employment. Consequently, any factor that in fluences either frictional or structural unem ploy ment (or both) is relevant to the determination of the natural rate. In the short run, before expectations o f inflation change, there is a trade-off between inflation and unemployment (shown as SR curves in figure II. Suppose the econom y initially is at point A, with Past research suggests that the most important influences on the natural rate are demographic or institutional. Demographic factors involve such characteristics o f the work force as age, sex and racial distribution. One o f the most prominent demographic factors in the post-World War II period occurred when the baby-boom generation expected inflation equal to zero. If monetary and fiscal policy becom e expansionary, employers will came o f age and entered the labor market. In re cent years, this generation has swelled the size o f Percent Unem ploym ent rate “For a survey, see Santomero and Seater (1978). http://fraser.stlouisfed.org/ FEDERAL RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis 47 Chart 1 Inflation and Unemployment Rates P ercent 15.0 10.0 Percent Annual Data 10 Inflation rate* ISCALE 5 .0 0.0 U nem ploym ent rate SCALED -5.0 1945 2 50 55 60 65 70 75 80 85 1 990 •C o n s u m e r prices_________ the prime-age working group (age 25 to 54). Age groups like this require time to develop the skills necessary to match the requirements of job vacan cies. Such adjustments eventually take place, but the process is usually longer than, say, the tem po rary nature of turnover or even cyclical unem ployment. Institutional factors can have an effect on the natural rate, for example, the pattern of shifting demand across industries, the minimum wage, and government policies that influence jobinformation search (personal and em ployer taxes, unemployment benefits). CHOOSING BENCHMARK YEARS To analyze unemployment trends, one must begin with the choice of certain benchmark years that are representative o f full employment.5The 5“ Full employment” is defined here as the employment counter part of the natural rate of unemployment. 6This was probably recognized first by the Council of Economic benchmark years in this study are ones that occur late in business expansions and are free o f the influences of war. It is difficult to identify any years in the 1948-55 period as benchmark years; they are obviously influenced by the economic conditions associated with W orld War II, the Ko rean War and their aftermath. For this reason, 1956 is chosen as the first benchmark year. This year is recognized generally as one of “full em ploym ent” without serious inflation and the other benchmark years chosen — 1973 and 1979 — are also ones that occur late in business expansions and are generally free o f wartime influence.6 Chart 1 summarizes inflation and unem ploy ment, with the benchmark years highlighted. As one can see, the relationship between the two is highly variable, reflecting a host of complex factors. A direct examination of the inflation- Advisers in the early 1960s. Actually both 1955 and 1956 are used; de Leeuw and Holloway (1983) use 1955, while Cagan (1979) uses 1956. JULY/AUGUST 1988 48 Chart 2 Unemployment Rates Percent Percent A nn u al D ata ■ ■ m m b b i m suganaii^! H A M n ib w Civilian unemployment rate ' 1^ r A ii V 1 1 M 1\ 1 M 1 \ 1 \ \ ll \ J / \ N -J / > 1 \ 1 \ 1 r A J \ V. \ A \ 50 55 60 65 unem ploym ent relationship yields little informa tion about h ow low the unemployment rate can go. Chart 2 summarizes the civilian unemployment rate along with that for prime-age males for the 1948-87 period. Civilian unemployment in the three benchmark years ranges from 4.1 percent of the labor force in 1956 to 5.8 percent in 1979. Com pared with 1979, it appears that the econom y reached full em ployment in late 1987 when the unemployment rate fell below 6 percent; com pared with the earlier benchmarks, however, there seems to be room for further em ployment expansion. The structure o f unemployment, especially as it reflects a changing composition o f the labor force, is an additional consideration in assessing the nearness o f actual em ployment to full em ploy ment. Chart 2 also shows unemployment for prime-age males, the group that has the lowest turnover rate in the labor force. The unem ploy FEDERAL RESERVE BANK OF ST. LOUIS * \ /II M ' ! Une mployment rate: \ "W 1945 i A \ *».— ' i i / \ \ h \ / / \ \ prime-age males ...L l J ................. 70 75 80 85 1990 ment rate for this group was 2.9 percent in 1956, 2.6 percent in 1973, and a somewhat higher 3.4 percent in 1979. The 1987 unem ploym ent rate for this group averaged 5 percent, again suggesting there is room for further expansion in employment. A direct examination o f those unemployment measures that are considered most important does not provide a clear-cut conclusion about whether the 1987 levels o f unemployment indicate an econom y approaching full employment. There fore, w e examine the com position o f the labor force in further detail. CHANGES IN DEMOGRAPHIC FACTORS The overall unem ploym ent rate reflects a w eighted average o f many unem ploym ent rates, as table 1 indicates. As this table shows clearly, cer tain groups typically have higher or low er unem- 49 Table 1 Table 2 Unemployment Rate by Age and Sex _______ (selected years) Composition of Labor Force by Age and Sex (selected years) 1956 Total 4.1% 1973 4.9% 1987 1979 5.9% 6.2% Total' 1956 1973 1979 1987 100.0% 100.0% 100.0% 100.0% Male 16-19 20-24 25-54 55-64 65 + 11.0 6.9 2.9 3.5 3.5 13.9 7.3 2.6 2.4 3.0 15.9 8.7 3.4 2.7 3.4 17.8 9.9 5.0 3.7 2.6 Male 16-19 20-24 25-54 55-64 65 + 67.7 3.7 5.2 45.6 9.3 3.9 61.0 5.2 8.0 37.8 7.9 2.1 57.9 4.9 8.1 36.1 6.9 1.9 55.2 3.4 6.5 37.9 5.8 1.6 Female 16-19 20-24 25-54 55-64 65 + 11.0 6.3 4.1 3.7 2.3 15.3 8.4 4.4 2.8 2.9 16.4 9.6 5.2 3.2 3.3 15.9 9.4 5.1 3.1 2.4 Female 16-19 20-24 25-54 55-64 65 + 32.2 2.8 3.7 20.6 3.9 1.2 38.9 4.3 6.3 22.4 4.7 1.2 42.1 4.3 6.9 25.3 4.5 1.1 44.7 3.2 6.0 30.4 4.1 1.0 'May not equal 100 due to rounding. ployment rates than the overall average. The teen age group is always the highest, follow ed hv the 20-24 year-old group. Consequently, to under stand more fully the significance o f a given unem ploym ent rate, one must examine both the relative importance o f each age group in anv given year and the growth rate o f each age group over time. Table 2 summarizes the relative importance o f the different age-sex groups for the benchmark years. One striking observation is the change in the ratio o f males to females that has taken place since 1956. This changing proportion, however, may not be critical in interpreting what has hap pened to the overall unem ploym ent rate: as table 1 shows, female unemployment is not always above that for males.7 What is important in interpreting movements in the unemployment rate over time is the shifting importance o f age groups. Obviously, the rise in importance o f the 16-19 and 20-24 vear-old groups from 1956 to 1979 was an important factor in interpreting unemployment trends for that period. As table 1 shows, the unemployment rate was always highest for these groups. Significantly, however, these youngest age groups have declined as a proportion o f the labor force from 1979 to 1987. Table 3 Growth of Labor Force by Age and Sex (selected years, annual rates) Total 1956-73 1973-79 1979-87 1.8% 2.7% 1.7% Male 16-19 20-24 25-54 55-64 65 + 3.9 4.3 0.6 0.7 -1 .8 1.5 2.9 1.9 0.4 0.3 - 2 .7 -1 .1 2.3 -0 .5 - 0 .3 Female 16-19 20-24 25-54 55-64 65 + 4.3 5.0 2.3 2.9 1.5 2.9 4.2 4.8 1.8 1.5 - 1 .9 - 0 .2 4.0 0.6 0.6 For the 1979-87 period, the 25-54 year-old group grew fastest, reflecting the maturation o f the 16-19 and 20-24 year-old groups o f the 1970s. Table 3 summarizes the growth o f the labor force by age group between the benchmark years. To aid in analyzing the effects o f changes in the underlying demographics, a w eighted unem ploy ment rate, w here the weights are based on the com position o f the labor force, is com m only used.8 7For an analysis of women in the labor market, see Shank (1988). 8See Clark (1977), Flaim (1979), Cain (1979), Antos, Mellow and Triplett (1979) and Cagan (1979). JULY/AUGUST 1988 50 Table 4 Alternative Unemployment Rates Overall Rate Fixed Weight1 Prime-Age Male Overall Less Fixed-Weight 1956 4.1% 4.3% 2.9% -0 .2 % 1.2 1973 4.9 4.4 2.6 0.5 2.3 1979 1980 1981 1982 1983 1984 1985 1986 1987 5.9 7.2 7.6 9.7 9.6 7.5 7.2 7.0 6.2 5.3 6.6 7.1 9.2 9.2 7.2 7.0 6.8 6.1 3.4 5.2 5.5 8.0 8.2 5.9 5.6 5.6 5.0 0.6 0.6 0.5 0.5 0.4 0.3 0.2 0.2 0.1 2.5 2.0 2.1 1.7 1.4 1.6 1.6 1.4 1.2 Overall Less Prime-Age Male ’Calculated using average of labor force composition in 1956 and 1987 (see table 2). Table 4 summarizes various unem ploym ent rates and provides information about h ow the changing com position o f the labor force influences the over all unemployment rate. A comparison o f the alter native rates with the overall rate shows that dem o graphic shifts w ere most pronounced in the 1956-79 period.9Changes in the com position of the labor force shifted the unem ploym ent rate upward by 0.8 percentage points (overall less fixed-weight column) compared with an actual rise o f 1.8 percentage points (5.9 minus 4.1). In other words, the labor market pressure o f 4.1 per cent in 1956 w ould have changed to 4.9 percent in 1979 because o f a shift in the com position o f the labor force toward the youngest groups. There was also a w idening o f the difference between the overall unem ploym ent rate and that for prime-age males, reaching 2.5 percentage points in 1979, up from 1.2 percentage points in 1956. This differential yields the same general con clusion: considering demographic changes, the natural rate o f unem ploym ent rose quite sharply between 1956 and 1979. Since 1979, the com position o f the labor force has shifted back toward the older groups, which suggests that the natural rate o f unemployment has declined. The difference between the fixedweight measure and the overall rate has narrowed to almost zero. The smaller differential means that dem ographic considerations no longer loom as 9For a similar table and analysis, see Council of Economic Advisers (1978), p. 170. http://fraser.stlouisfed.org/ FEDERAL RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis large in determining if the econom y is at full em ployment. For dem ographic reasons alone, the 1987 natural rate o f unem ploym ent is only about 0.3 percentage points higher than that o f 1956, and 0.5 percentage points low er than that o f 1979. This interpretation is supported by the change in the differential between the overall rate and that for prime-age males. Thus, although methods vary in calculating the effects o f changing demographics on the unemployment rate, there seems to be little doubt that changes in the com position o f the la bor force since 1979 have produced a low er natu ral rate o f unem ploym ent.10 CHANGES IN INSTITUTIONAL FACTORS As noted earlier, in addition to the age-sex com position o f the labor force, many other factors influence the unem ploym ent rate. These factors are discussed in this section along with a sum mary of their recent trends. Minimum Wage Federal minimum wage legislation was intro duced in the United States in 1938; by 1985, 80 percent o f the U.S. labor force was em ployed in sectors subject to its coverage. By paying the lowest-incom e workers a higher wage than the competitive market w ou ld pay in the absence o f a '“See Cain (1979). 