The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
Economic mg Review M I FEDERAL RESERVE BANK OF ATLANTA OCTOBER 1986 THE DOLLAR ABROAD Impact on Prices NONBANK ACTIVITY Greater Risks for Banks? U.S. TRADE Shifting Shares President Robert P. Forrestal Sr. Vice President and Director of Research Sheila L. Tschinkel Vice President and Associate Director of Research B. Frank King Financial Institutions and Payments David D. Whitehead, Research Officer Peter Abken Larry D. Wall Robert E. Goudreau Macropolicy Robert E. Keleher, Research Officer Mary S. Rosenbaum Thomas J . Cunningham Jeffrey A. Rosensweig Joseph A. Whitt, Jr. Regional Economics Gene D. Sullivan, Research Officer William J. Kahley Jon Moen Joel R. Parker W. Gene Wilson Visiting Scholars Russell Boyer William Hunter Economic Review Public Information and Publications Bobbie H. McCrackin, Director Public Information Larry J . Schulz, Public Information Coordinator Linda Donaldson Editorial Harriette Grissom, Publications Coordinator Melinda Dingler Mitchell Ann L. Pegg Graphics and Typesetting Cheryl B. Birthrong Steve Gelinas Word Processing Belinda Womble Distribution George Briggs Vivian Wilkins Ellen Gerber The Economic Review seeks to inform the public about Federal Reserve policies and the economic environment and, in particular, to narrow the gap between specialists and concerned laymen Views expressed in the Economic Review are not necessarily those of this Bank or the Federal Reserve System Material may be reprinted or abstracted if the Review and author are credited Please provide the Bank's Research Department with a copy of any publication containing reprinted material. Free subscriptions and additional copies are available from the Information Center, Federal Reserve Bank of Atlanta 104 Marietta Street N.W., Atlanta, Ga 303032713 (404/521-8788). Also contact the Information Centerto receive Southeastern Economic Insight a free newsletter on economic trends published by the Atlanta Fed twice a month. The Review is indexed online in the following data-bases- ABI/lnform, Magazine Index Management Contents, PAIS, and the Predicasts group. ISSN 0732-1813 V O L U M E LXXI, NO. 8, O C T O B E R 1986, E C O N O M I C R E V I E W Table of Contents The Dollar and Prices: An Empirical Analysis 4 Joseph A. Whitt, Jr., Paul D. Koch, and Jeffrey A. Rosensweig An alternative approach to the data suggests links between the dollar's value and U.S. prices from 1973 to 1985. Nonbank Activities and Risk 19 Larry D. Wall Does adding nonbank operations jeopardize the financial stability of bank holding companies? This article considers several angles. Economic Briefs 36 The Changing Pattern of U.S. Trade: 1975-1985 Statistical Summary Finance, Construction, General, Employment FEDERAL RESERVE BANK OF ATLANTA 3 43 The Dollar and Prices: An Empirical Analysis Joseph A. Whitt, Jr., Paul D. Koch, and Jeffrey A. Rosensweig Do changes in the dollar's value on foreign exchange markets affect prices in the United States? This article, based on an Atlanta Fed working paper "The Dynamic Relationship Between the Dollar and U.S. Prices: An Intensive Empirical Investigation," suggests a significant connection between moves in the dollar and periods of inflation and disinflation. 4 OCTOBER 1986, E C O N O M I C R E V I E W W h e n floating exchange rates were adopted in the early seventies, they brought wide fluctuations in the values of world currencies on foreign exchange markets. Economists have been trying ever since to determine how, and if, changes in the value of the dollar on foreign markets affect domestic prices. The interaction is complex. Like many macroeconomic relationships, the effect of the dollar's value on domestic price levels involves complicated lag patterns overtime as well as possible feedback from prices to exchange rates. Significant correlation between moves in the dollar's value and fluctuations in U.S. prices could have important policy implications, since it suggests that price stability will be hard to attain as long as exchange rates continue to fluctuate widely. Periods of unexpected dollar depreciation would be associated with worsening inflation, while unexpected appreciation would be associated with disinflation. Using a methodological approach especially suited to the complexities ushered in by floating exchange rates, our research confirms a group of previous studies that used other methods to discover a significant dollar-price level relationship and suggests that changes in the dollar's value may be associated with larger price changes than previous studies have indicated. Our results provide a stylized description of past relationships and are not suitable for direct or precise extrapolation into the future. Most existing research, which uses different methodologies than ours, estimates that a permanent 10 percent drop in the dollar's value is associated with an eventual increase of 1 to 2 percent in the consumer price index. Two recent studies find essentially no impact. Our study finds a definite price response, with the increase being approximately 4 and 3/4 percent. All of these estimates are based on the limited data available on the past relationships between the dollar and prices. U p until a few years ago most Americans had little direct experience with exchange rate changes, though other countries have certainly experienced them. However, since the advent of floating exchange rates, the dollar's value has moved up or down several times by international Joseph A. Whitt, Jr. and Jeffrey A. Rosensweig are economists in the Research Department of the Federal Reserve Bank of Atlanta. Paul D. Koch is an associate professor of economics at Kansas State University and was a visiting scholar at the Atlanta Fed. FEDERAL RESERVE BANK OF ATLANTA Chart 1. Inflation and Value of the Dollar Dollar Trade-weighted dollar index Inflation Percent Periods of marked decline in the dollar's value such as 1971-73 and 1977-78 appear to be followed by rises in inflation. Source: Federal Reserve Board of Governors and U.S Bureau of Labor Statistics quarterly data from 1970:1 to 1985:2. more than 10 percent within a 12-month period. For a variety of reasons, estimated relationships involvingthe dollar and prices based on past data may not necessarily remain unchanged in the future. Moreover, the dollar is not the only variable which affects inflation. Nevertheless, our finding of a strong past association between the dollar and subsequent inflation suggests that the dollar bears close watching to help gauge the outlook for inflation. Perhaps the most well-known measure of the overall price level is the consumer price index (CPI), which measures the cost of living for an average household in the United States. It is not obvious that there should be an important link between the value of the dollarand the CPI. After all, U.S. merchandise imports, which are presumably the items directly affected by exchange rate changes, constituted less than 10 percent of the U.S. gross national product in recent years. Nevertheless, a casual look at the behavior of inflation (in terms of the consumer price index) compared with an index of the dollar's exchange value from 1970 to mid-1985 (as measured by the Federal Reserve Board's trade-weighted index) shows a relationship (Chart 1). During this period phases of dollar depreciation such as 1971 -73 and 1977-78, indicated by declines in the dollar index, tended to be followed by increases in inflation, 5 whereas the dollar's appreciation in 1981-84 is matched by subsequent disinflation. To understand this relationship fully, a dynamic mathematical model that can account for patterns of change in a given time period is essential. During the era of fixed (or pegged) exchange rates when the values of currencies held steady except for occasional large jumps, modeling the impact of shifts in the dollar's value was fairly straightforward. W i t h floating exchange rates, however, the values of currencies are always moving. W e cannot say simply, for example, that a drop in the value of the dollar at one particular point generated the increase in prices at another particular point; rather, a whole pattern of changes in the dollar's exchange rate must be compared with a whole pattern of changes in prices. Another complicating factor is that the impact on prices of changes in the dollar's value is felt only incrementally over time. For example, the effects of a particular shift may be noticeable in three months, more acute in four months, peaking at six months, diminishing at eight months, and disappearing in twelve months. The aftermath of a change in the dollar's value moves through the rest of the economy like a wave rather than affecting it in a more easily discernible one cause, one effect pattern. To determine when and to what extent the effects of alterations in the dollar's value have emerged, when they have crested and when they have ebbed, w e have used a group of statistical techniques, time series analysis, to develop mathematical models of the lag structure. They describe, as it were, the size and shape of the wave. Other studies of the interplay between the dollar's value and domestic prices attempt to model the specific channels, such as import prices or wages, through which the dollar might affect domestic prices. In addition, these studies attempt to gauge the impact of the dollar in the absence of change in other factors, such as real gross national product and monetary policy. Our time series analysis enables us to create a model that will reflect the nature and extent of the price response to the dollar's movement regardless of how the dollar affects prices, whether through import prices, wages, or some other channel. Our approach differs from others in that it does not provide a detailed breakdown of the dollar's impact in terms of the specific channels through which the impact travels. In addition, while a few other studies attempt to provide an explanation of why the dollar moved (in terms of changes in 6 foreign fiscal or monetary policy, for example), our approach does not; w e focus solely on the observed interaction of prices and the dollar's value during the period of floating exchange rates. Our results indicate that during the period w e analyze, exchange rate movements were followed by substantial changes in the price level. If wide fluctuations in exchange rates continue and the linkage between the dollar and prices persists, then achieving price stability, which is one of the goals of monetary policy, may be difficult. To reduce the volatility of the dollar, some economists advocate a return to greater fixity of " T h e aftermath of a change in the dollar's value moves through the rest of the economy like a wave rather than affecting it in a more easily discernible one cause, one effect pattern." exchange rates, perhaps through a system of "target zones" as in the European Monetary System, which links the currencies of a number of European countries. Other economists insist that the current floating rate system would exhibit less volatility if governments would follow more stable and predictable policies. W h i l e the debate between those advocating a return to fixed rates and those committed to floating rates is beyond the scope of this paper, our results indicate that there is an important link between the dollar and the future price level which bears consideration. Measuring the Impact of Exchange Rate Changes Interest in the effect of exchange rate changes on U.S. prices has grown considerably since the breakdown of the Bretton Woods system of pegged exchange rates in the early 1970s. Near the end of OCTOBER 1986, E C O N O M I C REVIEW World War II, delegates from many nations met in the resort of Bretton Woods, N e w Hampshire, to plan an international monetary system that would promote economic growth and international trade in the post-war world. In the system which grew out of this meeting, governments intervened in the foreign exchange markets to maintain pegged exchange rates, often foryears at atime. For example, the exchange rate between the dollarand the British pound was maintained at $2.80 per pound from 1949 to 1967. The breakdown of the Bretton W o o d s system introduced a period of floating exchange rates that has continued to the present day. "If wide fluctuations in exchange rates continue and the linkage between the dollar and prices persists, then achieving price stability, which is one of the goals of monetary policy, may be difficult." As shown in Chart 1, the years of floating exchange rates have been marked by large fluctuations in both the exchange value of the dollar and the U.S. inflation rate. Not surprisingly, economists have been drawn to investigate the apparent relationship between moves in the dollar's value and U.S. prices. Prior to the collapse of Bretton Woods, analyses of U.S. inflation tended to focus solely on domestic determinants. Monetarist literature concentrated on the U.S. money supply as the source of inflation, while Keynesian literature saw the rate of inflation as a result of excess demand, which is identified with the unemployment rate in the Phillips curve framework used in most of these studies. ' This approach seemed to work well in explainingthe modest inflation of the 1960s, but it has proved inadequate since then. A common tactic in more recent studies has been to add energy or food prices as additional factors to explain inflation, based on the rationale thatthese prices have been subject to large externally generated supply shocks caused by OPEC and weather conditions. 2 Other investigators have tried to FEDERAL RESERVE BANK OF ATLANTA determine the role of exchange rates as an influence on U.S. prices. The idea that a connection exists between exchange rates and overall price levels can be traced to the theory of purchasing power parity. It states that the exchange rate between any two national currencies adjusts to maintain equality between the purchasing power of a currency at home (in terms of real goods and services) and its purchasing power abroad after conversion into the foreign currency. As a result, depreciation in the exchange rate should be associated with a proportionate increase in the ratio of domestic to foreign price levels. For example, suppose the exchange value of the dollar fell (or depreciated) by 5 percent, while foreign prices rose 2 percent. According to purchasing power parity, the ratio of U.S. to foreign prices should rise by the amount of the depreciation (5 percent). To obtain this result, the U.S. price level would have to rise by approximately 5 + 2 = 7 percent. 3 However, empirical analysis suggests that purchasing power parity is not by itself sufficient to explain price movements in recent years.4 Another strategy to account for the last decade of price fluctuations has been to modify the descendants of Keynesian models of the 1960s to include exchange rates or import prices as additional explanatory variables. This approach has been mathematically specified in a number of ways, which are reviewed in the study by P. Hooper and B. Lowrey (1979). In the single-equation method, the domestic price level is viewed as a function of labor costs (wages), demand pressure (unemployment), and import prices. The effect of exchange rate changes on the domestic price level is then inferred indirectly from statistical calculations that measure the impact of import prices, c o m b i n e d with analysis that considers the consequences of exchange rate changes for import prices. These studies, adjusted for comparability by Hooper and Lowrey, indicate that the long-run effect of a 10 percent depreciation of the dollar, with no change in labor costs or demand pressure, is a rise of 0.8 to 1.5 percent in consumer prices.5 This approach treats labor costs, demand pressure, and import prices as separate sources of inflationary pressure, so that exchange rates affect domestic price levels only through their direct effect on import prices. However, it is plausible that exchange rate changes might affect, perhaps after a lag, both labor costs and demand pressure. If this were the case, it would mean that the total 7 impact of exchange rate changes on the domestic price level might be greater than their direct effect alone would indicate. In attempting to account for some of these additional channels by which exchange rates could affect domestic prices, other studies have developed more complicated structural simultaneous equation models that incorporate exchange rate effects on labor costs and demand pressure. Studies using this approach, as adjusted for comparability by Hooper and Lowrey, estimate that a 10 percent dollar depreciation will eventually result in a 0.8 to 2.7 percent rise in consumer prices, with nearly all results below 2 percent. 6 More recently, R. Dornbusch and S. Fischer (1984) and J. D. Sachs (1985) obtain somewhat larger estimates, using different measures of exchange rates and prices. In contrast, two recent studies suggest that exchange rates have little or no effect on U.S. prices. W . T. W o o (1984) analyzes the G N P consumption deflator, a broadly based measure of consumer prices, excluding food and energy, with quarterly data from the second quarter in 1975 to the first quarter in 1984. Employing the single-equation approach, he finds that neither import prices nor the exchange rate have a significant impact on the consumption deflator excluding food and energy, once wages and oil prices are incorporated. J. E. Glassman (1985) argues that earlier studies overstate the effect of exchange rates on domestic inflation because of the strong correlation between exchange rate movements and energy price shocks. In his view, relative energy prices (and perhaps food prices) should be treated as separate explanatory variables for overall inflation. The results of his analysis, using quarterly data that cover both fixed and flexible exchange rate regimes, suggest that exchange rates have no significant effect on U.S. inflation. An Alternative Approach In ouropinion, the behaviorof inflation and the dollar's value under the flexible exchange rate system have been substantially different than under the Bretton W o o d s arrangement. Both inflation and especially the dollar have shown dramatically greater volatility since the end of Bretton Woods. W e have therefore based our estimates solely on data from the period of floating exchange rates.7 Moreover, it is generally 8 agreed that the relationship between the dollar and U.S. prices involves significant time lags, but theory provides little guidance as to the length of the lags or their pattern.8 Therefore, time series analysis is a natural approach to analyzing the data, because it is especially designed to uncover the length and patterns of lagged effects. 9 A model of the interaction between the dollar and prices based on time series analysis accounts for the impact of the dollar on prices regardless of the channel, whether through import prices, labor costs, demand pressure, or some other factor. O n the basis of our preliminary results, w e allowed for " B o t h inflation and especially the dollar have shown dramatically greater volatility since the end of Bretton W o o d s . " longerand less restrictive lag structures than those imposed in most previous studies. Time series analysis requires a sizable amount of information on the relevant variables. Nowthat the floating exchange rate regime has lasted more than a decade, it is feasible to apply these statistical methods using monthly data that allow us to extract the maximum amount of information available. The sample period begins in April 1973, afterthe floating exchange rate regime was fully in place, and ends in June 1985. For reasons consistent with the methodology, none of the data are seasonally adjusted. In creating a mathematical model of the relationship between the dollar and prices, w e tested to determine the validity of our assumption that domestic prices respond to changes in the value of the dollar. W e also had to determine if feedback existed so that changes in prices also influenced the value of the dollar. To accomplish this, w e considered the price level in any particular month as a function of its own past history plus a shock or innovation that had nothing to do with the price level's past history. In ascertaining OCTOBER 1986, E C O N O M I C REVIEW whether or not the shock or innovation in the price level could be predicted at least in part using past values of the dollar, w e found that changes in the dollar's value did anticipate changes in the price level. However, past values of the price level did not appear to help in predicting shifts in the exchange rate of the dollar. To summarize movements in the value of the dollar (e t ), w e use the Federal Reserve Board trade-weighted dollar index.10 Several price measures are examined because economic theories which distinguish between traded and non-traded goods, or between export and import goods, sug- prices. The natural logarithm of each variable is used throughout to normalize the wide range in values.11 Identifying the Dynamic Relationships Between the Dollar's Value and Prices To investigate the distributed lag relationship between the dollar and U.S. prices, w e start with the hypothesis that the price level in any month (p t ) is a weighted sum of current and past values of the exchange value of the dollar, plus a (possibly complicated) random error term, n1t:12 (1) p t = a D e t + anet_i + a2et_2 + . . . + n 1 t " A model of the interaction between the dollar and prices based on time series analysis accounts for the impact of the dollar on prices regardless of the channel, whether through import prices, labor costs, demand pressure, or some other factor." gest that exchange rates may not have a uniform link to all domestic prices. Traded goods prices are usually assumed to be highly responsive to exchange rates, while the prices of non-traded products, such as certain services and housing, are not. In the case of exports and imports, it is sometimes argued that because the United States is such a large factor in world markets, dollar prices of U.S. exports are determined by U.S. costs of production and are affected little by exchange rates, but that dollar prices of U.S. imports are somewhat more responsive to exchange rates. W e investigate the dynamic relationships between the value of the dollar and the following three measures of U.S. domestic prices (p it ): p 1 t = CPI, all items; p 2t = CPI, services; p 3 t = PPI, all finished goods. The "CPI, all items" is an overall index, while the "CPI, services" is one component of the overall CPI containing mostly nontraded items, and the "PPI, finished goods" is a proxy for traded goods FEDERAL RESERVE BANK OF ATLANTA = oo I k = 0 a k e t . k + n 1t for all t within the sample period, where the a k 's are coefficients (numbers that remain fixed throughout the sample period). The coefficients a k represent the dynamic response of prices to current and past movements in the dollar. The random error term, n qt , incorporates the variation in prices not attributable to exchange rate changes. A hypothetical example may help clarify the meaning of equation (1). Suppose that a 4 = -0.1, a 5 = -0.2, a 6 = -0.1, and all the other a k 's were zero. In this case, the dynamic response of prices to dollar movements has a fairly simple form. Suppose that the exchange rate fell (depreciated) 5 percent in December. There would be no associated price movement during December (because ay = 0) or in the first three months afterward. I n the fourth month (April), the price level would rise by (-0.1) x (-5 percent) = 0.5 percent. In May, the price level would rise by a further (-0.2) x (-5 percent) = 1.0 percent. In June, the price level would rise by a further (-0.1) x (-5 percent) = 0.5 percent; the price response would then be complete, and the total change in the price level w o u l d be approximately 2 percent. Likewise, w e consider the hypothesis that the exchange value of the dollar in any month (et) is a weighted sum of current and past values of the price level, plus its own random error term, n 2t : (2) e t = b 0 p t + pt.n + b 2 p t . 2 + . . . + n 2t CO = X b k p it . k + n 2t k = 0 where the b k 's are coefficients that remain fixed throughout the sample period. 9 Figure 1. Cross-Correlation Function [rvu(k)] Between the Dollar's Value (et) and Prices (p 1t ) a Plots rYU(k) Lag Correlation -30 -29 -28 -0.01724 -27 -26 -25 -24 -23 -22 -21 -20 -19 -18 -17 -16 -15 -14 -13 -12 -11 -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 -1 9 8 7 6 5 4 3 2 1 0 1 0.11459 0.17314 0.02221 * * * 0.07157 0.11162 0.03111 0.14187 0.06311 0.01808 -0.02976 0.01994 -0.04380 * * * * * * . * * 0.05195 0.04916 -0.03353 0.07853 0.05449 -0.06844 * * * . * * 0.00113 -0.05858 -0.05162 -0.02201 -0.09904 -0.18163 0.11752 -0.18686 -0.03005 -0.13349 -0.00768 ri n a. e. a • * * . . ** * * * * * * * * * * * * • I * 2 3 4 5 6 7 8 9 1 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 0.06681 -0.10105 0.08960 0.05591 -0.09598 -0.07684 -0.24460 0.03512 -0.12776 -0.11459 0.02794 -0.05063 -0.01678 -0.01567 -0.11829 0.04237 -0.03237 -0.03141 -0.05219 -0.07148 -0.04695 -0.02545 0.09644 0.05588 -0.13809 0.01927 0.05630 0.03816 -0.03125 0.03976 0.03378 • ** * • • • ** • * * • • • . ** * * * * * • * * * * • • * • • • ** • * • • • • . ** • • • * * • * • • • * * • * • • • * • ** • * • • • * * * • * * • * • • • • • * * « • ^ h e Fed Trade-Weighted Dollar Index is represented by e,, and Pn represents the CPI, all items. Ninety-five percent confidence intervals appear as "." in the plots. Table 1. G r a n g e r Test Results 3 H-,: the dollar (et) does not Granger-cause prices (pit) H 2 : prices (pit) do not Granger-cause the dollar (et) Price Measure H-,: et-/-pit CPI, all items (p 1t ) 1.92 (.014) 1.18 (.283) CPI, services (p 2t ) 2.05 (.008) 1.11 (.350) PPI, all finished goods (p 3t ) 2.36 (.004) 1.06 (.410) H2: PitT^^ a The variable e, represents the Federal Reserve trade-weighted dollar index. In all cases, 36 lags on the dependent variable are included. The numbers in each column are the F-statistics and (in parentheses) the marginal significance levels for the hypothesis being tested. For p, t and p2\<30 l a 9 s on the right-hand-side variable are always included, while for p 3 „ 18 lags on the right-hand-side variable are included. The longer lags were included for the consumer price measures because, a priori, w e would expect consumer prices to respond less rapidly to exchange rate (or other) shocks than producer prices; in addition, the cross-correlation functions suggested longer lags for the consumer price measures. To obtain preliminary information about these relationships, L. D. Haugh (1976) suggests scrutiny of the univariate residual cross-correlation function, which essentially shows when current movements in prices match past patterns of fluctuation in the dollar's value, and vice versa.15 If the price level (pt) and the value of the dollar (et) were not related to one another, then all the coefficients a k and b k in equations (1) and (2) would be zero. In that case, the residual crosscorrelations would be small in magnitude and randomly distributed about zero. Figure 1 shows the univariate residual crosscorrelation function using the overall CPI as a measure of prices. The first column gives the lag length, in months. Negative lag lengths involve the relationship of the dollar's exchange rate in any given month to previous fluctuations in prices, while positive lag lengths describe the price level versus previous exchange rates. The column labeled "Plots" represents the cross-correlation function; large cross-correlations which could indicate a relationship show up as several asterisks, while the smallest cross-correlations have no asterisks. The vertical line in the center of the 12 "Plots" column represents the zero axis. W h e n asterisks extend beyond the dots to the right and left of the center line (a confidence interval of 95 percent), we can say that we are "95 percent sure" that these correlations are different from zero. The bottom half of Figure 1 (positive lags) gives the cross-correlations between price residuals and past exchange rate residuals; these cross-correlations are closely related to the a k 's in equation (1). Observe that while the cross-correlations at lags 2 and 3 are mildly positive, nearly all the crosscorrelation at lags 4 to 24 are negative, with an especially large negative cross-correlation at a lag of six months. The negative sign of these crosscorrelations suggests that depreciation of the dollar (a decline in e t ) tends to be followed by an opposite movement (an increase) in prices. The top half of Figure 1 (negative lags) gives the cross-correlations between exchange rate residuals and past price residuals; these cross-correlations are closely related to the b k 's in equation (2). There is some indication of negative correlation for the early months (lags -2 to -7), with a fairly random pattern otherwise.14 The cross-correlation functions involving the exchange rate and the OCTOBER 1986, ECONOMIC REVIEW Table 2. Cumulative Response of U.S. Prices Following a 10 Percent Change in the Dollar (et) Through CPI, all items CPI, services PPI, all finished goods 12 months -.159 -.133 -.218 24 months -.336 -.390 -.417 36 months -.426 -.499 -.517 48 months -.463 -.530 -.554 co months -.485 -.542 -.571 other price series are not presented here to save space, but they show similar patterns. Formal statistical tests can further help to determine how exchange rates and prices are related. As mentioned earlier, if p t a n d e t w e r e not related to one another, then the estimated residual crosscorrelations should be small in magnitude and randomly distributed about zero. P. D. Koch and S. S. Yang (1986) provide a test that checks for the existence of a lead/lag relationship in the entire set of cross-correlations. Their test is particularly useful because it can discern patterns in the crosscorrelations. For example, suppose that none of the cross-correlations, taken individually, are significantly different from zero, but a group of crosscorrelations for adjacent lags are all fairly strongly negative. Unlike some other tests, the Koch-Yang test is designed to detect such non-random patterns in the cross-correlations and to reject the hypothesis that the variables are unrelated if the non-random pattern is sufficiently strong. W h e n applied to our data, the Koch-Yang test indicates that the hypothesis which contends that the price level and the dollar are unrelated can be rejected for all three price series.15 The test results, along with the residual cross-correlations, also suggest that the lag distributions in question are fairly long, extending over many months. Having observed an apparent relationship bet w e e n the dollar's value and domestic prices, further investigations were in order to determine whether shocks to the value of the dollar lead movements in prices, shocks to prices lead dollar movements, or both. Using the test proposed by C. W . J. Granger (1969), w e examined statistically FEDERAL RESERVE BANK OF ATLANTA whether movement in each variable, dollar and prices, could be predicted more accurately using its o w n past history plus values of the other