51 Chart 3 Minimum Wage as a Percent of Average Hourly Earnings P erce n t 60 30 1945 P ercent 60 A nnual Data 50 55 60 65 70 75 NOTE: Average hourly earnings are for the private nonagricultural sector. minimum wage, this legislation raised the average level o f real wages above their competitive level." In re s p o n s e , th e q u a n tity o f la b o r services s u p p lie d w ill e x c e e d th e q u a n tity d e m a n d e d , w ith th e d iffe re n c e b e in g c la s s ifie d as u n e m p lo y e d . When the minimum wage was first legislated in 1938, it was $.25 per hour and covered about 40 percent o f the nation’s nonsupervisorv employees. Over the 1938-81 period, it was raised 15 times, reaching $3.35 per hour in 1981 and has not been changed since. By 1985, 87 percent o f nonsupervisoiy employees were subject to the minimum wage The minimum wage law has had its greatest effect on teenage em ployment with little effect on other age groups. Because they have fewer skills and less education, teenagers’ marginal products are typically below those o f older, more experi enced workers. Consequently the minimum wage is much more likely to be above the competitive w a g e for this g ro u p . To assess the impact o f minimum wage legisla tion, the minimum wage must be viewed relative to average hourly earnings. The comparison measure used here is average hourly earnings for workers in the private nonagricultural sector. The movement o f the minimum wage relative to this measure from 1947 to 1987 is shown in chart 3. After a large jump in 1950, the minimum wage relative to average hourly earnings fluctuated be tween 45 percent and 55 percent, before dropping below 45 percent in 1972. It then rose from about 40 percent in 1973 to 46 percent in 1981. With the minimum wage constant at $3.35 per hour since 1981, however, a steady decline in the relative 11For a survey, see Brown, Gilroy and Kohen (1982). JULY/AUGUST 1988 52 Chart 4 Replacement Ratio and Unemployment Rate Annual Data P ercent 10 1945 NOTE: Replacement ratio is the average of unemployment benefits paid weekly as a percent of average weekly earnings in the private nonagricultural sector. minimum wage has occurred since then, reaching 37 percent in 1987. Although the relative minimum wage declined from 1957 to 1973, the coverage increased from 45 percent to 75 percent, primarily because o f the rapid growth o f teenagers in the labor force. The minimum wage may have pushed the unem ploy ment rate upward from 1973 to 1979, but this trend was sharply reversed from 1979 to 1987. Since the last benchmark year o f 1979, the m ini mum wage movements have had a positive effect on the labor market; its decline has reduced the natural rate o f unemployment. other things equal, individuals w ill search longer for a job, lowering the amount o f work that they are willing to supply at a given real wage. Also, individuals w ho are not in the labor force w ill be inclined to enter it to obtain a job and be eligible for unem ploym ent benefits in the future. Unemployment Benefits One important measure in assessing the effect o f unemployment benefits is the ratio o f average unemployment benefits paid weekly relative to average weekly earnings. T his ratio is called the replacement ratio. Chart 4 shows the ratio from 1947 to 1987 as a solid line. Because this ratio shows cyclical movem ent throughout the period, the unem ploym ent rate is also charted (dashed line i. An increase in unemployment benefits relative to wages lowers the cost o f job search.'- As a result, Generally, the replacement ratio and the unem ployment rate move in tandem. From 1965 to 1973, l2For discussion and references, see Parkin (1984) and Cagan (1979). http://fraser.stlouisfed.org/ FEDERAL RESERVE Federal Reserve Bank of St. Louis BANK OF ST. LOUIS 53 however, the replacement ratio rose quite sharply relative to unemployment. From 1973 to 1979, it then declined slightly relative to the unem ploy ment rate. But since 1984, the divergence between these measures has been sharp. social insurance. Introducing (or raising) this tax reduces the quantity o f labor dem anded for a given real wage and, because o f the higher cost o f labor, may also lengthen the amount o f time em ployers take in searching for workers. A closer examination reveals that the source of this recent divergence is chiefly a slowing of aver age weekly earnings while unem ploym ent benefits have continued to rise at relatively rapid rates. Despite this development, the replacement ratio seems to have had a recent upward effect on the natural rate o f unemployment. This effect is dam pened somewhat by two considerations: (1) recent changes in tax law whereby unemployment benefits became partially subject to taxation in 1979, and com pletely so in 1987, and (2) a general tightening o f eligibility requirements in recent years.13Thus, the actual value o f the replacement ratio in 1987 relative to 1979 is less than shown in the chart. It is impossible, however, to say how much the change is without further research. Thus, increased tax rates, whether applicable to em ployers or employees, reduce em ploym ent and may increase unemployment. To show what has happened to the tax wedge, em ployee and em ployer taxes are com bined into a summary mea sure and plotted against the unem ploym ent rate in chart 5. This tax w edge measure incorporates personal incom e taxes (federal, state and local), em ployer and em ployee contributions for social insurance, and sales and excise taxes.15Using 1956 as a reference point, the tax w edge has increased from about 21 percent to more than 32 percent by 1987. The rise was relatively rapid from 1956 to 1973, slightly slower from 1973 to 1979 and even slower from 1979 to 1987. These trends suggest that taxes contributed to an increase in unem ploy ment before 1979; since then, the tax w edge has had little effect, except perhaps to reduce unem ploym ent somewhat since 1981. Taxes Another factor o f considerable im portance in determining unem ploym ent trends is the role o f taxes in influencing the labor markets.14Again, the analysis is com plex and the conclusions are not clear-cut. As an aid in understanding the m acroec onom ic effects, it is useful to think in terms o f the effects on labor supply and dem and separately. Focusing first on labor supply, the tax w edge is the difference between the real wage that the em ployer is willing to pay and the after-tax value of that wage to the workers; the size o f this w edge is important in the work-vs .-leisure decision that people make. An increase in the tax w edge will reduce labor services offered at a given real wage and may encourage a longer job-search by reduc ing the cost o f being unemployed. On the dem and side o f the labor market, the relevant tax is the em ployer’s contribution for 13See Abraham (1988) and, for a state-by-state summary of unemployment legislation in 1987, see Runner (1988). '“Meyer (1981). 15The tax wedge for households (or suppliers of labor services) is W _ W (1 - t „ ) P (1 + tc) ’ P where W is the nominal wage, P is the price level, tp is the personal tax rate and tc is the consumption tax rate. This ex pression can be manipulated to give Demand Shift Recent research has suggested that shifts in industry dem and also have an effect on the natu ral rate.16This effect is com m only called frictional unemployment. I f changing tastes, technology or relative factor prices induce rapid shifts in indus try demands for labor, there w ill be greater uncer tainty in labor markets and increased search time for both the em ployee and the employer. Unemployment that occurs for these reasons is a healthy reflection o f a dynamic econom y. For our discussion, however, only the long-run m ove ments in the com position o f industrial output are relevant. Chart 6 is one attempt to capture this phenom enon; it shows the three-year moving average o f the sum o f the absolute percentage w (tc+g P (1+g Chart 4 shows the value of the expression following W/P plus the employer’s contribution rate for social insurance. For further discussion, see Parkin (1984), pp. 184-85. 16For a discussion and critique of this literature, see Johnson and Layard (1986). See also Lilien (1982), Lilien and Hall (1986), and Rissman (1986). JULY/AUGUST 1988 54 Chart 5 Estimate of Tax Wedge and Unemployment Rate change in sectoral em ploym ent shares.'7By this measure, there was a downw ard trend in the d e gree o f shifting em ployment until the mid-1960s; since then the measure o f shifting em ployment has m oved upward, although it has varied sub stantially around the trend. Focusing on the benchmark years, there is a slight downward movem ent from 1956 to 1973 to 1979, follow ed by an upward movement from 1979 to 1987. From 1982 to 1987, however, the measure dropped sharply.18Viewed in this perspective, it is 17The measure of demand shift (in year t) is 10 2 i= 1 E „_ E, E, , where Eit is employment in the ith industry in year t and E, is total employment in year t. Data used were employees on nonagricultural payrolls by major industry. See Council of Economic Advisers (1988), pp. 296-97. ,80ne factor operating during this period was the sharp swing in the value of the dollar, rising sharply from 1980 to 1985, and then falling sharply to 1987. If these developments affected http://fraser.stlouisfed.org/ FEDERAL RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis unlikely that shifts in the structure o f the econom y have influenced the natural rate o f unem ploym ent substantially. The relationship between “ dem and” shifts and the unem ploym ent rate appears, rather, to be a shorter-run phenom enon. Other Factors The above list o f factors, w hile not exhaustive, summarizes most o f the factors that influence unem ploym ent trends. Government regulations, however, also affect labor markets. For example, mainly manufacturing exports, the effect could have been to raise frictional unemployment (and the natural rate) when the dollar was rising, and lower such unemployment when it was falling. But since the focus here is on 1987 vs. 1979, a period over which the trade-weighted exchange rate rose only 10 percent, there would seem to be little net effect on the natural rate. Furthermore, manufacturing employment as a percent of total non-agricultural employment has shown little sensitivity to exchange rate movements, even of the magnitude experienced in the 1980s. 55 Chart 6 Measure of Demand Shift and Unemployment Rate regulations im posed by the Occupational Safety and Health Administration (OSHA) in the interest o f safety and health can divert funds that w ould normally be used for investment spending.1” These regulations can act like an em ployer tax, driving a w edge between the wage the em ployer is willing to pay and the actual cost. Another example o f regulations that had an important effect on unem ploym ent are work regis tration requirements for various government p ro grams like welfare and food stamps. For example, in 1972, legislation was passed that required w el fare mothers w ho w ere able to work to register for work.20Although some found jobs, others were added to the count o f the unemployed. SUMMARY OF FACTORS AFFECTING NATURAL RATE direction in w hich the structural factors have been operating between the benchmark years since 1956. No attempt is made to estimate precisely the magnitude o f the effects on the natural rate of unemployment. The most obvious change in recent years is the shifting age distribution o f the labor force, which has reduced the unem ploym ent rate. In other words, the babv-boomers, w ho made their pres ence felt throughout the 1970s by pushing up the natural rate o f unemployment, are now in the prime-age group. Having accumulated skills, edu cation and experience, this group is now market ing its productive skills, thus reducing the natural rate o f unem ploym ent by about one-half o f a per centage point from 1979. The role o f these factors is brought together in table 5. Shown are general conclusions about the The minimum wage has had a favorable effect in reducing the trend o f unem ploym ent since 1981, but no attempt was made here to estimate the magnitude o f effect. Cagan, however, estimated 19For a broad survey, see Licht (1988). “ Clarkson and Meiners (1977). JULY/AUGUST 1988 56 Table 5 Summary of Effects on the Natural Rate of Unemployment Factor 1956-73 1973-79 1979-87 World W ar II period. In general, the key factors that influence the natural rate o f unemployment have served to reduce it in the 1980s. As a result, the current natural rate appears below the 6 per cent rate estimated in 1979“ Shifts in the age structure o f the labor force alone have reduced it about one-half o f a percentage point. Other favor able developments, as noted in table 5, may have reduced it even further. Demographic change t f? 1 Minimum wage T? T 1 Unemployment benefits t 1? t? SUMMARY Tax wedge t T 0 Demand shifts i? 4? t? Unemployment rates below 6 percent in late 1987 and early 1988 have raised questions about how far the rate can fall before inflation again emerges. The fact that inflation has shown no clear signs o f acceleration suggests that structural changes in the U.S. econom y have reduced the natural rate o f unem ploym ent below what it was in 