variable, or simply on the basis of its o w n past history alone (Table 1). In determining the direction of Granger causality, w e test t w o hypotheses: H n : e t does not Granger-cause p it ; H 2 : Pit does not Granger-cause et. The resulting F-statistics and marginal significance levels are presented in Table 1. Larger values of F (smaller marginal significance levels) pointtoward rejection of the hypothesis being tested. Using standard significance levels of either .10 or .05, observe that H 2 (p t does not Granger-cause e t ) is not rejected for any price series. However, at those same significance levels, H-, (e t does not Granger-cause p t ) is rejected for all of these price series. Including past values of the dollar helped significantly to predict prices, but including past values of prices did not help to predict the dollar. The results of this test suggested that shocks to exchange rates lead price movements, but not vice versa.16 The Lag Between Dollar Movements and Price Fluctuations As discussed in the Appendix, the results of the Koch-Yang and Granger tests indicate that it is appropriate to estimate the distributed lag going from exchange rate movements to prices, as modeled in equation (1). Our estimates imply that a permanent ten percent drop in the dollar w o u l d be associated with an increase in the C P I of 13 Table 3. Selected Estimates of the Cumulative Impact of 10 Percent Depreciation of the Dollar on U.S. CPI Approach Time Series Model Koch, Rosensweig & Whitt Single-Equation Rise in CPI After One Year Long-Run Rise in CPI 1.6 4.9 Regression Spitaeller Modigliani-Papademos Dornbusch 1.5 0.4a 1.7 0.6a 1.5 2.2 Dornbusch-Krugman Simplified Structural Kwack Model 1.8 Full Structural Model Artus-McGuirk 1.8 M P S (Thurman) Berner and others Structural Model with Endogenous Exchange b Estimate 0.7 1.5 - 2.0 c,d * * * 1.5 Rate M C M (Hernandez-Cata and others) a Estimate 2.7 0.26 - 0.92 a,d a,d 0.8 - 1.5 assumes no change in oil prices. assumes no change in oil prices, as well as policy responses which prevent any changes in the path of real GNP. c Estimate is for the personal consumption expenditure deflator, rather than the CPI. dA range, rather than a point estimate, is given because the size of the impact on consumer prices depends on the reason for the exchange-rate change or on the policy reaction to the exchange-rate change. All estimates are based on depreciation of the dollar, as measured by the Federal Reserve Board trade-weighted dollar index; estimates for studies other than our own are as adjusted for comparability by Hooper and Lowrey (1979) or Glassman (1985). 14 OCTOBER 1986, ECONOMIC REVIEW approximately 1.6 percent after one year, 3.4 percent after two years, 4.3 percent after three years, and ultimately 4.85 percent. A rise in the dollar would have opposite effects (see Appendix). The distributed lags shown in Table 2 between the dollar's value (e t ) and other price indexes (p it ) also indicate substantial responses following exchange rate changes. The results for services are particularly noteworthy. Since most of the services in the CPI are not traded internationally, it seems plausible that exchange rates would have little direct effect on the prices of services. The strong response for price indexes based on services may reflect indirect effects from exchange rate changes to wages, interest rates, or aggregate demand conditions. Our results confirm that the lags in question are fairly long. For all three series, over half of the total cumulative change in prices comes more than a year following the exchange rate change, and sizable effects occur even beyond two years. Table 3 presents our results for the overall CPI and compares them with results from other studies that attempt to estimate the narrower structural link between the dollar and U.S. prices. The other studies hold constant, depending on the study, such variables as real G N P , monetary policy, or wages. Our results show by far the largest long-run rise in the CPI following a depreciation of the dollar. All of the estimates in Table 3 are subject to qualification. The various studies use data from differenttime periods; if the relationship between the dollar and prices changes through time, then estimates based on one time period might not apply to another time period. Secondly, all of the estimates are based on assumptions about the specification of the relationships in question, such as lag structure for included variables and the properties of the residual error term. In addition, the single equation regression models and the structural models require assumptions about which possible explanatory variables to include and which to exclude. Since there is no foolproof method of testing for or eliminating specification error, all of the estimates are subject to this source of error. Finally, the estimates in the table are not entirely comparable, because regression models and the structural models attemptto estimate the partial effect of the dollar on prices, holding constant other variables such as wages, monetary FEDERAL RESERVE BANK OF ATLANTA growth, or oil prices; moreover, different studies hold different variables constant. By contrast, our time series analysis attempts to estimate the price response to dollar movements during our sample period through all effective channels, without holding other variables constant. Since the end of our sample period, the dollar has depreciated considerably. If the dollar were the only determinant of inflation, our estimate would suggest that a substantial pickup in inflation is likely in the next year or two. The other studies in Table 3 suggest a significant but more modest pickup in inflation, while the estimates of W o o and Glassman would suggest no perceptible response. However, other factors suggest a more sanguine outlook for inflation. In particular, compared to the episodes of depreciation during the 1970s, the recent depreciation has coincided with a period of relatively weak economic activity in the world as a whole, as well as of weak and even declining prices of raw materials, especially oil. Nevertheless, based on our results, w e are inclined to think that the other estimates in the table, and especially those by W o o and Glassman which find essentially no price response to changes in the dollar, may understate the price response to the dollar.17 Conclusion The exchange value of the dollar has fluctuated considerably in recent years. Most previous studies of the link between the dollar and prices find that dollar depreciation has a modest but significant impact on inflation, though two recent studies suggest no impact. Our own findings indicate that during the sample period studied, exchange rate movements were followed by substantial changes in the price level. Indeed, our results suggest that, if anything, the majority of previous studies may have understated the inflationary response following dollar depreciation. Moreover, there were long lags between changes in the dollar and the associated changes in prices. If these historical patterns persist, then progress toward a policy goal of price stability may be incompatible with a continuation of wide fluctuations in the value of the dollar. 15 Figure 2. The Distributed Lag Relationship Between the Dollar (et) and Prices (p1t) A O » X % C £ 0 — Plot of Distributed L a g Coefficients 0.1 ICD IQ_ 6 S 8-1 _J w 0 i 1 3 ÌZ QJ W > Q O Months (k) -0.1 M I I I I I I I I I 1 II 0 11 I M 12 V A . £ j = 0 a I I I I I I I I 1 I I I M 24 I I I I I I I I I I I I I I I I I I I I I I i I I I I 36 48 60 S u m of Distributed L a g Coefficients i J 0.1 txSL r -0.1 - -0.2 - -0.3 - -0.4 - Months (k) -0.5 ' ' l ' I 0 'I I I I I 1I I I I 1 1 t I I I I M 12 24 I 1I I I I I 1I I I I I M 36 I I I I I I I I I I I I I I I I I I I I 48 60 The top frame shows the particular price responses at various lags, while the bottom frame shows how the cumulative price response builds through time. The response is largely complete after four years. Source: Federal Reserve Bank of Atlanta. 16 O C T O B E R 1986, E C O N O M I C R E V I E W NOTES Estimating the Distributed Lag Relationships 1See H. G. Johnson (1969) and R. J. Gordon (1970). 2See K. M. Carlson (1980), J . A. Tatom (1981), and A. S. Blinder (1979). 3 For Our results based on the univariate residual cross-correlation functions and Granger tests indicate substantive relationships in which the dollar leads prices over long distributed lags. The corresponding cross-correlation functions reveal that the first few coefficients in these distributed lags are often positive, while the next twelve to twenty-four coefficients are mostly negative. The Granger tests reveal no empirical support for the alternative hypothesis that movements in prices lead the dollar. In terms of equations (1 ) and (2), these results indicate that all of the b k 's are approximately zero but that at least some of the a ^ s are different from zero. Therefore, w e confine our attention to estimating the coefficients in equation (1). Equation (1) w a s estimated using constraints suggested by the specification test results. Leaving the contemporaneous coefficient and the first three monthly lag coefficients unconstrained (because some of the first few cross-correlations were positive), w e restricted the remaining coefficients to lie along a damped polynomial, the "modified Almon lag" proposed by P. Schmidt [Schmidt (1974), G. S . Maddala (1977) pp. 363-364], This lag structure combines an Almon lag with a Koyck lag, allowing the polynomial portion implied by the Almon lag to dominate over the first several coefficients, while the Koyck lag eventually damps the coefficients toward zero. The noise model for n 1 t is estimated simultaneously to enable consistent and asymptotically efficient estimation of equation (1 ); see Box and Jenkins (1976), pp. 388-395. For further details on the estimating procedures, s e e Koch, Rosensweig, and Whitt (1986). The estimated distributed lag coefficients, a k ; k = 0,1,2,..., fortheoverall C P I are plotted in the first frame of Figure2. In the k second frame, the cumulative sum of coefficients, I i=o aj, is plot- ted for k=0,1,2,... 60. This quantity represents the cumulative response after k months in the C P I following a sustained one percent increase in the value of the dollar. The contemporaneous and first three lagged coefficients are all positive, as suggested by the cross-correlation function, though only the coefficient at lag two months is statistically significant at the .05 level. The coefficients beyond lag three months are all negative. The cumulative sum of the coefficients turns negative beyond lag four months and grows toward an eventual sum of approximately -.485. Table 2 gives the cumulative sum of the coefficients at various lags for all our price series. For the CPI, the sum through twelve monthly lags is -.16; through twenty-four lags, -.34; through thirty-six lags, -.43; and the infinite sum converges to approximately -.485. FEDERAL RESERVE BANK OF ATLANTA a review of the literature on purchasing power parity, see Officer (1976). "^See I. B. Kravis and R. E. Lipsey (1978), and J . A. Frenkel (1981). 5 The original studies include R. Dornbusch (1978), F. Modigliani and L. Papademos (1975), and E. Spitaeller (1978). In other studies, R. Dornbusch and P. Krugman (1976) and R. J . Gordon (1982) obtain similar through larger estimates. ^These estimates typically assume no change in certain other variables which may affect the price level, such as the growth rate of the money supply or real gross national product. The original studies are J . L. Prakken (1979), S. Y. Kwack (1977), J . R. Artus and A. K. McGuirk (1981), S. S. Thurman (1977), R. Berner and others (1975), E. S. Urdang (1978), and E. Hernandez-Cata and others (1978). 7 Most of the earlier studies discussed above included at least some data from the period before generalized floating began. 8 Other studies have estimated the lagged effects of the dollar on prices, but they have usually imposed restrictive lag patterns without much justification. 9G. E. P. Box and G. M. Jenkins (1976) provide a useful introduction to time series analysis. The use of time series methods in macroeconomics is somewhat controversial; for a critical analysis see T. F. Cooley and S. F. LeRoy (1985). d e p r e c i a t i o n of the dollar corresponds to a fall in this index. W e have also employed the Morgan Guaranty dollar index in this analysis, with robust results. For more details on the construction and rationales for these and other indexes of the value of the dollar, see D. Deephouse (1985) and J. A. Rosensweig (1986). 11 The data for e t and p t often display trends and systematic seasonal movements, implying nonstationarity. This problem is addressed by detrending and deseasonalizing each series with a time trend and eleven seasonal dummies in the regression models. The results may then be interpreted as the relationships between the dollar and prices, abstracting from systematic movements due to trend and seasonality. W e have also estimated the relationship between e, and p 1 t using two other common methods to account for the nonstationarity implied by trend and seasonality: (i) take the first log difference of each variable and include eleven seasonal dummies in the distributed lag model; (ii) take a first and a twelfth difference on the log of the data, before estimating the time series model. W e prefer to employ the levels of e, and p lt rather than their first or twelfth differences, since it is the relationship between their levels that concerns us. Hence, the method outlined in the text is preferred to those listed here. The distributed lag relationships are found to be robust, regardless of the method used. 12 ln particular, the random error term is not restricted to be a white noise process. In the case of the estimated model for p, t , the best model for the random error term is a fairly complicated MA process. 13 To do this, univariate Box-Jenkins models are first estimated separately for p, and e t ; the residuals from these models are then calculated; and finally, the cross-correlation function between the two sets of residuals is obtained. 14 The line for lag zero in the middle of Figure 1 gives the contemporaneous correlation, which is not significantly different from zero. 15 For a more detailed description of the Koch-Yang test and its results for these data, see P. Koch, J . A. Rosensweig, and J. Whitt (1986). 17 16 Granger's definition of "causality" differs from the traditional philosophical concept in that it is purely predictive (Granger, 1969). Throughout this study, the term refers to Granger causality. 1 7 Such understatement might arise in single-equation or structural models because significant channels through which the dollar affects prices are erroneously omitted from a model, or perhaps Artus, J . R., and A. K. McGuirk. "A Revised Version of the Multilateral Exchange Rate Model," IMF Staff Papers, vol. 28 (June 1981), pp. 275-309. Berner, R., and others. "International Sources of Domestic Inflation," in Studies in Price Stability and Economic Growth, Papers 3 and 4, International Aspects of Recent Inflation, Joint Economic Committee, 95 Cong. 1 Sess. Washington: Government Printing Office, 1975, pp. 1-41. Blinder, A. S. Economic Policy and the Great Stagflation. New York: Academic Press, 1979. Box, G. E. P., and G. M. Jenkins. Time Series Analysis, Forecasting and Control. San Francisco: Holden Day, 1976. Carlson, K. M. "The Lag from Money to Prices," Federal Reserve Bank of St. Louis, Review, vol. 62 (October 1980), pp. 3-10. Cooley, T. F., and S. F. LeRoy. "Atheoretical Macroeconomics: A Critique," Journal of Monetary Economics, vol. 16 (November 1985), pp. 283-308. Deephouse, D. "Using a Trade-Weighted Currency Index," Federal Reserve Bank of Atlanta, Economic Review, vol. 70 (June/July 1985), pp. 36-41. Dornbusch, R. "Monetary Policy Under Exchange-Rate Flexibility," in Managed Exchange-Rate Flexibility: The Recent Experience. Federal Reserve Bank of Boston Conference Series no. 20, October 1978. Dornbusch, R., and S. Fischer. "The Open Economy: Implications for Monetary and Fiscal Policy," National Bureau of Economic Research, Working Paper 1422, 1984. Dornbusch, R., and P. Krugman. "Flexible Exchange Rates in the ShortRun," Brookings Papers on Economic Activity, no. 3,1976, pp. 537575. Frenkel, J. A. "The Collapse of Purchasing Power Parities During the 1970s," European Economic Review, vol. 16 (May 1981), pp. 145165. Geweke, J . "Testing the Exogeneity Specification in the Complete Dynamic Simultaneous Equation Model," Journal of Econometrics, vol. 7 (April 1978), pp. 163-185. "Comparison of Tests of the Independence of Two Covariance-Stationary Time Series," Journal of the American Statistical Association, vol. 76 (June 1981), pp. 363-373. Glassman, J . E. "The Influence of Exchange Rate Movements on Inflation in the United States," Board of Governors of the Federal Reserve System, Working Paper Series no. 46, April 1985. Gordon, R. J. "Inflation, Flexible Exchange Rates and the Natural Rate of Unemployment," in M. N. Baily, ed., Workers, Jobs and Inflation. Washington, D.C.: The Brookings Institution, 1982. "The Recent Acceleration of Inflation and Its Lessons forthe Future," Brookings Papers on Economic Activity, no. 1,1970, pp. 841. Granger, C. W. J . "Investigating Causal Relations by Econometric Models and Cross-spectral Methods," Econometrica, vol. 37 (July 1969), pp. 424-438. Haugh, L. D. "Checking the Independence of Two Covariance Stationary Time Series: A Univariate Residual Cross-Correlation Approach," Journal of the American Statistical Association, vol. 71 (June 1976), pp. 378-385. Hernandez-Cata, E., and others. "Monetary Policy Under Alternative Exchange-Rate Regimes: Simulations with a Multi-Country Model," in Managed Exchange Rate Flexibility: The Recent Experience, Federal Reserve Bank of Boston Conference Series, no. 20, October 1978. Hooper, P., and B. Lowrey. "Impact of the Dollar Depreciation on the U.S. Price Level: An Analytical Survey of Empirical Estimates," Board of 18 because of measurement error. For instance, if import prices are measured with error, then in a regression model using import prices as one of the variables explaining overall inflation, the estimated coefficient on import prices should be biased toward zero, leading to an understatement of the estimated impact of the dollar on overall inflation. Governors of the Federal Reserve System, Staff Study no. 103, April 1979. Johnson, H. G. "A Survey of Theories of Inflation," in his Essays in Monetary Economics, 2nd ed. London: George Allen and Unwin, 1969. Koch, P. D., and J . F. Ragan, Jr. "Investigating the Causal Relationship Between Quits and Wages: An Exercise in Comparative Dynamics," Economic Inquiry, vol. 24 (January 1986), pp. 61 -84. Koch, P. D., J. A. Rosensweig, and J . A. Whitt, Jr. "The Dynamic Relationship Between the Dollarand U. S. Prices: An Intensive Empirical Investigation," Federal Reserve Bank of Atlanta, Working Paper 86-5, July 1986. Koch, P. D., and S. S. Yang. "A Method for Testing the Independence of Two Time Series that Accounts for a Potential Pattern in the CrossCorrelation Function," Journal of the American Statistical Association, June 1986. Kravis, I. B., and R. E. Lipsey. "Price Behavior in the Light of Balance of Payments Theories," Journal of International Economics, vol. 8 (May 1978), pp. 193-246. Kwack, S. Y. "Price Linkages in an Interdependent World Economy: Price Responses to Exchange Rates and Activity Changes," in J . Popkin, ed., Analysis of Inflation 1965-1974. Cambridge, Mass.: Ballinger Pub. Co. for the National Bureau of Economic Research, 1977. Maddala, G. S. Econometrics. New York: McGraw-Hill, 1977. Modigliani, F., and L. Papademos. "Targets for Monetary Policy in the Coming Year," Brookings Papers on Economic Activity, no. 1,1975, pp. 141-166. Officer, L. H. "The Purchasing-Power Parity Theory of Exchange Rates: A Review Article," International Monetary Fund Staff Papers, vol. 23 (March 1976), pp. 1-60. Prakken, J . L. "A Quarterly Model of Linkages Between the Exchange Rate and Domestic Prices and Wages," paper presented at the Midwest Economic Association, April 1979. Rosensweig, J. A. "A New Dollar Index: Capturing a More Global Perspective," Federal Reserve Bank of Atlanta, Economic Review, vol. 71 (June/July 1986), pp. 12-22. Sachs, J. D. "The Dollar and the Policy Mix: 1985," Brookings Papers on Economic Activity, no. 1, 1985, pp. 117-185. Schmidt, P. "A Modification of the Almon Distributed Lag," Journal of the American Statistical Association, vol. 69 (September 1974), pp. 679681. "The Small Sample Effects of Various Treatments of Truncation Remainders in the Estimation of Distributed Lag Models," The Review of Economics and Statistics, vol. 57 (August 1975), pp. 387389. Spitaeller, E. "An Analysis of Inflation in the Seven Main Industrial Countries, 1958-1976," IMF Staff Papers, vol. 25 (June 1978), pp. 254277. Tatom, J . A. "Energy Prices and Short-Run Economic Performance," Federal Reserve Bank of St. Louis, Review, vol. 63 (January 1981), pp. 3-17. Thurman, S. S. "The International Sector" and "The Price Sector," Board of Governors of the Federal Reserve System, Mimeo, April 1977. Urdang, E. S. "An International Flow of Funds Model with an Endogenous Exchange Rate," University of Pennsylvania, Mimeo, March 1978. Woo, W. T. "Exchange Rates and the Prices of Nonfood, Nonfuel Products," Brookings Papers on Economic Activity, no. 2,1984, pp. 511 530. Zellner, A., and F. Palm. "Time Series Analysis and Simultaneous Equation Econometric Models," Journal of Econometrics, vol. 2 (January 1974), pp. 17-54. O C T O B E R 1986, E C O N O M I C R E V I E W Nonbank Activities and Risk Non bank subsidiaries have become an increasingly attractive investment choice for bank holding companies, but regulators are concerned about the possibility of new risks. Larry D. Wall The number of bank holding companies (BHCs) with nonbank affiliates has grown dramatically since 1976, especially among BHCs with assets in excess of $10 billion. Whatever the risks involved in nonbank activities (such as mortgages, consumer financing, and data processing), bankers apparently see potential for gain. Regulators, however, must watch the financial condition of BHCs closely because of the effect the parent company can have on its bank subsidiaries. Financial problems at the bank holding company level can endanger affiliated banks. How do nonbank affiliates affect the stability and profitability of BHCs? Existing regulation of BHC activities and regulators' inquiries into further risk-based capital regulation assume that nonbanks could increase an organization's riskiness, but the relationship between nonbank activities and increased risk is by no means clear-cut. Even though existing nonbank subsidiaries may be riskier than their bank affiliates, this does not The author is a financial Research Department. FEDERAL RESERVE BANK OF ATLANTA economist in the Atlanta Fed's 19 necessarily imply that the nonbank subsidiaries increase the riskiness of the holding company. It is possible that gains from diversification of an organization's portfolio, especially geographic diversification, could actually reduce B H C risk. In either case, the potential effect of nonbank subsidiaries on BHC risk may have important implications for the regulation of BHCs. If nonbank activities make BHCs riskier, then regulators might want to require BHCs with nonbank subsidiaries to hold higher levels of capital. Currently both banks and BHCs must maintain primary capital equal to 5.5 percent of their assets.1 Since not all assets are equally risky and not all the risks taken by the bank are reflected on its balance sheet, the Board of Governors of the Federal Reseive System has proposed supplemental guidelines that include adjustments for the riskiness of on- and off-balance sheet items.2 Although this proposal does not differentiate between bank and nonbank subsidiaries of the BHC, the Board recognizes that nonbank affiliates could add to an organization's risk and recently requested comments on a proposal to establish a new risk category that would " t a k e account of the higher risks associated with certain nonbanking activities "3 If, on the other hand, nonbanks reduce B H C risk, then perhaps regulations limiting the type of activities that can be undertaken by BHCs could be relaxed. Current regulations confine the domestic involvement of BHCs' nonbank subsidiaries to activities closely related to banking. Practically speaking these nonbanks are restricted to the same kind of activities allowed to banks. In spite of these restrictions on nonbank activities, nonbank subsidiaries may be reducing B H C risk by providing greater geographic distribution than is permitted for banks. Proponents of deregulation contend that diversification into currently prohibited activities would reduce risk by allowing even greater diversification.4 Opponents contend that many activities considered for deregulation, such as investment banking, are far riskier than commercial banking and would tend to increase the riskiness of BHCs. 5 This study finds that the importance of nonbank subsidiaries for BHCs increased markedly during the period from 1976 to 1984 (Chart 1). M o r e BHCs invested in nonbank subsidiaries, especially those BHCs with assets in excess of $10 billion. These results also show that nonbank subsidiaries are generally less profitable than their 20 banking affiliates. Indications of lower profitability combined with other studies' findings that nonbank subsidiaries are riskier may seem to suggest that nonbank subsidiaries weaken the performance of BHCs. However, this conclusion must be balanced by consideration of the effect of nonbank activities on diversification. Two recent studies find that on average nonbank subsidiaries may tend to reduce the riskiness of BHCs slightly. Another study suggests on the contrary that investment in nonbank subsidiaries is associated with increased B H C risk. In the analysis presented here, however, the effect of nonbank subsidiaries on overall B H C profitability as measured by consolidated return on equity appears to be small. This research also concludes that while nonbank subsidiaries may be associated with increased risk, this does not imply that nonbank subsidiaries cause BHCs to be riskier. Even given the correlation between nonbanks and risk, it is possible that nonbanks have no effect on BHC risk or actually reduce risk. The implications of this andother studies is that nonbank subsidiaries should be considered in risk-based capital regulation. However, riskbased capital regulation should take account not only of the stand-alone riskiness of nonbank activities but also their effect on B H C diversification. These findings are neutral for deregulation. Although nonbank subsidiaries do not significantly increase risk, neither do they decrease it sufficiently to support deregulation. W h y the Concern About B H C Risk? Bank holding companies are not regulated out of concern forthe financial condition of BHCs per se but rather because a B H C s financial problems can affect its subsidiary banks. Not only is a subsidiary influenced by the overall financial condition of its parent BHC, but it can also be exposed to losses if other affiliates of its B H C fail. O n e way to protect subsidiary banks would be to impose regulations that would prevent banks from sharing in the losses of their affiliates. Then, since the financial problems and failures of its nonbank affiliates would not influence the financial condition of other bank subsidiaries, regulations designed to protect the safety and soundness of the BHC, such as capital regulation and limits on permissible activities, could be eliminated, according to some analysts.6 Others claim that even though the current system attempts to protect banks from problems in OCTOBER 1986, E C O N O M I C REVIEW their nonbank affiliates through regulations that limit a bank's ability to transfer resources to B H C affiliates, including limits on dividends and loans to affiliates, additional restrictions may be needed. According to Robert Eisenbeis (1983) and Larry Wall (1984), although existing guidelines reduce a BHC's ability to use its banking subsidiaries to help nonbank affiliates, regulations may not be sufficient to prevent a transfer of resources. Banks may, for example, attempt to aid their financially troubled B H C parent or their nonbank affiliates in order to protect the good name of the bank, or because the management of the BHC controls the management of its subsidiary banks. (Often times the same individuals manage the B H C and the bank.) Even if a bank could not transfer any resources, it would still be vulnerable to the failure of its nonbank affiliates if it depended on them for vital services or if the affiliate's failure impaired the public's image of the bank.7 Anthony Cornyn and others (1986) review the evidence and contend that both B H C managers and the public view BHCs as integrated entities with the health of each subsidiary depending on the overall condition of the BHC. Eisenbeis suggests that the only way to protect banking subsidiaries completely from problems in their nonbanking subsidiaries would be to require the BHC to operate as a passive portfolio manager. He notes that doing so would, however, eliminate the advantages of banks' affiliation with BHCs. 8 Chart 1. Growth in the Number of BHCs with Positive Investment in Nonbank Subsidiaries ^447 / • f 406 340 322 9 312 ><•305 Nonbank Risk in a Larger Context ^ 2 9 8 The most obvious way of analyzing the effect of nonbank subsidiaries on the stability of BHCs and their banking subsidiaries is to examine the activity's expected return and the variability of its return. However, this simple approach to risk measurement can be highly misleading unless banks are financially isolated from their nonbank affiliates. If transactions between bank and nonbank subsidiaries are permitted, then the riskiness of an activity should be analyzed in terms of its effect on the entire BHC's expected return and variability of return. An activity that, in isolation, appears to be relatively chancy may substantially reduce a BHC's riskiness when the risk position of the entire organization is taken into consideration. 