1979. This article examined some o f these struc tural factors. NOTE: Arrow indicates direction of effect on the natural rate of unemployment. Question mark with arrow indi cates the effect is probably small. A zero indicates no effect on the natural rate. that the minimum wage contributed to an in crease in the natural rate o f .45 percentage points from 1956 to 1977.21The relative minimum wage in 1987 was below that in 1956. With the decline in the proportion o f teenagers in the labor force, however, the magnitude o f the effect on the natu ral rate o f unem ploym ent is probably less than Cagan estimated. Unemployment benefits generally appear to have affected recent unem ploym ent trends nega tively. The replacement ratio has risen quite dra matically since 1984. This is misleading, however, because starting in 1987, unemployment benefits became fully taxable by the federal government, w hile eligibility requirements have been tightened in recent years. These developments have raised the cost o f being unem ployed and have reduced the trend o f unemployment. Taxes w ere a factor in the 1956-73 period (and to some extent from 1973 to 1979), increasing un employment, both by reducing the cost o f the job search (reducing foregone earnings) and increas ing the tax w edge between what employers pay to labor and workers receive. Since 1979, however, the upward trend o f taxes has slowed, suggesting that the tax w edge has not worsened. These devel opments are assessed as having no effect on the natural rate in the 1979-87 period. Despite considerable fluctuation in the shares o f sector employment, demand-shift factors do not appear to have been a factor during the post21Cagan (1979). http://fraser.stlouisfed.org/ FEDERAL RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis Several o f these factors w ere found to have re duced the natural rate o f unem ploym ent in recent years, when com pared with previous experience from 1956 to 1979. The age com position o f the labor force, the minimum wage, individual and em ployer tax rates are a few o f the factors that have m oved favorably. Any conclusions about un em ployment benefit ratios, however, require fur ther study. For the unem ploym ent rate to con tinue to decline depends critically on the course o f future government actions, namely, legislation relating to the minimum wage, tax rates and un em ployment benefits. REFERENCES Abraham, Katharine G. “ Has the Natural Rate of Unemploy ment Fallen?” mimeo (University of Maryland, June 1988). Antos, Joseph, Wesley Mellow, and Jack E. Triplett. “ What is a Current Equivalent to Unemployment Rates of the Past?” Monthly Labor Review (March 1979), pp. 36-46. Brown, Charles, Curtis Gilroy and Andrew Kohen. “The Effect of the Minimum Wage on Employment and Unemployment,” Journal of Economic Literature (June 1982), pp. 487-528. Bureau of Labor Statistics. How the Government Measures Unemployment, Report 742 (U.S. Government Printing Office, September 1987). Cagan, Phillip. Persistent Inflation: Historical and Policy Essays (Columbia University Press, 1979). Cain, Glen G. "The Unemployment Rate as an Economic Indicator,” Monthly Labor Review (March 1979), pp. 24-35. 22Weiner (1986) and Cagan (1979). 57 “A New Estimate of Potential GNP,” The 1977 Economic Report of the President, Hearings before the Joint Clark, Peter K. Economic Committee, Congress of the United States, Part 1 (GPO, 1977), pp. 39-55. Inflated Unem ployment Statistics: The Effects of Welfare Work Registration Requirements (University of Miami School of Law, March Clarkson, Kenneth W., and Roger E. Meiners. 1977). Council of Economic Advisers. dent (GPO, 1978). Economic Report of the Presi ________ . Economic Report of the President (GPO, 1988). de Leeuw, Frank, and Thomas M. Holloway. “ Cyclical Adjust ment of the Federal Budget and Federal Debt," Survey of Current Business (December 1983), pp. 25-40. Flaim, Paul O. “The Effect of Demographic Changes on the Nation's Unemployment Rate,” Monthly Labor Review (March 1979), pp. 13-23. Gordon, Robert J. Co., 1987). ________ . Macroeconomics, 4th ed. (Little, Brown and “ The Role of Wages in the Inflation Process,” American Economic Review, Papers and Proceedings (May 1988), pp. 276-83. Johnson, G. E., and P. R. G. Layard. “The Natural Rate of Unemployment: Explanation and Policy," in Orley Ashenfelter and Richard Layard, eds., Handbook of Labor Economics, vol. II (North-Holland, 1986), pp. 921-99. Licht, Walter. “ How the Workplace has Changed in 75 Years,” Monthly Labor Review (February 1988), pp. 19-25. Lilien, David M. “ Sectoral Shifts and Cyclical Unemployment,” Journal of Political Economy (August 1982), pp. 777-93. Lilien, David M., and Robert E. Hall. “ Cyclical Fluctuations in the Labor Market,” in Orley Ashenfelter and Richard Layard, eds., Handbook of Labor Economics, vol. II (North-Holland, 1986), pp. 1001-35. Meyer, Laurence H., ed. The Supply-Side Effects of Economic Policy, Proceedings of a Conference cosponsored by Wash ington University in St. Louis and Federal Reserve Bank of St. Louis, October 24-25, 1980 (May 1981). Office of Management and Budget. Budget of the United States Government: Fiscal Year 1989 (GPO 1988). Parkin, Michael. Macroeconomics (Prentice-Hall, 1984). Rissman, Ellen R. “ What Is the Natural Rate of Unemploy ment?” Federal Reserve Bank of Chicago Economic Perspec tives (September/October 1986), pp. 3-17. Runner, Diana. “ Changes in Unemployment Insurance Legis lation During 1987,” Monthly Labor Review (March 1988), pp. 9-16. Santomero, Anthony M., and John J. Seater. “ The InflationUnemployment Trade-Off: A Critique of the Literature," Jour nal of Economic Literature (June 1978), pp. 499-544. Shank, Susan E. “ Women and the Labor Market: The Link Grows Stronger,” Monthly Labor Review (March 1988), pp. 3-8. Weiner, Stuart E. “ The Natural Rate of Unemployment: Con cepts and Issues," Federal Reserve Bank of Kansas City Economic Review (January 1986), pp. 11-24. JULY/AUGUST 1988 58 H. Alton Gilbert ft. Alton Gilbert is an assistant vice president at the Federal Re serve Bank of St. Louis. Dawn M. Peterson provided research assistance. A Comparison of Proposals to Restructure the U.S. Financial System G L _7lN C E the 1930s, commercial banks have been for banking organizations to underwrite securities. permitted to offer only a limited range o f financial services. At the same time, firms engaged in nonfinancial activities, as well as some in financial industries, have not been permitted to own banks. Such restrictions w ere intended to limit the risk of bank failure, to avoid conflicts o f interest and to prevent undue concentration o f financial power.1 In recent years, however, the separation between banking and other activities has been relaxed somewhat; w hat’s more, Congress is considering further relaxation, including expanding the powers 'These restrictions have not been applied to the ownership of banks by individuals. Individuals who own bank stock may own and operate firms in any other industry. Under the Change in Bank Control Act of 1978, individuals and groups of individuals acting in concert must apply to the appropriate federal supervi sory agency for permission to acquire the stock of a bank over certain percentages of ownership. See Spong (1985), pp. 9 4 95. The bank supervisory agencies may deny permission to purchase bank stock under the following conditions: (1) The purchase would create a monopoly in any part of the banking industry, (2) The financial condition of the acquiring party could ad versely affect the bank, or (3) The competence, experience or integrity of the proposed ownership would not be in the interest of the bank’s deposi tors. http://fraser.stlouisfed.org/ FEDERAL RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis One major reason for permitting the comm on ownership o f banks and firms in other industries is based on concern about the role o f banks in financial intermediation in the future. Some bank customers have found cheaper sources o f credit and other financial services outside the banking industry. Consequently, some analysts say, restric tions must be relaxed if banks are to survive.2The purpose o f this paper is to describe several major proposals for changing banking restrictions and to 2Corrigan (1987), Federal Deposit Insurance Corporation (1987) and Huertas (1986, 1987). 59 examine the concepts that underlie these pro posals. Table 1 CURRENT RESTRICTIONS ON BANKING ACTIVITY At present, the activities o f federally insured commercial banks are lim ited essentially to ac cepting deposits, holding relatively low-risk secu rities and making loans. Banking organizations may acquire firms engaged in financial activities through bank holding companies (BHCs) — corpo rations that ow n one or more banks. In the Bank Holding Company Act (BHCA), Congress autho rized the Federal Reserve Board to determine what activities are permissible for BHCs; these activities, according to the act, should be “ so closely related to banking as to be a proper incident thereto.” Banks generally can engage in most activities that BHCs are allowed to pursue.3A major distinction between banks and the nonbank subsidiaries o f BHCs involves opportunities for geographic expan sion. The nonbank subsidiaries o f BHCs may have offices throughout the nation, whereas nationwide branch banking is not permitted. BHCs are subject to the supervision o f the Fed eral Reserve, w hich periodically inspects them to determine whether they are operating in a sound manner and in compliance with regulations, in cluding the capital requirements set by the Fed eral Reserve.4 On several occasions, the Federal Reserve Board has ruled that BHCs could not un dertake certain activities because they w ere not closely related to banking, might result in conflicts o f interest or might have subjected the BHCs to greater risk.3 3Spong (1985), pp. 95-98. The major exception to this involves the nonbank banks. The BHCA, which gave the Federal Re serve jurisdiction over the acquisitions of banks by corpora tions, defined a bank as one that accepts demand deposits and makes commercial loans. Acquisitions of institutions that did not accept demand deposits or make commercial loans were not subject to the jurisdiction of the Federal Reserve in its capacity as regulator of BHCs. These limited-service banks are commonly called nonbank banks. The Competitive Equality Banking Act of 1987 (CEBA) closes that loophole in the law. It places restrictions on the growth and activities of nonbank banks acquired on or before March 5,1987, and requires firms that acquired nonbank banks after that date to sell them or restrict their activities to those permissible for BHCs. The following restrictions apply to nonbank banks acquired on or before March 5, 1987: Restrictions on Credit Relationships Between Commercial Banks and Their Nonbank Affiliates_________________ Restrictions in section 23A of the Federal Reserve Act: 1. 2. 3. 4. 5. Loans by banks to nonbank affiliates must be fully and adequately collateralized. Total credit to any one nonbank affiliate is limited to 10 percent of the bank's capital. Combined credit to all nonbank affiliates is limited to 20 percent of the bank’s capital. Purchases by banks of unsound assets from nonbank affiliates are forbidden. Bank transactions with affiliates (including transactions covered by the statute and transactions specifically exempt) are to be on terms and conditions that are consistent with safe and sound banking practices. Restrictions in section 23B of the Federal Reserve Act: 1. 2. 3. 4. A bank's transactions with affiliates must be on terms and under circumstances, including credit standards, similar to those offered to nonaffiliate companies. A bank, acting as a fiduciary, shall not purchase securities issued by an affiliate, unless such purchases are specified in the fiduciary agreement. A bank shall not purchase securities being underwritten by a securities affiliate. A bank shall not state or suggest that it is responsible for the obligations of its affiliates. NOTE: Legislation in 1982 removed most of the restrictions on transactions between commercial banks that are subsidiaries of the same corporation. If a corporation owns 80 percent or more of the shares of its subsidiary banks, the only restriction on transactions between the subsidiary banks is that one bank may not sell low quality assets from another bank in the same organization. See Rose and Talley (1982). (3) Beginning in August 1988, their assets may not rise by more than 7 percent in any 12-month period. CEBA also imposes restrictions on the daylight overdrafts of nonbank banks. 4Gilbert, Stone and Trebing (1985). 5Volcker (1986), pp. 436-38. The following are some of the activities not permissible for BHCs and the dates of denials for those activities by the Federal Resen/e Board: underwriting general life insurance (1971), real estate brokerage (1972), land investment and development (1972), operating a savings and loan association (1974), operating a travel agency (1976) and acting as a specialist in foreign exchange options on a security exchange (1986). (1) They may not engage in new activities, (2) They may not market the goods or services of affiliates or have their banking services marketed through nonbank affiliates, except through those marketing arrangements in effect before March 5, 1987, and JULY/AUGUST 1988 60 Table 2 Proposals to Restructure the Financial System Features Corporate structure required of firms that own banks Association of Bank Holding Companies (LaWare (1987)) Association of Reserve City Bankers (1987) Robert Heller (1987) Federal Deposit Insurance Corporation (1987) FSHCs would own BHCs and holding companies that own firms engaged in financial activities in addition to banking. FSHCs would direct ly own banks and firms in other industries. BHCs could acquire banks and firms engaged in financial activities. Non financial firms could acquire BHCs. Firms in any industry could buy banks, and banks could engage in nonbanking activities through their own subsidiaries. No Yes Yes Yes Restrictions on transactions between banks and their affiliates Keep current restrictions Eliminate section 23B of the Federal Reserve Act (see table 1). Keep current restrictions Impose uniform restrictions on dividends and lending limits of banks. Make these restrictions and those in sections 23A and 23B of the Federal Reserve Act apply to transactions between banks and their subsidiaries. Supervisory authority of regulatory agencies Supervision of banks and BHCs unchanged. No one agency supervises FSHCs, which may own BHCs and holding companies that own firms in financial industries other than banking. Subsidiaries of FSHCs in nonbanking industries subject to supervision by their regulatory authorities. Same as for the Association of Bank Holding Companies. No comment on the supervisory powers of the Federal Reserve over BHCs. Nonbank subsidiaries of BHCs subject to supervision by their own government authorities. Firms that buy banks not subject to supervision by bank supervisors. Banks required to report all transactions with affiliates or subsidiaries to bank supervisors, which could audit the terms of the transactions. Obligation to support bank subsidiaries None None BHCs must absorb losses of bank subsidiaries. Nonbanking firms must absorb losses of their BHCs. None Restrictions on assets of banks Current restrictions Current restrictions Current restrictions Current restrictions Ownership of banks by nonfinancial firms permitted http://fraser.stlouisfed.org/ FEDERAL RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis 61 Gerald Corrigan (1987) Robert Litan (1987) Firms that engage in financial activities exclusively could purchase banks. Firms engaged in any activities could buy banks, subject to restrictions on the assets held by those banks. No Yes Keep current restrictions Prohibit banks owned by nonbanking organizations from lending to affiliates. Firms that own banks subject to supervision by the federal bank supervisors, including exercise of powers to limit risks (such as capital require ments) and aggregate concentra tion in the financial system. Nonbank firms that own banks not subject to bank supervisors except to verify that those banks held only the designated safe assets. No formal obligation, but general commitment to be a source of strength for bank subsidiaries. None Current restrictions Bank subsidiaries of nonbanking firms may hold only designated low-risk, liquid assets. Some banks offer financial services through their ow n subsidiaries. The Com ptroller o f the Currency determines w hich activities are permis sible for subsidiaries o f national banks; these are generally restricted to activities that are permissi ble for national banks themselves. In recent years, state governments have allowed subsidiaries o f state-chartered banks to engage in a variety o f new activities; among these are insurance, real estate investment and securities underwriting.6 All federally insured commercial banks are sub ject to restrictions on transactions with their affili ates; these restrictions are shown in table 1. Thus, for example, total loans to affiliates are limited to 20 percent o f the bank's capital. Additional restric tions apply to sales o f assets to banks and pur chases by banks o f securities issued by nonbank affiliates or underwritten by securities affiliates, as w ell as restrictions on loans by banks to their of ficers, directors and m ajor stockholders.7 PROPOSALS FOR RESTRUCTURING THE U.S. BANKING SYSTEM This section describes six proposals for restruc turing the U.S. banking system. Although others could be included, particularly those dealing with the entry o f banks into specific industries, the following proposals encompass the range o f op tions being considered in current policy debates. The key features o f these six proposals are sum m arized in table 2. Each proposal w ould permit banking organizations to engage in a broader range o f activities than currently allowed. Essen tially, the proposals allow nonbanking services to be offered through corporate entities (affiliates or subsidiaries) distinct from the banks themselves. There are two primary differences among the proposals. First, they differ on w hether to permit nonfinancial firms to acquire banks or BHCs. These differences reflect conflicting views on the 6Federal Deposit Insurance Corporation (1987), p. 106. This paper focuses on the issues involved in the common owner ship of commercial banks and firms in other industries. Non banking firms may offer a wide range of banking services by acquiring savings and loan associations (S&Ls). Corporations in any industry other than securities underwriting may acquire one S&L each. Regulations prohibit lending by S&Ls to their nonfinancial parent organizations and restrict other types of transactions that could benefit the parent organization at the expense of the S&L subsidiary. See Federal Home Loan Bank Board (1986). 7Spong (1985), pp. 55-58. JULY/AUGUST 1988 62 policies necessaiy to avoid conflicts o f interest, decreased or unfair com petition among firms of fering financial services and undue concentration o f econom ic resources. These issues have been discussed extensively elsewhere; they are not ana lyzed in this article." Second, the proposals differ on the policies necessary to limit the risk assumed by banks. Note that the proposals have some comm on features designed to limit banking risk. Each proposal in table 2 requires banking organizations to offer nonbanking services through subsidiaries or affili ates; moreover, each includes restrictions on banks lending to their nonbank subsidiaries or affiliates. These proposals rely in part on the legal concept o f “corporate separateness,” under which the creditors o f a corporation have no legal claim on the assets o f a stockholder, even if that stock holder is another corporation. Thus, creditors of the nonbanking units o f a firm that also owns banks w ould have no claim on its banks’ assets.3 Several proposals include special features to limit the risk o f bank failure that might result from affiliation o f banks and nonbanking firms. The Heller proposal (Heller (1987)) requires BHCs to absorb all losses incurred by their bank subsidi aries; nonfinancial firms that acquire BHCs w ould absorb all losses incurred by their BHCs. The FDIC proposal (Federal Deposit Insurance Corporation (1987)) requires bank supervisors to audit transac tions between banks and their nonbank affiliates or subsidiaries to determine whether they are “Rose (1985). 9Black, Miller and Posner (1978). ’“Similar proposals have been made by Kareken (1986), Gilbert (1987), Tobin (1987) and Forrestal (1987). Tobin proposes limiting the assets of all banks to short-term, low-risk assets. "T he factors that determine the expected value and variance of profits of a firm that buys a bank and a nonbanking firm can be expressed in the following equations: E(B + N) = E(B) + E(N), V(B + N) = V(B) + V(N) + 2C0V(B,N), where E refers to expected value, V to variance, B to the profits of the bank, N to the profits of the nonbanking firm and COV to the covariance of the profits of the bank and the nonbanking firm. Holding constant the covariance of the two profit streams, a higher variance in the profits of the nonbanking firm means a higher variance in the profits of the combined firms. The vari ance of the combined profit streams depends on the covari ance of the two profit streams. Finally, as the size of the non banking firm rises relative to the size of the bank, the variance of the combined profit streams converges to the variance of the profits of the nonbanking firm. An analysis of the proposals to restructure the financial system involves an analysis of the mean and variance of the http://fraser.stlouisfed.org/ FEDERAL RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis detrimental to the banks. The Corrigan proposal (Corrigan (1987)) relies on direct supervision o f the firms that buy banks to limit the risk they assume. Finally, the Litan proposal (Litan (1987)) requires banks purchased by nonbanking firms to hold only low-risk liquid assets.10 A FRAMEWORK FOR ANALYZING THE RISK OF BANK FAILURE The proposals for changing bank regulations are concerned with their likely effect on bank failures. This section illustrates h ow the probability o f bank failure is affected when banks and nonbanking firms combine. Key Factors Affecting the Profits and Risks o f Combining Ranks and Nonbanking Firms 11 If a bank offers nonbanking services, the effect on both the expected rate o f return and the varia bility o f returns to the bank’s shareholders, as w ell as the risk o f failure for the bank, depend on five factors. Suppose a bank merges w ith a nonbanking firm. One important factor is the average level o f expected profits or rate o f return for the nonbank ing service. A second factor is the “risk” associated with the prospective nonbanking service; risk is often measured by the standard deviation o f the profits or rates o f return. A third factor is the cor relation between the profit rates o f the bank and returns to shareholders of a firm that buys a bank and a non banking firm and operates them under the conditions of the various proposals. One approach to this analysis might involve expressing the mean and variance of the profits of the firm that buys the bank and the nonbanking firm in terms of the mean and variance of the profits of the bank and the nonbanking firm separately, as indicated in the equations above. The problem with this approach is that the distribution of returns to share holders is not the same as the distribution of profits. In some outcomes, losses exceed the investment of the shareholders; losses to shareholders, however, are no larger than their in vestment in the firm. The distinction between the distribution of profits and the distribution of returns to shareholders is espe cially important for this study, since the various proposals involve different rules for truncating the losses to shareholders. Analysis of the mean and variance of returns to shareholders must be based on specific distributions of the profits of the bank and the nonbanking firm, as presented in the text, not on the expected value and variance of the profits. 