285 Diversification. Diversification into a variety of holdings or over a broad geographic area can defuse risk, even when some of the individual FEDERAL RESERVE BANK OF ATLANTA 1976 1978 1980 1982 1984 Source: Federal Reserve Bank of Atlanta. holdings themselves are risky. The misleading nature of activity-by-activity analysis can be illustrated with a simple example. Suppose that regulators are considering allowing a B H C to invest in three different assets: A, B, and C. Each of the assets could yield one of six equally likely rate of return outcomes, as presented in Table 1. For example, a one-in-six probability exists that outcome 2 will occur, in which case asset A would yield a return of 0 percent, asset B a 9 percent 21 Table 1. Possible Gains from Diversification of Return on Assets c Portfolio of A and B Portfolio of all 3 A B_ 1 -4.0 2.0 14.0 -1.0 4.0 2 0.0 9.0 4.0 4.5 4.3 3 3.0 -5.0 14.0 -1.0 4.0 4 7.0 9.0 -6.0 8.0 3.3 5 7.0 -2.0 8.0 2.5 4.3 6 7.0 9.0 -8.0 8.0 2.7 Expected Return 3.3 3.7 4.3 3.5 3.8 Standard Deviation of Return 4.2 8.7 3.7 0.6 Source: Federal Reserve Bank of Atlanta. return, and asset C a 4 percent return. The comparative merits of these assets can be examined by looking at their profitability and riskiness. The profitability of each asset is measured by its expected return and the riskiness is measured by the standard deviation of its returns. (Higher values for the standard deviation suggest greater risk.) Assets A and B are relatively low risk assets, but B offers a slightly higher expected return for a little more risk. Asset C is substantially riskier and has a somewhat higher expected return than either asset A or B. If a banking organization were permitted to invest in only one asset, then asset C would be the least desirable since it has a much higher standard deviation and only a slightly larger expected return. However, asset C could be highly desirable if a BHC were investing in a portfolio of assets. The returns from a portfolio with equal combinations of assets A and B are presented in the fourth column. This limited portfolio results in expected returns that are midway between the returns of assets A and B and a standard deviation lower than either asset A or asset B taken singly. 22 However, a portfolio of assets A, B, and C formed with equal weights on all three assets produces even better results. The expected return on the three-asset portfolio is slightly greater than on the two-asset portfolio, and the standard deviation of returns on the three-asset portfolio is substantially lower—0.6 percent versus 3.7 percent. Thus, if our hypothetical banking organization is allowed to invest in more than one asset, then C should probably be permitted. The effect of diversification is demonstrated more generally in Figure I. 9 The vertical axis on Figure 1 represents an asset's expected return, and the horizontal axis measures the standard deviation of the return. The dots in Figure 1 represent different assets, such as bank assets or the sort of stocks and bonds individuals might invest in. If an investor purchased only one asset, he could expect the riskand return associated with the individual dot he chooses. However, if the investor purchased a portfolio of several assets, then the potential combinations of risk and return would include all the combinations in the circle. O C T O B E R 1986, E C O N O M I C R E V I E W Figure 1. Effect of Portfolio Diversification Figure 2. Benefits of Portfolio Expansion Through diversification an investor expands the options for combinations of risk and returns from one dot (representing a single asset) to all the dots or combinations in the circle. Investors seeking the maximum return for the minimum risk will select a point on the upper lefthand side of the circle, the efficient frontier, as these points are superior to other alternatives. The effect of expanding the list of assets available to an investor can be seen in Figure 2. The inner circle represents the choices available when tight controls are placed on permissible assets; the outer circle represents the choices when the list of permissible assets is expanded. The exact placement of the two circles depends on the particular assets under consideration. However, expanding the list of permissible assets can never shrink the circle, because investors can always choose to ignore assets on the expanded list. Figure 2 illustrates that wider asset selection is always preferable. Suppose for instance that an investor initially constrained to the inner circle chose his assets so that he would obtain the risk and return at point a. W h e n the list of assets is expanded, the investor can choose to maintain the same return while reducing his risk through diversification (point b) or he can maintain the same risk and receive a higher return through diversification (point c). Management Influences. Portfolio theory is inadequate by itself to determine the effect of BHCs, according to Stephen A. Rhoades (1985). The discussion of diversification implied that the efficient frontier is dictated solely by the restrictions on BHC assets, but in fact the way assets are managed significantly affects the location and FEDERAL RESERVE BANK OF ATLANTA Wider asset selection is almost always preferable because it expands the permissible combinations of expected returns and risk. shape of the efficient frontier. Furthermore, management's choice of a portfolio will determine which point on the new efficient frontier an organization selects—one with less risk or greater profit—when the list of potential assets is expanded. Incompetent management and weak internal controls can increase the level of risk for any given return, while good management with strong internal controls can decrease the risk for any given level of returns. Management's strategy, as well as its quality, can influence the efficient frontier. Some managers may choose to run a tightly controlled organization to promote cooperation among its subsidiaries, while other BHCs may allow their subsidiaries to operate with relative independence so that they can respond to conditions in their individual markets. W h i c h strategy will produce the most desirable efficient frontier is debatable; but, in any event, the location of the frontier is likely to change depending on management's strategy. Figure 2 demonstrates how management can control the location of its organization on the efficient frontier. Suppose that a B H C is at point a 23 prior to deregulation. After deregulation it could move to point b, which has lower risk forthe same return, or to point c, which has the same risk with higher return, or to some point in between. However, there is no requirement that the B H C lie somewhere between points b and c on the new frontier. Management could choose a portfolio that lies at point d on the curve, a position with greater returns and greater risks, or point e, with lower returns and lower risks. W h e t h e r or not management would use nonbank subsidiaries to increase a BHC's risk exposure is an empirical question. Regulators are concerned that because FDIC deposit insurance removes some of the negative incentives to increased risk-taking, management will use nonbank subsidiaries to increase risk exposure. However, nonbank subsidiaries are not the only way to increase risk; if managed properly (or improperly) traditional banking activities, too, can be extremely risky.10 For example, banks can already make or lose substantial sums of money in commercial lending. H o w Profitable are Nonbanks? Before considering the impact of nonbanks on B H C risk, it is important to consider the profitability of nonbank subsidiaries and their effect on the financial condition of parent BHCs. A look at the data from 1976 to 1984 shows that investment in nonbanks has grown dramatically among BHCs with assets greater than $10 billion. Even though the percent invested in nonbanks has shrunkamongsmaller BHCs—those with assets of under $1 billion and those with assets between $1 billion and $10 billion—the number of BHCs with nonbank subsidiaries has increased across the board (Table 2).11 This suggests that for whatever reason, BHCs find nonbanks attractive. The analysis in this study indicates that nonbank subsidiaries are more profitable for BHCs with assets over $10 billion. Examining the extent of B H C investment in nonbanks and the profitability of bank and nonbank subsidiaries, as well as parent BHCs, may offer some insight into the role of nonbanks in B H C investment strategies. Extending previous studies, this research includes all BHCs with positive investments in nonbanks, examines time trends, and splits the sample into three subsamples based on B H C size.12 B H C involvement in nonbank activities is measured in two ways: first, the number of BHCs 24 with positive investment in nonbanking subsidiaries and, second, median B H C investment in nonbanking subsidiaries as a percentage of total B H C investment in subsidiaries. The number of BHCs with positive investment in nonbank subsidiaries is increasing for all three size categories. BHCs with consolidated assets of less than $1 billion showed the largest increase in the number of BHCs with nonbank subsidiaries. The number of nonbank subsidiaries and BHCs with consolidated assets in excess of $10 billion doubled over the nine-year period of the study. The numbers in Table 2 cannot be used to determine the number of BHCs starting nonbank activities in each of the three size categories. The change in the total numberof BHCs with nonbank activities does not reflect the total number of BHCs starting nonbank subsidiaries, because the acquisition of one B H C with nonbank affiliates by another with nonbank subsidiaries reduces the number of BHCs reporting nonbank activities. In the smallest size category, the change in the number of BHCs with nonbank activities is also less than the number of small BHCs starting nonbank activities, because some BHCs with nonbank activities shifted size categories as a result of asset growth. Similarly, in the case of BHCs with assets in excess of $10 billion, the change may be greater than the number of BHCs starting nonbank operations because some of the increase reflects the fact that smaller BHCs have changed categories due to asset growth. Median parent BHC investment in nonbank subsidiaries as a proportion of total parent investment in subsidiaries varies across the different size categories in recent years (Table 3).13 Median proportion of investment in nonbank subsidiaries by all three size categories was small in 1976 with the largest BHCs having the smallest proportion. W h i l e the median proportion has fallen since 1976 for BHCs with assets below $10 billion, it has increased dramatically forthe largest BHCs, jumping from 1.8 percent in 1976 to 12.2 percent in 1984. The returns for nonbank subsidiaries, banking subsidiaries, and consolidated BHCs are provided for each of the nine years in Tables 4-6. Before examining the results, however, several points need to be made about the data. First, any study that uses accounting data from corporations owned by other corporations must implicitly assume that the reported income is an accurate reflection of economic values and that the parent company is not manipulating interaffiliate transfer OCTOBER 1986, E C O N O M I C R E V I E W Table 2. Number of BHCs with Positive Investment in Nonbank Subsidiaries Consolidated Assets Year Below $1 Billion $1 Billion$10 Billion Over $10 Billion 1976 153 117 15 1977 159 123 16 1978 155 130 20 1979 154 137 21 1980 158 139 25 1981 172 142 26 1982 197 154 27 1983 206 172 28 1984 244 172 31 Source: Federal Reserve Board of Governors. Table 3. BHC Parent Investment in Nonbanking Subsidiaries as a Percentage of Total Parent Investment in Subsidiaries for BHCs with Positive Investment in Nonbanks Consolidated Assets Year Below $1 Billion $1 Billion$10 Billion 1976 3.0 4.0 1.8 1977 2.4 3.4 2.5 1978 2.5 2.4 4.7 1979 2.7 2.9 5.7 1980 2.4 2.4 3.9 1981 2.6 2.1 3.2 1982 2.4 2.4 8.4 1983 2.3 2.2 11.1 1984 1.6 2.3 12.2 Source: Federal Reserve Board of Governors. FEDERAL RESERVE BANK O F ATLANTA 25 Over $10 Billion Tables 4-6. Median Return on Equity for B H C s With Positive Investment in Nonbanking Subsidiaries Table 4. For BHCs With Consolidated Assets Below $1 Billion Year Bank Subsidiaries 1976 11.0 1977 11.4 1978 1979 1980 1981 1982 1983 8.0 Consolidated BHC 11.2 9.1 12.6 11.5 10.0 12.6 14.0 7.6 13.3 9.3 12.9 12.7 5.8 12.8 4.9 12.0 12.7 1984 Nonbank Subsidiaries 12.0 6.9 5.2 13.2 11.6 11.7 11.5 Table 5. For BHCs with Consolidated Assets Between $1 Billion and $10 Billion Year Bank Subsidiaries 1976 11.4 1977 11.1 1978 1979 13.3 7.7 1981 13.7 13.5 12.6 1984 7.7 9.7 13.6 1983 7.0 12.4 1980 1982 Nonbank Subsidiaries 13.5 Consolidated BHC 11.2 11.5 12.2 12.9 8.4 13.5 10.4 13.3 13.7 10.3 11.0 13.0 12.3 13.1 Table 6. For BHCs with Consolidated Assets Over $10 Billion Year Bank Subsidiaries 1976 10.5 1977 1978 1979 1980 1981 1982 1983 1984 Nonbank Subsidiaries 6.9 Consolidated BHC 10.8 11.0 11.2 13.0 11.8 8.3 13.3 13.6 13.6 13.0 12.0 12.0 11.5 9.3 7.9 12.8 12.3 11.1 10.2 14.4 14.2 13.8 12.6 12.3 11.8 Source: Federal Reserve Board of Governors. 26 O C T O B E R 1986, E C O N O M I C R E V I E W pricing to shift reported income from one subsidiary to another.14 This assumption is not subject to directtesting, because the fair market value of interaffiliate transactions is not observable. If shifting does, however, bias the reported figures of BHC subsidiaries, it may be toward lower bank riskand highernonbankrisk, since bank regulators are more concerned about the safety of bank subsidiaries than that of nonbank ones. Second, any study that uses consolidated B H C figures will reflect not only the results of operations among the bank and nonbank subsidiaries, but also those of the B H C parent. The profitability of services provided by the B H C parent must be considered as a relatively minor influence. BHC parents typically place most of the service functions in their subsidiaries so that the parent has few employees and minimal assets. A more significant influence is B H C double leverage practices: some parent BHCs have a greater investment in their subsidiaries' equity capital than they have equity capital of their own, and they fund the excess with debt issues. This double leverage results in both reduced net income and reduced equity for the consolidated BHC. A third consideration is that a subsidiary and BHC parent may report different figures for income and equity in cases in which the subsidiary is accounted for by the BHC parent through the purchase method of accounting. Then the income will generally be lower and the equity greater on the books of the parent BHC than on the books of the subsidiary for several years after the acquisition. The purchase method of accounting is required when one company acquires another and the transaction is paid for primarily with cash or debt. This method is not used for de novo subsidiaries created by the BHC, nor is it allowed for acquisitions financed exclusively by stock of the acquiring organization. The direction and extent of this distortion is hard to measure since it depends on how many subsidiaries each B H C has acquired, how much they are worth, and on the method used to finance the acquisitions. This consideration is an important one for any study that calculates return on investment using data from the parent company. Given these caveats, an examination of the data shows that the median return on parent investment in nonbank subsidiaries is consistently below that of the return on banking subsidiaries for the smallest BHCs (Table 4). In some cases, such as 1982, the return on the nonbank subsidiaries was less than half that of the banking subsidiaries. Furthermore, the lower median returns FEDERAL RESERVE BANK OF ATLANTA for nonbank subsidiaries are reflected in median returns that are lower for the consolidated BHCs than for BHCs investing in bank subsidiaries. One possible explanation for the results in the category for the smallest BHCs is that nonbank subsidiaries are inherently less profitable. This explanation implies that the proportion of B H C investment in nonbanking activities is falling because small organizations are reducing their exposure to less profitable activities. However, the increase in the number of small BHCs with nonbank subsidiaries appears to contradict this hypothesis, suggesting that smaller BHCs find nonbank activities desirable (Table 2). An alternative reason for the low profitability could be that many small banking organizations are opening de novo nonbank subsidiaries. Then the low profitability of the de novo subsidiaries would be reducing the median profitability of all the subsidiaries, and the small size of the new subsidiaries would be reducingthe median proportion of BHC investment in nonbanking subsidiaries. Another explanation is that many of the new nonbank subsidiaries may have been organized to provide services to their B H C affiliates. In this case the parent organization might not be concerned about the profitability of the nonbank activities, since the nonbank subsidiary's profits would be earned at the expense of the other subsidiaries. For the intermediate size B H C (consolidated assets between $1 billion and $10 billion), median nonbank subsidiaries appear to be somewhat less profitable than the median banking subsidiaries. The gap is narrower in the 1980s, though, than it was in the 1970s (Table 5). Nonbank subsidiaries helped make the median return on equity of consolidated BHCs somewhat lower than the median return for the BHCs' banking subsidiaries. The nonbank subsidiaries of the largest BHCs have been the most successful (Table 6). Although their median return is generally below that of the banking subsidiaries, the difference is less than in either of the two smaller size categories, and the return on equity for the consolidated BHC is above that of its banking subsidiaries. This result could, as previously mentioned, reflect double leveraging practices by the BHCs. Nonbanking subsidiaries of BHCs with total assets below $1 billion, then, reported much lower returns than did their banking affiliates. The nonbank subsidiaries of BHCs with assets above $1 billion did betterand the nonbank subsidiaries of BHCs with assets greater than $10 billion came closest to matching banking affiliates. 27 Determining Risk: Some Prior Studies W h i l e the potential benefits of BHC diversification into currently permissible activities and impermissible activities has been demonstrated by several studies, the actual effect of existing nonbank subsidiaries has received less attention. 15 A look at some prior studies of risk helps put the research presented here into perspective. Examining bond market reactions to bank acquisitions of discount brokers, Wall and Eisenbeis (1984) found that bond returns were not significantly affected by B H C diversification into discount brokerage. B H C diversification into discount brokerage, then, does not significantly affect the market's perception of the riskiness of a BHC. However, the Wall and Eisenbeis study is limited in that it only considers BHC diversification into discount brokers and does not analyze performance after acquisition. To analyze the effect of nonbank subsidiaries on the risk of capital impairment among BHCs (defined as the probability that cash flows from operations will be less than debt service costs), David R. Meinster and Rodney D. Johnson (1979) developed a special methodology and applied it to determine the impact of nonbank subsidiaries on First Pennsylvania Corporation and Philadelphia National Corporation over the period from the first quarter of 1973 to the third quarter of 1977. They found that nonbank subsidiaries reduced the riskiness of the BHCs they analyzed, but that the financial practices of BHCs caused the consolidated organizations to be slightly riskier than their banking subsidiaries.16 Meinster and Johnson provide a comprehensive examination of the effect of nonbank subsidiaries on BHC risk, but it applies to only two BHCs. This study also depends on the assumption that BHCs do not shift income between bank and nonbank subsidiaries. O n e study finds that BHCs are less risky than their subsidiary banks. Robert E. Litan (1985) examined the mean and coefficient of variation of after-tax earnings as a percentage of assets for 31 large banking organizations over the period from 1978 to 1983 to measure the riskiness of a firm's earnings.17 He found that the bank holding companies had higher mean returns and lower coefficients of variation of returns—implying lower risk—than their banking subsidiaries. However, Litan notes that the bank holding companies are less risky than their bank subsidiaries for only 16 of the 31 banking organizations in his sample. 28 Litan worked with a largersample than Meinster and Johnson, but his sample was still a relatively small proportion of the total number of BHCs with nonbank subsidiaries. (The sample does, however, include the holding companies that are likely to have the largest nonbank subsidiaries.) Litan shares with Meinster and Johnson the implicit assumption that the reported income of banking subsidiaries is an accurate reflection of economic income, and his results may also reflect B H C parent activities and double leverage practices. In a study that examined the risk of failure (defined as losses in excess of capital) for banking subsidiaries, nonbanking subsidiaries, the combination of bankand nonbankingsubsidiaries,and the consolidated B H C over the period from 1976 to 1984, Larry D. Wall (1986) found that nonbank subsidiaries did not increase the riskiness of BHCs and may have caused slight reductions in their riskiness. All BHCs that had nonbank subsidiaries at least six of the nine sample years were included in the analysis. Using data on BHC parent return on investment in subsidiaries for the B H C parent, the study found that nonbank subsidiaries by themselves are indeed much riskier than the banking subsidiaries. However, nonbank subsidiaries provided diversification benefits such that combining bank and nonbank subsidiaries caused a small, statistically insignificant, reduction in the risk of failure on average. Using a contingency table approach to provide statistically significant evidence that nonbank subsidiaries are risk-moderating, the study indicated that nonbank subsidiaries tended to decrease the riskiness of high-risk BHCs and increase the riskiness of low-risk organizations.18 Although it is more comprehensive than the other studies, the Wall (1986) study, like the previous two, assumes that reported income is not distorted by interaffiliate transactions. In the Wall study, however, interaffiliate transactions are assumed to bias the results towards nonbanks increasing risk, whereas all three studies using accounting data found evidence for a slight risk reduction. Wall's results may also be influenced by B H C parent activities and double leverage practices. The study uses subsidiary income and equity as repotted by the BHC parent, and the figures could differ from those recorded on the subsidiaries' accounting records. In a study that seems to contradict Wall's 1986 research, John H. Boyd and Stanley L. Graham (1986) show that the risk of failure for a BHC is OCTOBER 1986, E C O N O M I C REVIEW significantly positively related to its involvement in nonbank activities over the period 1971 to 1977. Their sample consisted of all domestic BHCs with total assets exceeding $5 billion at the end of 1983, except for six deleted due to missing data or special Federal Deposit Insurance Corporation involvement during the sample period. The study obtains its results by gauging a BHC's zscore (the sum of expected return on assets and the capital-to-asset ratio divided by the standard deviation of return on assets) against two measures of nonbank activity, the BHC's debt-toassets ratio and the log of its total assets. Neither measure of nonbank activity matched exactly with the percentage of B H C assets devoted to nonbank activities, but both are highly correlated with B H C investment in nonbank subsidiaries as a percentage of total investment in subsidiaries over the 1976 to 1983 period. Boyd and Graham also found that the proportion of nonbank activities does not have a statistically significant effect over the full sample period from 1971 to 1983 or the subperiod from 1978 to 1983. They note that the Federal Reserve phased in a "goslow" policy towards new nonbank subsidiaries beginning in 1974, and they assert that this policy was the primary cause for the difference in results between the 1971 to 1977 subperiod and the 1978to 1983 subperiod. Thus, they conclude that "when management was left more to its own devices, those BHCs with above-average nonbank activity also exhibited above-average risk."19 This suggests that careful regulation of BHC expansion may be appropriate. The results of the Boyd and Graham study may reflect BHC parent activities and double leverage, but unlike the other studies using accounting data, Boyd and Graham's results cannot be biased by interaffiliate transactions. However, the study's findingthat nonbank activities are associated with greater B H C risk is consistent with several different interpretations that could reflect three very different explanations: nonbanks causean overall increase in a BHC's riskof failure; nonbanks do not change the overall riskoffailure of BHCs;and nonbank subsidiaries cause a decrease in the overall risk of failure of BHCs. For example, the first explanation, that nonbank subsidiaries cause an increase in risk, could be interpreted as an indication that at least some BHCs cannot attain their desired risk level with their traditional banking subsidiaries. Nonbank subsidiaries enable BHCs to increase their risk because they allow expansion into the riskier FEDERAL RESERVE BANK OF ATLANTA aspects of banking and because nonbank subsidiaries are less tightly regulated than bank subsidiaries. This interpretation raises some concern about additional deregulation of activities, since BHCs could use expanded powers in nonbank activities to become riskier. The hypothesis that nonbank subsidiaries cause an increase in BHC risk is inconsistent with the findings of Litan and Wall showing that BHCs are slightly less risky than banking subsidiaries, and with Wall's finding that nonbank subsidiaries are risk-reducing. An interpretation consistent with the risk neutrality explanation is that greater nonbank activity is a sign of lower BHC risk aversion, but that nonbank subsidiaries do not cause an increase in overall BHC risk of failure. In this case the bank affiliates are often riskier than the nonbank subsidiaries. Management may invest in nonbank subsidiaries because the long-run risk/return reward is more favorable. This interpretation is not favorable for deregulation, because it suggests that BHCs may not use expanded powers to reduce risk through diversification. However, it may not be as unfavorable for diversification as the first interpretation since it allows for the possibility that BHCs will be able to attain their desired risk level with traditional banking services, thus implying that further deregulation will not automatically result in BHCs becoming riskier. This interpretation of Boyd and Graham's results does not necessarily conflict with Litan's and Wall's evidence that BHCs are less risky than banks, because both of those studies found that the difference between the riskiness of banks and BHCs is small. An interpretation consistent with the risk reduction explanation is that BHCs with the riskiest banks are investing in nonbank subsidiaries to reduce risk through diversification. For example, geographic restrictions on bank operations may not allow a BHC to obtain adequate geographic diversification. In this case the positive relationship between B H C risk and level of nonbank activities observed by Boyd and Graham occurs because existing nonbank subsidiaries are not sufficient to obviate the risk created by the banking subsidiaries. This interpretation of the risk reduction explanation suggests that deregulation of BHC activities could reduce the riskiness of banking organizations. The risk reduction interpretation is not consistent with Boyd and Graham's suggestion that the differences in results between the two subperiods were due to changes in Federal Reserve 29 policy. However, their study does not test the role of Federal Reserve BHC regulatory policy, and the possibility exists that the results could be traced to other factors such as disparities in the economic conditions between 1971-77 and 197883.20 Thus, even though Boyd and Graham's findings seem to suggest that deregulation could lead to increased risk-taking, their results can be interpreted in several ways and are not necessarily inconsistent with nonbanks' having a risk-reducing influence. The results of Litan and Wall could also be interpreted as supporting this finding. W h i l e the evidence seems to indicate that existing nonbank subsidiaries can significantly affect the riskiness of individual BHCs, whether or not they have increased the overall riskiness of banking organizations is less clear. Correlatingthe results of the three major previous studies—Litan (1985), Wall (1986), and Boyd and Graham (1986)— suggests a likely hypothesis, however. Litan and Wall found that nonbank subsidiaries may have caused a small decline in average risk, but they note that nonbank subsidiaries may have increased the riskiness of some BHCs. Boyd and Graham find that B H C risk is positively associated with its investment in nonbank activities, and one interpretation of Boyd and Graham's results conflicts with Litan's and Wall's research, namely, that nonbank subsidiaries cause BHCs to become riskier. However, Litan's and Wall's findings are consistent with two other interpretations of Boyd and Graham's results: that nonbank subsidiaries have a neutral effect on B H C risk orthat nonbanks cause a reduction in B H C risk. Litan's and Wall's results can be reconciled with those of Boyd and Graham if nonbanks have a neutral or risk-reducing effect on BHC risk. Do Nonbank Activities Increase B H C Risk? The hypothesis that nonbank activities have a neutral or risk-reducing effect can be tested empirically. If nonbank activities are indeed risk-reducing or neutral with regard to risk, then the relationship that Boyd and Graham observe between nonbank activity and BHC risk must exist because (1) differences in the riskiness of the banking subsidiaries cause differences in B H C risk and (2) nonbank activity is positively correlated with risk among banking subsidiaries. Both of these relationships can be subjected to empirical 30 tests. If either condition does not exist, the risk neutrality and risk reduction explanations for Boyd and Graham's findings would be doubtful. Before going to the empirical analysis one note of caution is necessary. Failure to find significant relationships may suggest rejection of the risk neutrality and risk reduction hypothesis. The existence of significant relationships, however, is not by itself grounds for rejecting the interpretation that nonbank activities cause BHCs to be riskier. There is no logical inconsistency between nonbank subsidiaries causing greater B H C risk and, either (1 ) bank subsidiaries being associated with greater B H C risk or (2) nonbank activity being positively correlated with bank risk. Empirical Analysis The following empirical analysis uses Wall's sample of 267 BHCs reporting positive investment in nonbank subsidiaries for six of the nine years from 1976 to 1984. The data is from parent and consolidated BHC financial statements filed with the Federal Reserve (FRY-6 and FRY-9). The advantage of using FRY-6 and FRY-9 is that it makes possible the calculation of the risk of failure among bank subsidiaries, nonbank subsidiaries, and the consolidated BHC. It also provides information on BHC investment in nonbank subsidiaries. A potential problem with FRY-6 and FRY-9 is that reliable data begins only in 1976. However, Boyd and Graham did not find a significant relationship between the proportion of B H C investment in nonbank