63 Table 3 Means and Standard Deviations of Profit Rates for Firms in Financial Service Industries, 1975-84 Industry Average after-tax return on equity (ROE) Standard deviation of ROE Commercial banks Thrift institutions Securities brokers Securities underwriters Large investment banks only Life insurance underwriters Property-casualty insurance underwriters Insurance brokers and agents All manufacturing 12.3% 3.4 13.0 16.4 21.5 13.7 11.9 12.2 13.1 1.3% 10.7 4.0 5.7 7.7 2.3 6.4 4.1 2.0 SOURCE: Litan (1987), p. 64. nonbanking firm. A fourth factor is the size o f the bank relative to the nonbanking firm. The third and fourth factors are important because the bank may actually reduce its risk by acquiring a non banking firm that has a higher coefficient o f varia tion o f profits than the bank. This possibility w ill be demonstrated later. The fifth factor that must be considered is the “ synergies” (increase in profits) involved in com bining banking and nonbanking services in the same organization. Offering banking and nonbank ing services through the same firm may reduce the cost of providing the services and may attract cus tomers w ho value the w ider array o f services of fered by the combined bank-nonbank firm. These synergies could produce profit rates that exceed the sum o f the profit rates o f banks and firms in the nonbanking industry operating as separate corporations. 12Eisenbeis and Wall (1984) survey the studies. For more recent studies, see Boyd and Graham (1988) and Macey, Marr and Young (1987). There is evidence that BHCs reduce their risk by offering nonbanking services. See Boyd and Graham (1986), Wall (1987) and Brewer (1988). The results of these studies do not indicate the effects on risk of banking institutions entering nonbanking industries as permissible under the proposals in table 2. The nonbanking activities permissible for BHCs now are primarily those permissible for banks. The diversification of risk achieved by offering the nonbanking services currently permissible for BHCs may reflect merely geographic diversifi cation. Some Empirical Estimates o f Rates o f Return and Risk A number o f studies have investigated the profit rates in banking and selected nonbank activities.12 One finding, demonstrated in table 3, is that both the average profit rate and its standard deviation are low er in banking than in several industries that banks w ould be perm itted to enter under the recent proposals.13Indeed, the standard deviation of return on equity, one measure o f risk, is lowest in table 3 for the banking industry. Another key finding o f these studies is that the profit rates of banks are not positively correlated with the profits of firms in many industries that they w ould be permitted to enter. Thus, banks could diversify their risk by entering many nonbanking industries, even if the profits o f firms in those industries are more variable than those o f banks. 13Some studies measure returns to shareholders using data on stock prices and dividends. These studies report similar pat terns: mean rates of return and variability of returns to share holders are higher in several of the industries that banking organizations would be permitted to enter than in the commer cial banking industry. See Boyd and Graham (1988), Eisemann (1976) and Macey, Marr and Young (1987). JULY/AUGUST 1988 64 Table 4 Variability of Profits of Hypothetical Firms formed through the Merger of Banks and Firms in Various Financial Industries, 1962-82 Coefficient of variation Item Banks Banks Banks Banks Banks Banks Banks Banks Banks Banks 0.22 0.18 0.24 0.22 0.22 0.15 0.29 0.15 0.20 0.20 alone plus savings and loan associations plus personal credit agencies plus business credit agencies plus securities and commodities brokers plus life insurance plus mutual insurers plus insurance agents plus real estate operators and lessors plus subdividers and developers NOTE: A time series of the profits of each hypothetical firm is formed by assuming that 75 percent of the assets of the hypothetical firm are devoted to banking and 25 percent are devoted to the nonbanking activity. The coefficient of variation is derived for the constructed time series. SOURCE: Litan (1987), p. 88. Table 4 illustrates the potential reduction in variability o f bank profits possible through m er gers with firms that offer other financial services. The table illustrates this with the coefficient of variation, a measure o f relative risk that is calcu lated by dividing the standard deviation o f the profit rates by the mean. The results demonstrate, using a hypothetical situation involving the rela tive size o f banking and nonbanking components of the firm, that the combined firm can have the same or even low er risk than the bank itself, even though risk is higher in the nonbanking industries. Because banks have not yet entered the various nonbanking industries, there is little evidence on the magnitude o f the synergies involved in com bining banks with other firms.14There is evidence, however, o f synergies for banks and selected finan cial activities. For example, before the separation o f commercial banking and investment banking in '“Several studies estimate the effects of the combination of services offered by banks on their costs. See Gilligan and Smirlock (1984) and Benston, et. al. (1983). The results of these studies are not relevant in estimating the effects of non banking services on the costs of banks, since the data are for banks subject to current limitations on the services they may offer. 15White (1986). 16Daskin and Marquardt (1983). FEDERAL RESERVE BANK OF ST. LOUIS the 1930s, securities affiliates o f comm ercial banks held a large share o f the investment banking busi ness.15In nations where comm ercial banking orga nizations may offer investment banking services, commercial banking organizations have large shares o f the investment banking business.16 An Illustration The effects o f permitting banking organizations to offer nonbanking services on the risk and re turns in banking are analyzed using two probabil ity distributions o f profits, one for a hypothetical bank and another for a nonbanking firm. These probability distributions, presented in table 5, are designed to reflect the results o f studies o f risk and returns in banking and various nonbanking industries summarized above. Profit distributions are combined in table 6 under various assump tions that reflect the proposals for restructuring 65 Table 5 Probability Distributions of the Profits of a Bank and a Nonbanking Firm Prior to Merger or Affiliation Bank Outcome Probability Profits Return to shareholders A B C 0.01 0.98 0.01 -$ 1 1 0 10 130 -$ 1 0 0 10 130 Nonbanking firm Outcome Probability Profits Return to shareholders A B C 0.05 0.90 0.05 -$ 1 1 5 15 145 -$ 1 0 0 15 145 Bank Nonbanking firm Expected return to shareholders as a percentage of capital 10.1% Coefficient of variation of returns to shareholders 1.6117 Expected loss to the FDIC $0.10 15.75% 2.4637 the financial system described in table 2. Table 7 shows the returns to shareholders and the ex pected loss to the FDIC for the four cases analyzed in table 6. The illustration is designed to be simple. Differ ences among the four cases might change under assumptions that w ould make the analysis more complex. For instance, the management o f the firm that buys the bank and the nonbanking firm is assumed to make no changes that affect the capital ratios or the probability distributions o f profits. Analysis o f the cases under alternative assumptions is beyond the scope o f this paper. The bank begins the current year with book value o f equity equal to $100. The market value o f the bank is assumed to equal its book value prior to financial restructuring, which permits the af filiation o f the bank with the nonbanking firm. As presented in table 5, the (discrete) probability dis tribution o f the bank's profits in the current year has three possible outcomes: a 1 percent chance o f a loss o f $110, which w ould cause the bank to fail, a 98 percent chance o f a profit o f $10 (a 10 percent return on equity) and a 1 percent chance o f a profit o f $130.17 Table 5 also presents the probability distribu tion o f profits o f a nonbanking firm that, begins the year with book value capital o f $100. The market value o f the nonbanking firm is also assumed ini tially to equal $100. The nonbanking firm is riskier than the bank: the coefficient o f variation o f its profits is higher than that o f the bank. This speci fication was chosen to reflect the greater variability o f profits shown in table 3 in some o f the indus tries that banking institutions wish to enter. The effects o f combining the bank and the non banking firm in the same corporation are exam ined using three indicators: the expected return to shareholders as a percent o f capital, the coefficient o f variation o f returns to shareholders o f the con solidated firm, and the expected loss to the FDIC from the bank’s failure. These measures are calcu lated in table 5 for both the bank and the non banking firm as separate organizations to provide benchmarks for comparison. The distribution of returns to shareholders differs from the distribu tion o f profits because losses to shareholders are lim ited to the amount o f their initial investment in the firm. Thus, losses to shareholders are limited to $100 for the bank and $100 for the nonbanking firm. The expected loss to the FDIC is calculated as follows. The bank fails in only one o f the three possible outcomes: a loss o f $110, with a chance o f 1 percent. The loss to the FDIC in that outcome w ould be $10, since the initial capital o f the bank is $100. Thus, the expected loss to the FDIC is $10 (loss to FDIC) X 0.01 (probability) = $0.10. In deriving the distribution o f returns to share holders in table 6, one must specify their invest ment, which determines their maximum loss and the denom inator used in calculating their ex pected rate o f return. The shareholders' initial investment is measured as the book value o f the com bined firms. The use o f book value, net o f any accounting goodw ill resulting from the acquisition o f the bank and the nonbanking firm, provides a 17The large profit of the bank associated with the small probabil ity might reflect the recovery on loans previously charged off as losses or a large favorable change in market interest rates on portfolios of assets and liabilities that do not have matched duration. JULY/AUGUST 1988 FEDERAL RESERVE BANK OF ST. LOUIS Table 6 Distributions of Returns to Shareholders for Various Combinations of a Bank and a Nonbanking Firm Outcomes from underlying profit distributions (bank, nonbanking Outcome firm) ( ) 2 (3) (4) Affiliation, corporate separateness Affiliation, Heller proposal Affiliation, corporate separateness; bank lends $10 at a zero interest rate to its nonbank affiliate (1) Merger Probability (bank x non banking firm) Return to shareholders Loss to FDIC Return to shareholders $25 -$100- $100= - $200 Loss to FDIC Return to shareholders $10 -$100- $100 = - $200 1 A, A 0.01 x0.05 = 0.0005 -$ 1 0 0 -$100 = - $200 2 A, B 0.01 x0.90 = 0.009 - 110 + 3 A, C 0.01 X0.05 = 0.0005 - 110 + 145 = 4 B, A 0.98 x 0.05 = 0.049 10- 115= - 105 5 B,B 0.98 x 0.90 = 0.882 10 + 15 = 25 6 B,C 0.98x0.05 = 0.049 10 + 145 = 155 7 C, A 0.01 X0.05 = 0.0005 130- 115 = 15 130- 100 = 8 C, B 0.01 X0.90 = 0.009 130 + 15 = 145 130 + 15 = 145 130 + 9 C,C 0.01 x0.05 = 0.0005 130 + 145 = 275 130 + 145 = 275 15= - 95 - 100 + 35 - 100 + 145 = 15= - 10- 100= - 85 10 - 110 + 45 10 - 110 + 145 = 15 = - Loss to FDIC 95 35 $10 Return to shareholders Loss to FDIC -$100-$100 = -$200 —100 + (15 +1.053) = -100 + (145 + 1.053) = 10.50 46.053 10.50 90 10- 100 = - 15 = 25 10 + 15 = 25 (10 —0.50) + (15 + 1.053) = 25.553 10 + 145 = 155 10+ 145 = 155 (10 - 0.50) + (145 + 1.053) = 155.553 30 130- 100 = 30 (130 —0.50 —10) —100 = 19.500 15 = 145 (130 - 0.50)+ (15+ 1.053) = 145.553 130 + 145 = 275 (130 - 0.50)+ (145+ 1.053) = 275.553 10 + 90 $20.50 83.947 (10 —0.50) —10) —100 = - 100.500 o> o> 67 Table 7 Returns to Shareholders and Losses to the FDIC Under Various Combinations of a Bank and a Nonbanking Firm Case number Means of combining the firms Expected return to shareholders as a percentage of capital Coefficient of variation of returns to shareholders Expected loss to the FDIC 1 Merger 12.51% 1.7754 $0.0125 2 Affiliation, corporate separateness 12.93 1.6278 0.1000 3 Affiliation, Heller proposal 12.88 1.6434 0.0050 4 Affiliation, corporate separateness; bank lends $10 at zero interest rate to nonbank affiliate 12.93 1.6860 0.1100 basis for specifying bankruptcy. Book value also provides a com m on denom inator for comparisons o f expected rates o f return in the various cases. The market value o f the firm that buys the bank and the nonbanking firm w ill exceed their com bined