activities between 1978 and 1983. Their finding of a relationship between the proportion of BHC investment in nonbank activities and B H C risk must be replicated with Wall's sample before the explanations of Boyd and Graham's results can be considered. All three limitations of the FRY-6 and FRY-9 data noted in the discussion of nonbank profitability also apply to this analysis: (1) reported income of bank and nonbank subsidiaries will depend on the pricing of interaffiliate transactions; (2) BHC parent activities and double leverage policies will influence reported consolidated income and equity; and (3) the reported income and equity of subsidiaries acquired via the purchase method of accounting may not be the same on the parent B H C s books as it is on the subsidiaries' financial statements.21 The risk variable used is a measure of the possibility that losses will exceed expected income OCTOBER 1986, E C O N O M I C REVIEW Table 7. Rank Order Correlations (Significance level in parentheses) Nonbank Proportion Leverage Ratio Log of Total B H C Assets g' of Bank Subsidiaries g' of Nonbank Subsidiaries -0.1493* (0.0150) -0.1754* (0.0040) 0.1758* (0.0040) 0.6612* (0.0010) 0.1642* (0.0070) 1.0000 -0.0574 (0.3500) 0.0769 (0.2100) -0.2082* (0.0010) 0.4477* (0.0010) 1.0000 0.3295* (0.0010) -0.0698 (0.2560) -0.1202* (0.0500) 1.0000 0.1605* (0.0090) 0.0662 (0.2810) 1.0000 0.1956* (0.0010) g' of Consolidated BHC Nonbank Proportion Leverage Ratio Log of Total B H C Assets g' of Bank Subsidiaries 'Statistically significant at 5 percent. Source: Calculated by Federal Reserve Bank of Atlanta from Federal Reserve Board of Governors data. plus capital. The measure is defined as g' = (1 + m) / s where g' = risk measure m = mean return on equity over the period from 1976 to 198422 s = standard deviation of return on equity over the period from 1976 to 1984. This risk measure is similar to Boyd and Graham's z-score. The primary difference is that the z-score is based on return on assets rather than return on equity. The g' measure is inversely related to B H C risk, higher levels of g' implying lower risk. Following Boyd and Graham, leverage is defined as consolidated B H C debt divided by consolidated assets, and the size variable is the logarithm of consolidated BHC assets. The empirical technique used to analyze the data is calculation of rank order correlations between the different variables. Regression analysis and commonly used correlation techniques assume a linear relationship between the variables of interest. However, it is highly unlikely that the risk measure used by this study is linearly related to the probability of failure. Rank order correlation eliminates the need to assume a linear relationship between the variables. The only assumption FEDERAL RESERVE BANK OF ATLANTA required for rank order correlation is that the risk measure be able to rank BHCs from the riskiest to the least risky. The principal disadvantage of this technique is that it does not allow risk to be regressed on three independent variables, so that only one pair of variables can be analyzed at a time. The correlation coefficients measure the closeness of the relationship between two sets of rankings. The coefficients may range between -1 and + 1. A coefficient of +1 indicates that the highest value in one ranking is associated with the highest value in the other ranking, second highest with second highest, third highest with third highest, and so forth. A value of -1 indicates a perfect inverse relationship between the rankings, and a value of zero implies no relationship between the rankings. The figure in parentheses gives the probability that the true rank order correlation coefficient is equal to zero, that is, the probability that no relationship exists between the variables. The closerthe number in parentheses is to zero, then, the more significant the correlation. In Table 7 w e see the strongest correlation, 0.6612 with a significance level of 0.0010, between the risk measure of consolidated BHCs and the risk measure of bank subsidiaries. Another strong correlation, 0.3295 with a significance level of 0.0010, exists between the leverage ratio and the log of total B H C assets. 31 The rank order correlations in Table 7 are similar to Boyd and Graham's regression results for the 1971 to 1977 subperiod. BHC risk is directly related to B H C nonbank activity and leverage, and inversely related to the logarithm of total B H C assets (recall that g' is inversely related to risk). Thus, these data may be able to offer evidence supporting or rejecting the risk neutrality and risk aversion explanations. The results of these tests do not reject the risk neutrality and risk reduction interpretation of Boyd and Graham's results. The riskiness of the banking subsidiaries is positively correlated with the riskiness of the BHC, suggesting that the bank subsidiaries may be the primary determinant of BHC risk. Furthermore, B H C investment in nonbank activities is positively associated with the riskiness of the banking subsidiaries. This suggests that B H C investment in nonbank activities could be the result of either management risk preferences or an attempt by management to diversify. None of the otherempirical results (Table 7) can be used to reject the risk increasing, risk reducing, or risk neutrality interpretations. However, several results provide interesting information on the risk structure of BHCs. The positive relationship of nonbank subsidiary risk with consolidated B H C risk, BHC leverage, and banking subsidiary risk is consistent with the possibility that management preferences influence the riskiness of the BHC's subsidiaries and determine the use of consolidated leverage to influence overall risk. The negative relationship between percentage investment in nonbank subsidiaries and nonbank subsidiary risk suggests that nonbank performance becomes more stable as the size of the subsidiaries increase. Conclusion Analysis of the contribution of nonbanks to the financial performance of BHCs suggests that at least some BHCs can gain from nonbank activities. The number of BHCs with nonbank subsidiaries has grown over the period from 1976 to 1984. The proportion of BHC investment in nonbank subsidiaries has increased markedly for BHCs with assets in excess of $10 billion. W h i l e the return on investment has been low for BHCs with assets below $1 billion, the returns have improved over the sample period for those with 32 assets between $1 and $10 billion and those with assets greater than $10 billion. Nevertheless, the riskiness of BHCs remains an issue for regulators, who must be concerned not only with the health of the B H C but with the stability of subsidiary banks, which can be seriously affected by failures among their affiliates. Advocates of deregulation suggest that nonbank subsidiaries will reduce the riskiness of BHCs through diversification. The opponents of deregulation claim that nonbank subsidiaries will engage in high-risk activities that will increase the riskiness of the parent BHC. This research concludes that while there is indeed a correlation between the proportion of nonbank activity and B H C risk, this correlation does not necessarily mean that nonbanks are the cause of the increased risk. This correlation could hold true even if nonbanks in fact decrease risk or have no impact upon BHC risk whatsoever. A composite of existing research, tested by empirical analysis, seems to support best the hypothesis that nonbanks either decrease BHC risk slightly or have little impact. Caution is nonetheless implied by the results of these studies. Nonbank activities appear to increase the risk for some BHCs and decrease it for others, indicating that the riskiness of nonbank activities should be taken into consideration in risk-based capital standards. Keep in mind, however, that nonbank subsidiaries should be analyzed in a portfolio context rather than on a stand-alone basis, because activities that appear highly risky in isolation could actually reduce the overall risk of a BHC. The indications for deregulation from studies of existing nonbank subsidiaries are, then, neutral to slightly unfavorable. Even though, according to this interpretation of the research, nonbanks are most likely to reduce BHC risk or leave it unaffected, the observed risk reductions are sufficiently small and the evidence is sufficiently ambiguous that w e cannot count on deregulation to reduce significantly the riskiness of the banking system. Furthermore, the low profitability of some bank subsidiaries could mean that nonbank activities have an adverse affect on the profits of certain BHCs. Thus, any deregulation of activities should be accompanied by careful monitoring of new nonbank subsidiaries. The author thanks David Whitehead for helpful and John Boyd for an insightful critique. comments OCTOBER 1986, ECONOMIC REVIEW NOTES 1See R. Alton Gilbert, Courtenay C. Stone, and Michael E. Trebing (1985). 2 Bank activities that create risk for the bank but do not increase bank assets are often referred to as off-balance sheet activities. An example of an off-balance sheet activity is bank issuance of letters of credit to corporations. When the bank issues a letter of credit it promises to lend money to the corporation at the firm's discretion. However, both the time period for the corporation to exercise its rights and the maximum amount of the loan are limited. In return the corporation pays a fee to the bank. Banks are assuming some risk with a letter of credit. The corporation may request a loan when it is having financial problems and cannot obtain loans elsewhere. However, no asset has been created on the bank's books until the corporation actually borrows money and, therefore, the banking organization does not need to hold additional capital. 3 Press release, Federal Reserve Board of Governors, January 24,1986, p. 27. 4 S e e Larry Wall (1984). 5 Failure to deregulate B H C activities may also endanger the banking system by limiting its ability to compete with nonbank firms. Banks may lose market share and may even become obsolete if the selling of traditional banking services in the future requires that the seller be able to provide services traditionally prohibited to banks. 6See Samuel Chase (1971), Samuel Chase and John J . Mingo (1975), and Samuel Chase and Donn Wage (1983). 7See Larry Wall (1984) and Samuel Talley (1985). 8 However, Robert J. Lawrence (1985) suggests that banks could be effectively insulated if they were in effect converted into mutual funds with sharp limits on permissible investments. ^ h i s example is based on a theoretical study of portfolios by Harry Markowitz (1952). 1 0 One reason for believing that deregulation will increase the riskiness of B H C s is that the regulators will be ineffective supervisors of nonbank activities. Regulators understand the risks inherent in traditional banking activities and can therefore properly supervise these activities. Some banks that would like to become riskier cannot because the regulators stop them. However, the regulators may not understand the risks inherent in some currently prohibited nonbanking activities (such as automobile manufacturing or retailing). B H C s could take excessive risks in nonbank affiliates without the regulators' recognizing the problem. Although this may be a legitimate concern, it does not necessarily imply that most nonbank activities must be prohibited. The problem could be solved in a variety of ways; for example, the regulators could hire experts in nonbanking fields and cover any additional costs by levying greater examination fees on to B H C s with selected nonbanking activities. 1 1 See Adi Kama (1979) for a discussion of nonbank subsidiary profitability in 1976. 1 2The source of the data is the "Bank Holding Company Financial Supplement" (FRY-9) for the period from 1975 to 1977 and the "Annual Report of Domestic Bank Holding Companies" (FRY-6) for the 1978 to 1984 period. 13 Median values are used rather than mean values to avoid distortions caused by unusual values of some variables for certain BHCs. In particular the proportion of investment figures are heavily skewed by a small number of B H C s with very substantial investment in nonbank subsidiaries. Also the return on investment figures for some nonbank subsidiaries have abnormally large absolute values, perhaps reflecting the fact that many of the nonbank affiliates are small in relation to their BHC and also the fact that many nonbank operations incurred significant start-up expenses. 14 This consideration is not limited to analysis of B H C subsidiaries; it applies to analysis of all corporations that are owned by other corporations. 15 For example, see Arnold A. Heggestad (1976), Johnson and Meinster (1974), John H. Boyd, Gerald A. Hanweck, and Pipat Pithyachariyakul (1980), Roger D. Strover(1982), Jeffrey Born, Robert A. Eisenbeis, and Robert S. Harris (1983), Robert A. Eisenbeis (1983), Larry D. Wall (1984), and Robert E. Litan (1984). 16BHCs increased risk by double-leverage. Double leverage exists when a B H C parent's equity investment in its subsidiaries exceeds the value of the parent's own stockholders equity. 17 The coefficient of variation of after-tax earnings as a percentage of assets is a measure of the riskiness of a firm's earnings. Higher coefficients of variation imply greater risk. 18A contingency table is a way of indicating whether two different classifications are dependent on each other. 1 9 John H. Boyd and Stanley Graham (1986), p. 16. 20Admittedly, there is no obvious test for the effect of Federal Reserve policies for approving nonbank subsidiaries on B H C risk. possible way to avoid the problems created by purchase accounting would be to gather income and equity data from the individual subsidiaries. Unfortunately, the only comprehensive source of financial information about individual nonbank subsidiaries of B H C s and about B H C percentage ownership of bank subsidiaries is the "Annual Report of Domestic Bank Holding Companies" (FRY-6) and that data set cannot be used to aggregate partially owned subsidiaries. The ownership figures on the Y-6 file contain numerous errors (such as ownership percentages greater than 100 percent) and omissions. Furthermore, the interpretation of the ownership figures is questionable for tiered BHCs. If a B H C owns 60 percent of subsidiary A and subsidiary A owns 70 percent of subsidiary B, then the B H C s share of B's earnings is 42 percent (60 percent times 70 percent). The Y-6 file may show an ownership interest of 70 percent, however, to reflect the fact that the B H C controls over half of B's stock. 210ne 22 Return on equity for the bank and nonbank subsidiaries is defined as the B H C parent's dividends from subsidiaries plus the parent's interest in the undistributed income of the subsidiaries divided by parent's investment in the subsidiaries. Consolidated return on equity is defined as the net income of the consolidated B H C divided by the equity of the consolidated BHC. REFERENCES Born, Jeffrey, Robert A. Eisenbeis, and Robert S. Harris. "The Benefits of Geographical and Product Expansion in the Financial Services Industry." Paper presented to the Financial Management Association Meeting in Atlanta, October 1983. Boyd, John H., and Stanley L. Graham. "Risk Regulation, and Bank Holding Company Expansion into Nonbanking," Federal Reserve Bank of Minneapolis, Quarterly Review, vol. 10 (Spring 1986), pp. 2-17. FEDERAL RESERVE BANK OF ATLANTA Boyd, John H., Gerald A. Hanweck, and Pipat Pithyachariyakul. "Bank Holding Company Diversification." Proceedings of a Conference on Bank Structure and Competition, Federal Reserve Bank of Chicago, 1980, pp. 102-121. Chase, Samuel B. Jr. "The Bank Holding Company as a Device for Sheltering Banks From Risk." Proceedings of a Conference on Bank Structure and Competition, Federal Reserve Bank of Chicago, 1971, pp. 38-49. 