book value. If this were not the case, the combination o f these firms in the same corpora tion w ould not benefit the shareholders. The profits o f the bank and the nonbanking firm are assumed to be statistically independent and, thus, uncorrelated. This assumption simplifies the analysis; it is also consistent with some o f the evi dence cited previously for several industries that banks could enter. For each outcom e for the profits o f the bank, there are three possible out comes for the profits o f the nonbanking firm. If com bined into one firm, there w ould be nine pos sible outcomes for the returns to shareholders of the consolidated firm, as table 6 illustrates. Tables 6 and 7 ignore the existence o f synergies from combining a bank with a nonbanking firm; they assume that there is no increase in the joint profits resulting from low er costs or a w id er array o f services to offer customers. As previously m en tioned, it is difficult to determine the magnitude o f such synergies, given that such combinations have been unlawful for many years. Such synergies, o f course, must exist to make such combinations attractive to shareholders; investors can easily obtain the benefits o f diversification by owning shares o f firms with uncorrelated profits. In this paper, however, assumptions about the size o f the synergies are unnecessary; the relevant com pari sons are made between the various cases. An in crease in the levels o f profits for each outcome w ould not alter the differences among the four cases examined in tables 6 and 7, unless the syn ergies eliminate bankruptcy in all outcomes. Merger o f the Bank and the Nonbanking Firm: The Simplest Case Each proposal described in table 2 calls for the new activities o f banking organizations to be con ducted through corporate entities that are sepa rate from banks. This feature o f the proposals reflects the view that the chances o f bank failure and the potential loss to the FDIC w ou ld be higher if the organizations that ow n banks offered non banking services through their bank subsidiaries, rather than through subsidiaries that are separate from the banks. JULY/AUGUST 1988 68 This view is not valid under all circumstances, as case 1 in tables 6 and 7 illustrates. In this case, the bank begins offering nonbanking services by merging with the nonbanking firm that has the profit distribution presented in table 5. The capital o f the bank after the merger is $200. Given the underlying profit distributions in table 5, there is only one outcom e in which the bank fails: in out come # 1, the returns from the banking and non banking activities yield the largest possible losses. In that outcome, the shareholders lose their total investment. The bank remains in operation in all o f the other outcomes. In outcomes # 2 and # 3, in which the losses from banking operations are large enough to make the bank fail if operating as a separate corporation, the profits from the non banking operations and the increased capital of the bank resulting from the m erger keep the bank from failing. The expected loss to the FDIC in case 1 depends on what happens to the liabilities o f the nonbank ing firm after the merger. Suppose the nonbanking segment o f the m erged firm continues to borrow from the same sources it used before the merger. If the claims o f these lenders are subordinated to the claims o f depositors, the merger might reduce the expected loss to the FDIC, perhaps to zero. In this illustration, however, the m erged organi zation converts all o f its liabilities to federally in sured deposits. If the bank involved in the merger goes bankrupt, the FDIC absorbs losses above the capital o f $200. In outcom e # 1, because the bank's maximum loss after its merger with the nonbanking firm is $225, the loss to the FDIC is $25. Although the maximum loss to the FDIC is larger after the merger, the expected loss ($25 X 0.0005) is actually smaller after the merger (com pare tables 5 and 7). The effects that a m erger have on the possibility o f bank failure and the expected loss to the FDIC depend on the size o f the nonbanking firm relative to the bank. To illustrate, suppose the bank merges with a nonbanking firm whose distribution o f profits is 10 times as large for each outcom e as that presented in table 5 and whose capital is $1,000. In this case, w hich is not shown in the table, the expected loss to the FDIC w ould be $2.04, much larger than the expected loss shown in table 7. Thus, in considering a restructuring of the financial system, the size o f the bank relative to the nonbanking firm is an important determ i nant o f the expected loss to the FDIC. http://fraser.stlouisfed.org/ FEDERAL RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis Affiliation o f a Bank with a Nonbanking Firm If banks combine with nonbanking firms, one way to limit the FDIC's expected loss is to require that banks remain separate corporations within their parent organizations and limit FDIC insur ance only to the deposit liabilities o f the banks. Within such structures, the principle o f corporate separateness w ould prevent the nonbanking firm ’s creditors from claiming the assets of the bank. The risk and return characteristics o f a holding company that buys the bank and the nonbanking firm are presented in case 2. Under this case, la belled “ affiliation, corporate separateness,” losses to shareholders o f the holding com pany resulting from losses by the nonbank subsidiary are lim ited to the capital o f the nonbank subsidiary. The bank does not rescue the nonbank subsidiary by ab sorbing the additional losses. In turn, if the bank has losses that exceed its capital, the nonbank subsidiary does not rescue the bank by absorbing the additional losses. There is assumed to be no lending among units o f the holding company. The holding com pany lends to neither the bank nor the nonbank subsidiary, and the bank lends noth ing to the nonbank affiliate. The nonbank affiliate borrows, instead, from nonaffiliated lenders; the liabilities o f the bank are covered by FDIC insur ance. The expected return to the shareholders is higher and the variability of returns is low er in case 2 than under a similar combination o f firms arranged through a merger. Thus, the sharehold ers benefit more from a combination o f the bank and the nonbanking firm as affiliates o f a holding company than through the m erger of these firms. The benefit to the shareholders, however, comes partly at the expense o f the FDIC. The FDIC's ex pected loss is the same in case 2 as in the ben chmark case in table 5 but higher than under the merger. Under affiliation and corporate separate ness, the outcomes in which the FDIC is exposed to losses are determined by the probability distri bution o f the bank’s profits. Under the merger illustrated in case 1, in contrast, losses in out comes # 2 and # 3 that w ou ld make the bank fail are absorbed by the profits o f the nonbank seg ment o f the m erged firm and the capital contrib uted by the nonbanking unit. Under affiliation and corporate separateness, however, the expected loss to the FDIC does not depend on the size of the bank relative to its nonbank affiliate. 69 IMPLICATIONS FOR THE PROPOSALS Merger or Affiliation The cases in tables 6 and 7 indicate that, under some conditions, the risk o f FDIC loss w ould be low er if a bank engages in a nonbanking activity directly, rather than through affiliation with a non banking firm. In considering proposals for finan cial restructuring, therefore, it is unnecessary to prohibit the direct offering o f nonbanking services through banks under all circumstances. The Financial Services Holding Company (FSHC) Proposal The proposals by the Association o f Bank Hold ing Companies (LaWare (1987)) and the Associa tion o f Reserve City Bankers (1987) w ould permit FSHCs to acquire banks as subsidiaries under the condition o f affiliation and corporate separate ness. The bank could not use its assets to rescue a failing nonbank affiliate, and the FSHC w ou ld not be required to rescue a failing bank. A comparison o f case 2 in table 7 with table 5 shows how the formation o f FSHCs can affect risk in banking. Affiliation o f a bank with a nonbanking firm reduces the probability that the bank w ill fail only if affiliation yields synergies that raise the profits o f the bank for each possible outcome. Thus, affiliations between banks and nonbanking firms that facilitate diversification o f risk for share holders o f banking firms reduce the probability of bank failure and the expected loss to the FDIC only if there are synergies from combining banking and nonbanking firms in the same organization. The Heller “Double Umbrella” Proposal The distribution o f returns to shareholders un der the Heller (1987) proposal is presented under case 3 in table 6. The implications o f this proposal can be illustrated by comparing the distribution of returns to shareholders under various outcomes in cases 2 and 3. Under the Heller proposal, the losses o f the bank and nonbank subsidiary in out com e # 1 absorb all o f the capital o f the holding company. The FDIC has a loss o f $10 in that out come, the amount by which the loss o f the bank exceeds its capital. In outcome # 2, the bank has a loss that exceeds its capital, but the holding com pany is required to cover that loss, drawing on its profit o f $15 from the nonbanking subsidiary and its capital. The holding com pany also covers the large loss o f the bank in outcom e # 3. In outcomes # 4 and # 7, in contrast, the holding company does not absorb all o f the losses o f the nonbanking subsidiary. Instead, the nonbanking subsidiary goes bankrupt. The holding com pany writes off its investment o f $100, and nonaffiliated lenders ab sorb the additional loss o f $15 in each o f these outcomes. The minimum level o f synergies necessary to make combinations o f banks and nonbanking firms attractive to investors is higher under the Heller proposal than under the FSHC proposal. The diversification o f risk illustrated in case 2 could be achieved through a mutual fund that buys shares in firms in banking and nonbanking industries. Any synergies w ould make the share holders’ expected rate o f return higher with the bank and nonbanking firm combined in the firm under affiliation and corporate separateness than through a mutual fund. To make combinations o f banks and nonbanking firms under the Heller proposal attractive to shareholders, synergies w ould have to exceed a level necessary to com pensate the holding com pany for the expected cost o f bailing out the failing bank subsidiary. The synergies necessary to make the affiliation o f banks with nonbanking firms profitable under the Heller proposal w ou ld be different for each potential combination o f firms. For case 3, the synergies w ould have to raise the returns to share holders by $0,095 to make them equal to the ex pected returns to shareholders in case 2, and even more to compensate shareholders for the higher variability o f returns in case 3. The Corrigan Proposal Corrigan (1987) assumes that the methods of insulating banks built into the proposals for FSHCs w ill be ineffective. This view is based on evidence that BHCs are integrated organizations that have used all o f their resources, including those o f their bank subsidiaries, to support any nonbank subsid iary in danger of failing. Corrigan also expresses concern that, in approving the acquisition o f banks by nonbanking firms, the federal supervi sory authorities w ill extend the federal safety net to the parent organizations themselves. The Effects o f Loa ns to N o n b a n k Affiliates on S to ck h o ld e r Wealth — The Corrigan proposal reflects these views on the relationship between banks and their parent organizations. Case 4 in tables 6 and 7 examines whether such concerns reflect rational, profit-m axim izing behavior. The JULY/AUGUST 1988 70 Corrigan proposal assumes that firms are willing to risk the assets o f their bank subsidiaries to aid their nonbank subsidiaries. One w ay for a holding com pany to do this is to allow the bank to lend directly to the nonbank subsidiary. To illustrate this, the bank in case 4 lends $10 to the nonbank affiliate at a zero interest rate, thus subsidizing the nonbank subsidiary at the expense o f the bank. Several assumptions have been made to derive the probability distribution o f returns for share holders o f the holding company. First, the bank loan is assumed to be subordinated to other debt o f the nonbank affiliate. If the nonbank affiliate goes bankrupt, therefore, the bank absorbs the first $10 o f losses to creditors. Second, the interest rate on riskless assets is assumed to be 5 percent. The distribution o f profits for the bank is derived by subtracting $0.50 from the profits for each pos sible outcom e presented in table 5; this reduction reflects the opportunity cost o f foregoing an alter native investment o f $10 at the riskless rate. The nonbank subsidiary saves $1,053 in interest expense on the $10 it borrows from the bank; this is the amount that a risk-neutral lender charges to compensate for the risk-free rate o f 5 percent and the 5 percent chance o f losing the $10 principal and foregoing the interest incom e if the nonbank ing firm goes bankrupt.18 The effects o f this loan on the distribution o f shareholders' returns are illustrated in table 6 under case 4. In outcomes # 1, # 4 and # 7, the bankruptcy o f the nonbanking firm imposes an additional loss o f $10 on the bank. In outcom e # 1, in w hich the bank has its largest losses, the FDIC absorbs a loss o f $20.50 ($10 loss from the underly ing distribution in table 5, $0.50 loss o f interest incom e on the loan to the nonbank affiliate and $10 loss on the loan to the nonbank affiliate). The cost saving by the nonbank affiliate due to the zero interest loan from the bank raises the returns to shareholders by $1,053 in all outcomes except those in which the nonbank affiliate goes bankrupt. The return to shareholders is $0.01 higher in case 4 than in case 2; this difference is 18The interest rate that the nonbank affiliate would pay to borrow from a nonaffiliated lender is determined by calculating the rate that would make the expected return on such a loan equal to the risk-free interest rate. Let rl be the interest rate on the loan and rs the risk-free rate. In lending $10 to the nonbank affiliate, there is a 95 percent chance of collecting the principal plus interest at the rate rl and a 5 percent chance of losing the principal and collecting no interest. The expected returns on the alternative investments are calculated as follows: http://fraser.stlouisfed.org/ FEDERAL RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis not large enough, however, to raise the expected rate o f return in table 7 by 1 basis point. The im portant difference between the distributions of returns in case 4 and case 2 is that the coefficient of variation o f the returns is higher in case 4. Thus, it is not in the shareholders’ interest to have their bank lend to its nonbank subsidiary, even at a subsidized rate. Such loans make their returns more variable. Typically, bank supervisors w ou ld make such a loan even less attractive to the shareholders. Be cause the loan to the nonbank affiliate raises the expected loss to the FDIC, bank supervisors w ould require the bank to maintain a higher capital ratio. Though the bank could raise its capital ratio by reducing its total assets w hile keeping its capital unchanged, the asset reduction w ould reduce the level o f profits for each possible outcom e the bank faces. This analysis is consistent with evidence that few banks make loans to their nonbank affiliates up to the limits allowed by regulation. Bose and Talley (1983) examine transactions among affiliates of 224 o f the 229 BHCs that filed reports with the Federal Beserve from the fourth quarter o f 1975 through the fourth quarter o f 1980. In 1980, 27 percent o f the BHCs had no transactions among affiliates. Am ong the 16 BHCs in which the bank subsidiaries made larger loans to the nonbank affiliates than the nonbank affiliates made to the banks, loans to the nonbank affiliates in 1980 were only 1.3 percent o f the capital o f the bank subsidi aries. B anking R isk un d er A ssum ptions O ther Than P r o fit M a xim ization — The distribution o f returns in cases 2 and 4 reflect the assumption that, if the bank does not lend to the nonbank affiliate, the affiliate's bankruptcy does not affect the bank’s profits. In a few cases, however, the bankruptcy of a nonbank subsidiary of a holding com pany has induced depositors to withdraw their deposits from the bank subsidiary.13The management o f a holding company, therefore, might justify loans from a bank subsidiary to a nonbank affiliate as a w ay to prevent the nonbank subsidiary from going rl x $10 x 0.95 - $10 x 0.05 = rs x $10. If rs is 5 percent, rl = [0.05 + 0.05] + 0.95 = 0.1053. ,9Cornyn, et. al. (1986). 71 bankrupt and thus make depositors less con cerned about the safety o f their deposits. In this case, the costs o f bailing out the nonbanking sub sidiary might be less than the cost o f adverse reac tion by depositors. There have been several cases in which the management o f a BHC used the resources o f a bank subsidiary to aid a nonbank affiliate in dis tress. In the mid-1970s, for example, the holding com pany that ow ned the Hamilton National Bank o f Chattanooga, Tennessee, arranged for the bank to buy low-quality mortgages from a mortgage banking affiliate. The mortgage purchase was an important factor that led to the failure o f the bank.20In October 1987, to cite another case, the Continental Illinois National Bank made a loan that exceeded its limit for loans to one customer to a subsidiary that deals in options. The subsidi ary suffered a large loss after the sharp fall in stock prices that month. The rationalization behind bank loans to bail out the nonbank affiliate overlooks an alternative that might be more favorable to the shareholders o f the holding company: let the nonbank subsidi ary go bankrupt and sell the bank to another party. Losses to the holding com pany w ou ld be limited to its investment in the nonbank subsidi ary, with nonaffiliated lenders forced to absorb any additional losses. If potential bidders are con cerned that the bank made loans to the failing nonbank affiliate or in some way assumed respon sibility for the debts o f that affiliate, the FDIC could facilitate the sale by offering to reimburse the winning bidder for any losses resulting from the failure o f the nonbank affiliate. Management o f the holding com pany may p re fer to have the bank absorb the losses necessary to bail out the failing nonbank affiliate, rather than sell the bank, which w ill result in the loss o f their jobs. It may be in managem ent’s interest to ar range for the bank to lend to the nonbank subsidi ary and pray that some favorable outcom e helps the holding com pany remain solvent. The possi bility o f such action is w hy government supervi sors must remain aware o f any financial problems in firms that own banks and must subject the bank subsidiaries o f those firms to particularly close supervision. “ Ibid., p. 186. 21Suppose the bank has a capital-to-asset ratio of 10 percent. For all federally insured commercial banks, the average ratio of loans to assets is about 60 percent. Thus, $600 is a reasonable The analysis in tables 6 and 7 o f a bank lending to its nonbank affiliate is based on the assumption that the loan is used for legitimate business pur poses. Loans from a bank to a nonbank affiliate, of course, could be made for fraudulent purposes. Suppose a bank is perm itted to make a loan o f any amount to an affiliate. One m ethod o f stealing from a bank w ould be to buy the bank through a holding company, arrange for a loan that ex ceeded the investment o f the holding company in the bank and disappear with the proceeds o f the loan. The potential for fraud indicates that it may be prudent to prohibit loans to affiliates that exceed the capital o f a bank. This prohibition w ould not prevent all forms o f fraud in banking, but its viola tion w ou ld indicate to the bank supervisors when a bank is vulnerable to this type o f fraud. It is also prudent to screen the background o f those w ho buy banks through holding companies, as the federal bank regulatory agencies do w hen individ uals buy banks. The FDIC (1987) proposal calls for greater au thority to audit the terms o f any loans banks make to affiliates or subsidiaries. This proposal does not indicate what bank examiners w ould look for in such audits. Audits to detect fraud w ould be ap propriate. The Safe Bank Proposal The so-called safe bank proposal (Litan (1987)) is intended to reduce the expected level and stand ard deviation o f profit rates o f banks subject to the “ safe bank” asset restrictions. As the appendix indicates, for each $100 o f assets shifted from busi ness loans to Treasury bills, the revenue o f the safe bank w ould decline by $1.26. The asset limitations for safe banks may be so restrictive that they w ould prevent many affiliations o f banks with nonbanking firms that w ould prom ote diversifica tion or benefit society through synergies. One way to evaluate the safe banking proposal is to compare the size o f the synergies necessary to make bank acquisitions profitable for nonbanking firms to the synergies necessary under alternative proposals. Suppose the bank had loans o f $600.21 If the bank becom es a safe bank by reinvesting the $600 in Treasury bills, its revenue falls by $7.56. It level for loans of the hypothetical bank with capital of $100 and a 10 percent capital ratio. JULY/AUGUST 1988 72 must, however, continue to pay competitive inter est rates on deposits after becom ing a subsidiary to avoid a decline in its deposits. Thus, synergies from the operation o f the bank as a subsidiaiy must be worth at least $7.56 to the holding com pany. This amount can be compared to the syner gies necessary to make the acquisition o f a bank subsidiary profitable under the Heller proposal, which is $0,095 for the case examined above. This large difference reflects the fact that the safe bank proposal imposes a significant opportu nity cost on a nonbanking firm that buys a bank under each possible outcome. The Heller pro posal, on the other hand, imposes a loss on the nonbanking firm under an unlikely outcom e — the failure o f the bank subsidiary. These com pari sons suggest that few er combinations o f banking and nonbanking firms that w ould prom ote diversi fication o f risk and, possibly, more efficient use of resources w ou ld be viable under the safe bank proposal than under the Heller proposal. CONCLUSIONS Several barriers separating banking from other industries have been rem oved in recent years, w hile Congress debates a more com plete restruc turing o f the financial system. Much evidence indicates that banking organizations could diver sify risk by affiliating with firms in a w ide variety of other industries, even those with m ore variable profits than the banking industry. This paper illus trates the potential for risk diversification through the com m on ownership o f a hypothetical bank and nonbanking firm. The illustration has several implications for current proposals for restructuring the financial system. Banks are not necessarily made safer by requiring that all nonbanking activities be con ducted through separate subsidiaries. On the con trary, banks may be less vulnerable to failure if some nonbanking activities are offered through the banks directly. Moreover, the expected loss of federal deposit insurance funds may be low er even if the nonbanking activities are financed through insured deposits. The major proposals for restructuring the finan cial system w ould permit firms in various indus tries to buy banks and operate them as separate subsidiaries. Some o f the proposals build in safe guards to prevent nonbanking firms from using the resources o f their bank subsidiaries in ways that w ould increase both the chance for bank fail ure and the expected loss o f the federal deposit http://fraser.stlouisfed.org/ FEDERAL RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis insurance funds. These restrictions are based on the presumption that, without such safeguards, nonbanking firms w ou ld use the resources o f their bank subsidiaries to benefit their nonbank subsidi aries. The analysis in this paper indicates that the shareholders o f a holding com pany generally do not benefit by having their bank subsidiary lend at a subsidized interest rate to the nonbank subsidi ary. In fact, shareholders are made worse off by such transactions because the holding com pany profits becom e more variable. Transactions that benefit nonbank subsidiaries at the expense o f bank subsidiaries do not increase the sharehold ers' wealth. The greatest danger in banks lending to affiliates involves management o f holding com panies attempting to save their jobs by bailing out nonbank subsidiaries and fraudulent schemes to steal from banks through loans to affiliates. Tw o o f the proposals place special constraints on the nonbanking firms that buy banks to limit the risks o f bank failure. One proposal requires that the holding companies absorb all losses in curred by banks, up to the holding com pany’s total capital. The other proposal requires the bank subsidiaries o f nonbanking firms to hold only lowrisk liquid assets. Both proposals raise the level o f synergies necessary to make the acquisition o f banks by nonbanking firms profitable. Of these proposals, the safe banking proposal is the more restrictive. Some consolidations o f banking and nonbanking firms that w ou ld yield social benefits in the form o f higher profits and reduced variation in stockholder returns w ould not be attractive to shareholders under the safe banking proposal but w ould be attractive under other proposals. REFERENCES Association of Reserve City Bankers Emerging Issues Committee Proposal for a Financial Sen/ices Holding Company (March 19, 1987). Association of Reserve City Bankers. Benston, George J., et. al. “ Economies of Scale and Scope in Banking,” in Proceedings of a Conference on Bank Structure and Competition (Federal Reserve Bank of Chicago, May 2-4, 1983), pp. 432-55. Black, Fischer, Merton H. Miller, and Richard A. Posner. “ An Approach to the Regulation of Bank Holding Companies," Journal of Business (July 1978), pp. 379-412. Boyd, John H., and Stanley L. Graham. “ The Profitability and Risk Effects of Allowing Bank Holding Companies to Merge With Other Financial Firms: A Simulation Study,” Federal Reserve Bank of Minneapolis Quarterly Review (Spring 1988), pp. 3-20. ________ _ “ Risk, Regulation, and Bank Holding Company Expansion into Nonbanking,” Federal Reserve Bank of Minneapolis Quarterly Review (Spring 1986), pp. 2-17. 73 Brewer, Elijah, III. “A Note on the Relationship Between Bank Holding Company Risk and Nonbank Activity,” Federal Re serve Bank of Chicago Staff Memoranda SM88-5 (1988). Cornyn, Anthony, et. al. “ An Analysis of the Concept of Corpo rate Separateness in BHC Regulation from an Economic Perspective,” in Proceedings of a Conference on Bank Struc ture and Competition (Federal Reserve Bank of Chicago, May 14-16, 1986), pp. 174-212. Corrigan, E. Gerald. Financial Market Structure: A Longer View (Federal Reserve Bank of New York, January 1987). Daskin, Alan J., and Jeffrey C. Marquardt. “ The Separation of Banking from Commerce and the Securities Business in the United Kingdom, West Germany and Japan,” Issues in Bank Regulation (Summer 1983), pp. 16-24. Eisenbeis, Robert A., and Larry D. Wall. “ Bank Holding Company Nonbanking Activities and Risk,” in Proceedings of a Conference on Bank Market Structure and Competition (Federal Reserve Bank of Chicago, April 23-25, 1984), pp. 340-57. Eisemann, Peter C. “ Diversification and the Congeneric Bank Holding Company,” Journal of Bank Research (Spring 1976), pp. 68-77. Federal Deposit Insurance Corporation. Mandate for Change: Restructuring the Banking Industry (October 1987). Federal Home Loan Bank Board. “ Supplemental Materials From the Comptroller of the Currency, FDIC, and Federal Home Loan Bank Board,” in Structure and Regulation of Financial Firms and Holding Companies (Part 1), Hearings before a Subcommittee of the Committee on Government Operations, House of Representatives, 99 Cong., 2 Sess. (April 22, June 11, and July 23, 1986), pp. 255-385. Forrestal, Robert P. “ Regulations Must Evolve Along with Financial Services Industry,” Address to the Economics Club of Connecticut, December 4,1987, reprinted in the American Banker (December 23,1987), pp. 4-7. Gilbert, R. Alton. “ Banks Owned by Nonbanks: What is the Problem and What can be Done about It?,” Business and Society (Roosevelt University, Spring 1987), pp. 9-17. Gilbert, R. Alton, Courtenay C. Stone, and Michael E. Trebing. “The New Bank Capital Adequacy Standards,” this Review (May 1985), pp. 12-20. Gilligan, Thomas, Michael Smirlock, and William Marshall. “ Scale and Scope Economies in the Multi-Product Banking Firm,” Journal of Monetary Economics (May 1984), pp. 393-405. Heller, H. Robert. “ The Shape of Banking in the 1990s,” Address before the Forecasters Club of New York, June 26, 1987. Huertas, Thomas F. “The Protection of Deposits from Risks Assumed by Non-bank Affiliates,” in Structure and Regulation of Financial Firms and Holding Companies (Part 3), Hearings before a Subcommittee of the Committee on Government Operations, House of Representatives, 99 Cong., 2 Sess. (December 17 and 18,1986), pp. 325-60. ________ _ “ Redesigning Regulation: The Future of Finance in the United States,” Issues in Bank Regulation (Fall 1987), pp. 7-13. Kareken, John H. “ Federal Bank Regulatory Policy: A Description and Some Observations,” Journal of Business (January 1986), pp. 3—48. LaWare, John P. “ FSHCA — The Flexible Alternative for Financial Restructuring,” Issues in Bank Regulation (Fall 1987), pp. 25-27. Litan, Robert E. What Should Banks Do? (The Brookings Institution, 1987). Macey, Jonathan R., W. Wayne Marr, and S. David Young. “ The Glass-Steagall Act and the Riskiness of Financial Intermediaries,” mimeo (Tulane University, November 1987). Rose, John T. “ Government Restrictions on Bank Activities: Rationale for Regulation and Possibilities for Deregulation,” Issues in Bank Regulation (Autumn 1985), pp. 25-33. Rose, John T., and Samuel H. Talley. Financial Transactions within Bank Holding Companies, Staff Studies 123 (Board of Governors of the Federal Reserve System, May 1983). Spong, Kenneth. Banking Regulation: Its Purposes, Implementation, and Effects (Federal Reserve Bank of Kansas City, 1985). Tobin, James. “A Case for Preserving Regulatory Distinctions,” Challenge (November/December 1987), pp. 10-17. Volcker, Paul A. “ Appendices to the Statement by Paul A. Volcker, Chairman, Board of Governors of the Federal Reserve System,” in Structure and Regulation o f Financial Firms and Holding Companies (Part I), Hearings before a Subcommittee of the Committee on Government Operations, House of Representatives, 99 Cong., 2 Sess. (April 22, June 11, and July 23,1986), pp. 391-510. Wall, Larry D. “ Has Bank Holding Companies’ Diversification Affected Their Risk of Failure?” Journal of Economics and Business (November 1987), pp. 313-26. White, Eugene Nelson. “ Before the Glass-Steagall Act: An Analysis of the Investment Banking Activities of National Banks,” Explorations in Economic History (January 1986), pp. 33-55. JULY/AUGUST 1988 74 Appendix The Opportunity Cost Of Holding Safe Assets The safe bank proposal (Litan (1987)) w ould put the bank subsidiaries o f nonbanking firms at a disadvantage in com peting for deposits by restrict ing the return on their investments. This disadvan tage could be offset slightly by waiving deposit insurance premiums for the subsidiaries o f non banking firms. Under the requirements for holding only safe assets, the subsidiaries o f nonbanking firms w ould not expose the federal deposit insur ance funds to potential losses; therefore, an argu ment could be m ade for exempting “safe” banks from deposit insurance premiums. The opportunity cost o f investing in Treasury securities instead o f loans is estimated using data from the functional cost analysis program o f the Federal Reserve. A change in the com position o f a bank’s assets affects its interest revenue and ex penses. The functional cost data includes informa tion on interest incom e and expenses allocated to various categories o f loans, as w ell as expenses involved in purchasing and holding securities. Table A l indicates that the gross yields on loans almost always exceed those on three-month Trea sury bills. Net yields on loans, which reflect ex penses and losses, are low er than the net yields on Treasury bills in some years for mortgage and installment loans. Table A1 Gross and Net Yields on Bank Assets Treasury bills Real estate mortgage Installment loans Commercial and other loans Year Number of banks Gross Net Gross Net Gross Net Gross Net 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 86 96 99 98 109 102 85 80 59 63 76 90 82 81 75 4.07% 7.04 7.89 5.84 4.99 5.27 7.22 10.04 11.51 14.03 10.69 8.63 9.58 7.48 5.98 3.92% 6.88 7.72 5.67 4.83 5.11 7.08 9.86 11.28 13.81 10.54 8.47 9.43 7.31 5.75 7.58% 8.11 8.57 8.17 8.39 8.84 8.88 9.32 10.01 10.80 10.84 11.02 11.41 11.60 10.21 6.82% 7.35 7.77 7.36 7.46 7.89 7.93 8.39 9.29 9.88 9.95 9.95 10.31 10.33 8.50 10.19% 10.29 10.77 11.01 11.11 11.05 11.43 12.00 12.90 14.90 15.87 14.98 14.39 13.41 12.50 6.54% 6.65 6.90 6.81 6.91 7.31 8.02 8.57 9.18 10.94 11.96 11.07 11.10 10.16 9.11 6.71% 8.44 10.53 8.88 8.22 8.21 9.67 12.23 14.31 16.85 14.96 11.93 12.82 11.30 10.21 5.35% 7.21 9.09 7.17 6.39 6.46 8.16 10.68 12.62 14.86 12.36 9.26 10.34 8.91 7.73 NOTE: Data on the gross and net yields for the three categories of loans are derived from the functional cost accounting data. These data are for the banks with total assets greater than $200 million. The second column indicates the number of banks in that size category that reported data for the investment function each year. The choice of this largest size category in the functional cost accounting reports is based on the assumption that the safe banks owned by relatively large nonbanking firms would tend to have assets above this dollar level. Net yields on loans reflect adjustments of the gross yields for expenses in making and servicing loans and loss rates on the various types of loans. The gross yields on Treasury bills are the annual averages of yields on three-month Treasury bills, new issues. Net yields on Treasury bills are the gross yields minus the costs of buying and holding investments per dollar of investments in the functional cost accounting data. Under the safe bank proposal, safe banks could hold longer-term Treasury securities, but the longer-term securities have greater potential for capital gains and losses. This exercise uses the yields on short-term Treasury securities and ignores capital gains and losses. http://fraser.stlouisfed.org/ FEDERAL RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis 75 Table A2 isolates the comparisons between net yields on Treasury bills and those on three catego ries o f loans. Net yields on mortgages and install ment loans tend to fall below the net yields on Treasury bills in periods o f sharp increases in interest rates. The most stable spread is that be tween the net yield on comm ercial and other loans and the net yield on Treasury securities. On average, banks lose $1.26 in net incom e before incom e taxes per dollar transferred from com m er cial loans to Treasury bills. Table A2 Sacrifice of Income Before Income Taxes per $100 Dollars of Loans Shifted to Treasury Bills Loan Categories Year Real estate mortgages 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 $2.90 0.47 0.05 1.69 2.63 2.78 0.85 - 1 .4 7 - 1 .9 9 -3 .9 3 -0 .5 9 1.48 0.88 3.02 2.75 Mean .768 Installment loans $2.62 -0 .2 3 -0 .8 2 1.14 2.08 2.20 0.94 -1 .2 9 -2 .1 0 -2 .8 7 1.42 2.60 1.67 2.85 3.36 .905 Commercial and other loans $1.43 0.33 1.37 1.50 1.56 1.35 1.08 0.82 1.34 1.05 1.82 0.79 0.91 1.60 1.98 1.262 JULY/AUGUST 1988 F e d e ra l R e s e rv e Bank o f St. Lou is Post O ffice Box 442 St. Louis, Missouri 63166 The R e v ie w is pub lish ed si# tim es p e r year by the R esearch and P u b lic In fo rm a tio n D ep a rtm en t o f the Federal Reserve Bank o f St. Louis. S ingle-copy su b s crip tion s are available to the p u b lic f r e e o f charge. M a il requests f o r s ub scription s, back issues, o r address changes to : Research and P u b lic In fo rm a tio n D epartm ent, Federal Reserve Bank o f St. Louis, P.O. Box 442, St. Louis, M is s o u ri 63166. The views expressed are those o f the individual a u th ors and d o n o t necessarily re fle ct o fficia l p o s itio n s o f the Federal Reserve Bank o f St. L ou is o r the Federal Reserve System. A rticles herein may be rep rin ted p rovid ed the s o u rce is credited. Please provid e the B a nk’s R esearch and P u b lic In fo rm a tio n D epartm en t with a cop y o f rep rin ted m aterial.