33 Chase, Samuel B. Jr., and John J . Mingo. "The Regulation of Bank Holding Companies," Journal of Finance, vol. 30 (May 1975), pp. 281292. Chase, Samuel, and Donn L. Wage. Corporate Separateness as a Tool of Bank Supervision. Washington, D.C.: Samuel Chase and Company, Washington D. C. 1983. Cornyn, Anthony, and others. "An Analysis of the Concept of Corporate Separateness in B H C Regulation from an Economic Perspective," Federal Reserve Bank of Chicago, Bank Structure and Competition 1986, forthcoming. Eisenbeis, Robert A. "How Should Bank Holding Companies B e Regulated?" Federal Reserve Bank of Atlanta, Economic Review, vol. 68 (January 1983), pp. 42-47. Gilbert, R. Alton, Courtenay C. Stone, and Michael E. Trebing. "The New Bank Capital Adequacy Standards," Federal Reserve Bank of St. Louis, Review, vol. 67 (May 1985), pp. 12-20. Hannon, Timothy H. "Safety, Soundness, and the Bank Holding Company: A Critical Review of the Literature," unpublished working paper, staff of the Federal Reserve Board of Governors, 1984. Heggestad, Arnold A. "Riskiness of Investments in Nonbank Activities by Bank Holding Companies "Journal of Economics and Business, vol. 27 (Spring 1975), pp. 219-223. Jessee, Michael A., and Steven A. Seelig. Bank Holding Companies and the Public Interest: An Economic Analysis. Lexington, Massachusetts: D.C. Heath and Company, 1977. Johnson, Rodney D., and David R. Meinster. "Bank Holding Companies: Diversification Opportunities in Nonbank Activities," Eastern Economic Journal, vol. 1 (October 1974), pp. 1453-1465. 34 Kama, Adi. "Bank Holding Company Profitability: Nonbank Subsidiaries and Financial Leverage," Journal of Bank Research, vol. 10 (Spring 1979), pp. 28-35. Lawrence, Robert J . "Minimizing Regulation of the Financial Services Industry," Issues in Bank Regulation, vol. 9 (Summer 1985), pp. 2231. Litan, Robert E. "Assessing the Risks of Financial Product Deregulation." Paper presented to the American Economics Association in New York, December 1985. Markowitz, Harry. "Portfolio Selection,"Journal of Finance, March 1952, pp. 77-91. Meinster, David R. and Rodney D. Johnson. "Bank Holding Company Diversification and the Risk of Capital Impairment," The Bell Journal of Economics, vol. 10 (Autumn 1979), pp. 683-694. Rhoades, Stephen A. "Interstate Banking and Product Line Expansion: Implication From Available Evidence," Loyola of Los Angeles Law Review, vol. 18 (1985), pp. 1115-1164, especially pp. 1153-57. Stover, Roger D. "A Reexamination of Bank Holding Company Acquisitions," Journal of Bank Research, vol. 13 (Summer 1982), pp. 101108. Talley, Samuel H. "Activity Deregulation and Banking Stability," Issues in Bank Regulation, vol. 9 (Summer 1985), pp.32-38. Wall, Larry D. "Has BHC's Diversification Affected Their Risk of Failure?" mimeo, 1986. "Insulating Banks from Nonbank Affiliates," Economic Review, vol. 68 (September 1984), pp. 18-28. O C T O B E R 1986, E C O N O M I C R E V I E W Take Note! a sampler of recent articles in Southeastern Economic Insight Regional Economic Updates Atlanta Fed Dollar Index Thrift Diversification Foreign Investment in the Southeast plus a statistical summary page in each issue Southeastern Economic Insight covers conditions, trends, and forecasts for the region's industries and general economy. The newsletter Insieht is offered free of Return to: Information C e n t e r F é d é r a l R e s e r v e B a n k of Atlanta 104 Marietta S t r e e t N.W. Atlanta Georgia 30303-2713 P l e a s e start my subscription to Southeastern Economic Insight Name_ Address. City FEDERAL RESERVE BANK OF ATLANTA .State. -Zip. 35 The Changing Pattern of U.S. Trade: 1975 to 1985 Jeffrey A. Rosensweig, Gretchen Lium, and Kelly W e l c h The trade deficit and the increasingly global nature of the economy often dominate in discussions of recent U.S. trade developments. Frequently overlooked are significant changes in the geographic pattern of U.S. trade over the last few years. Since 1975 shares in the total dollar value of our trade, our imports, our exports, and our trade deficits have shifted away from the oilexporting countries and Latin America toward Asia. Asia's increasing preeminence in trade applies not only to Japan but to developing nations such as South Korea, Taiwan, and Singapore. These shifts affect our analysis of the impact of world events on the U.S. international trade situation and should be taken into account in evaluating the effectiveness of policies designed to alleviate the trade deficit. To appreciate the implications of changes in geographic patterns of U.S. trade, however, it is necessary to recognize their magnitude. Overall Trends in U.S. Trade Despite talk about growing global economic integration, total U.S. trade has remained relatively constant as a proportion of G N P since 1974.1 In current dollars, total trade more than doubled, growing from $207 billion in 1975 to $573 billion. Trade's share of G N P , however, increased only slightly between 1974 and 1980— from 14.2 percent to 17.5 percent. By 1985 it had declined to 14.4 percent. Total trade figures, however, mask a drastic change in the import-export trade mix. U.S. imports rose from just over $100 billion in 1975 to over $360 billion in 1985, an average nominal 36 growth rate (not adjusted for changing prices) exceeding 13 percent per annum. Export growth, on the other hand, barely exceeded 7 percent per annum, as exports rose from $107 billion to $ 213 billion. The much faster rate of growth in imports has upset the balance of U.S. trade from a small surplus in 1975 to a record deficit in 1985 of $148 billion (Chart 1). In 1975 exports constituted 6.7 percent of C N P , and imports represented a lower 6.6 percent share, signifying a small trade surplus. The import share of G N P rose to 9.1 percent in 1985, whereas the export share fell to a low for the period of 5.3 percent, resulting in the record trade deficit. Substantial geographic shifts in patterns of U.S. trade have accompanied the sharp rise in imports (Chart 2). The Asian share of U.S. trade has grown from one-fifth to nearly one-third, whereas the oil exporters' share has fallen from 14.2 percent to 5.8 percent of U.S. trade between 1975 and 1985.2 The share of U.S. trade accounted for by Latin America and by the Soviet bloc (Soviet Union and East European countries) declined significantly since 1975, as well. The shares with Canada, Europe, and Australia during the same ten-year period remained roughly constant. Asia: A Rising Force in U.S. Trade Trade with Asia (includingthe developing nations and Japan) as a percent of total U.S. trade rose moderately from 20.9 percent in 1975 to 24.2 percent in 1980. The Asian share then exploded OCTOBER 1986, E C O N O M I C REVIEW Chart 1. U.S. Trade Balance by Region (From 1975 to 1985) Canada Europe Oil Exporters Dev. Asia Japan Soviet Bloc Latin America The relatively record deficit rapid rate of growth in 1985. in imports has upset the balance of U.S. trade Source: Calculated by Federal Reserve Bank of Atlanta from data in Directions of Trade Statistics Fund. to reach 32.2 percent of U.S. trade by 1985. In nominal dollar terms, trade with Asia has quadrupled from $44.7 billion in 1975 to $185.2 billion in 1985 (Table 1). The majority of this growth can be traced to rapid escalation of U.S. imports from Asia, and not to export growth. Imports from Asia FEDERAL RESERVE BANK O F ATLANTA from a small surplus in 1975 to a 1986 Yearbook, International Monetary have grown at a nominal rate almost twice that of our exports to Asia—18.7 percent and 9.6 percent, respectively. As a result, the U.S. trade deficit with Asian countries amounted to $82 billion in 1985, or 55 percent of our total deficit. The U.S. deficit with Asia was only $3.5 billion in 1975. 37 Chart 2. U.S. Trade by Region as a Percentage of Total U.S. Trade Canada 20.9% > Europe 23.3% Japan 10.3% Developing Asia 10.7% \ f oil Exporters 14.2% Canada 16.2% . Other (Mainly Latin America) 20.7% 1980 Europe 23.5% Japan 11.3% Developing Asia 13.0% * Other (Mainly Latin Oil Exporters 15.5% Canada 20.3% America) 20.6% 1985 Europe 23.6% Japan 16.5% Developing Asia 15.7% Oil Exporters 5.8% Other (Mainly Latin America) 18.1% Since 1975 shares in the total value of our trade have shifted toward Asia and away from the oilexporting countries and Latin America. Source: Calculated by Federal Reserve Bank of Atlanta from data in Directions of Trade Statistics 1986 Yearbook, International Monetary Fund. 38 The ratio of U.S. exports to Asia relative to U.S. imports from Asia (X/M, where X/M equal to one indicates trade balance and less than one indicates a U.S. deficit) further emphasizes the deteriorating trade balance with Asia: it was about 6/7 in 1974-75 but fell to under two-fifths by 1985.3 If ships all carried equal nominal dollar values of goods, then for every five full ships now comingto the United States with goods from Asia, three would return empty. Over the past decade, U.S. trade with Japan has consistently accounted for approximately half of all U.S.-Asian trade. The remaining half is split among the developing countries of Asia, especially Taiwan, Hong Kong, South Korea, China, and Singapore. Japan is the second leading U.S. trade partner (Canada is first), with $95 billion in total trade in 1985. At $72.4 billion, U.S. imports constituted over three-fourths of this total. Since 1975, however, Japan's share of U.S. imports rose sharply to 20 percent from 11.7 percent as the dollar value of imports grew at an annual rate of 19.4 percent (Chart 3). Over the same period Japan's share of total U.S. exports also increased moderately. The dollar value of exports to Japan grew at a more modest 9 percent annual rate. Rapid growth of trade with Japan compares with world averages for the United States of 13.1 percent for imports and 7.1 percent for exports. The developing countries in Asia make up the remaining half of U.S.-Asian trade. U.S. trade performance is better with this group than with Japan, but here too the situation has become quite unbalanced. Over the past decade, U.S. imports from the Asian developing countries have been growing almost as rapidly as imports from Japan, at a compounded average annual rate of 18 percent. These developing countries have been somewhat more accepting of U.S. exports than Japan. U.S. exports to developing Asian countries have grown at a compounded annual rate of 10.2 percent over the past decade, in comparison to Japan's record of 9 percent average annual growth (Chart 4). The rapid growth in U.S. imports relative to exports within this group means that the X/M ratio has declined by half, from well over nine-tenths to under one-half, since 1975. In 1985, our $90.2 billion in total trade with this area resulted in a deficit of $32.2 billion. This exceeds our deficit with all of Europe but is only two-thirds of our deficit with Japan. OCTOBER 1986, E C O N O M I C R E V I E W Table 1. U.S. Foreign Trade by Region ($ Billions, nominal) Region 1975 Asia w/Japan 44.7 Europe 49.7 Total Trade 1980 1985 1975 Exports 1980 1985 1975 Imports 1980 1985 1975 Deficit 1980 1985 115.8 185.2 20.6 49.6 51.6 24.1 66.2 133.6 -3.5 -16.7 -82.0 112.3 135.5 28.2 64.5 53.6 21.5 47.8 81.9 6.6 16.7 -28.3 Canada 44.5 77.4 116.7 21.7 35.4 47.3 22.8 42.0 69.4 -1.0 - 6.6 -22.2 Oil Exporters 30.3 74.0 33.6 10.4 16.9 12.0 19.9 57.1 21.6 -9.5 -40.2 -9.6 Latin America 34.3 77.7 80.1 17.1 38.7 31.0 17.2 38.9 49.1 -.1 -.2 -18.1 3.1 4.2 3.8 2.5 3.1 2.9 .6 1.1 .9 1.9 2.0 2.0 U.S.S.R. & East Europe Developing Asia 22.7 62.1 90.2 11.0 28.8 29.0 11.7 33.3 61.2 -.7 Japan 21.9 53.8 95.0 9.6 20.8 22.6 12.3 33.0 72.4 -2.8 -4.5 -32.2 -12.2 -49.8 Source: Directions of Trade Statistics 1986 Yearbook, International Monetary Fund. Clearly, developing Asia has come to claim a significant sector of U.S. trade. In 1975, trade with developing Asian countries amounted to 10.7 percent of total U.S. trade; as of 1985, that portion had grown to 15.7 percent. Breaking down total trade into its export and import components, the results parallel the Japanese experience in which imports have grown much more than exports. The share of developing Asian countries in total U.S. imports expanded more than 50 percent, from an 11.1 percent share in 1975 to 16.9 percent in 1985. Similar comparisons for U.S. exports show a 3.4 percentage point increase, from 10.2 percent in 1975 to 13.6 percent in 1985. Europe: Share Remains Nearly One-Quarter, But the U.S. Balance Declines Europe has maintained a fairly stable position in the composition of total U.S. world trade. Over the ten years from 1975 to 1985, the percent of total U.S. trade with Europe has varied little, ranging from 21.9 percent to 23.7 percent. Europe's contribution to total U.S. trade in 1985 compared with 1975 shows an even smaller change, from 23.3 percent in 1975 to 23.6 percent. The exportimport mix of European trade, however, has FEDERAL RESERVE BANK OF ATLANTA changed dramatically. The United States experienced a trade surplus with Europe from 1974 to 1982, reaching a peak surplus of $16.7 billion in 1980, but faced a growing trade deficit of $28.3 billion dollars in 1985. During the course of five years, then, the U.S. trade balance with Europe in essence deteriorated by a huge $45 billion. The ratio of U.S. exports to Europe relative to U.S. imports from Europe further emphasizes the degenerating trade balance with Europe: X / M equaled 4 to 3 from 1975 to 1976, but ten years later it was halved to 2 to 3. Over the ten-year study period, U.S. imports from Europe have grown at a compounded rate of 14.3 percent per annum, whereas U.S. exports to Europe have grown at only 6.6 percent per annum. Canada: Stable Overall Share W i t h Rising Imports Canada is the United States' most active trading partner, accounting for 20.3 percent of this country's total world trade, with a dollar value of nearly $120 billion. During the mid- to late-1970s Canada claimed 20 to 21 percent of U.S. total trade; however, Canada's share of U.S. trade declined from 20.9 percent in 1975 to 16.2 percent in 1980 before recovering to 20.3 percent in 1984 and 39 Chart 1985. Canada's continuing importance as a major trade partner for the United States calls for special attention to growing deficits with Canada. In 1975 the United States had a relatively small trade deficit of $1.01 billion with Canada. By 1985 the deficit amounted to $22.2 billion, and it is still growing. The X/M ratio, another indicator of trade patterns, also reflects the size of the deficits; the nearly balanced ratio in 1975 had fallen to just more than two-thirds by 1985. Imports from Canada over the 1975 to 1985 period have grown at 11.8 percent per annum, compared to export growth of 8.1 percent yearly. Closer analysis of U.S. deficits with Canada provides additional insight into the present situation. I n 1982, the deficit with Canada as a percent of the total U.S. deficit peaked at 30.7 percent before declining to 14.9 percent in 1985. Nevertheless, over the same period the dollar value of our deficits with our largest trading partner has continued to grow. 3. U.S. Imports from Selected Regions as a Percentage of Total U.S. Imports 1975 Canada 21.6%. ^ Japan J 11.7% Europe 20.4% / Developing Asia ^ ^k / .— A Oil Exporters 18.8% 1 1 1% Other (Mainly Latin America) 1 6.4% 1980 Canada 16.3% Japan 12.8% / Developing W Asia \ 12.9% / \ —j ; \ "^ Y A Oil Exporters Latin America: Austerity Lowers Export Share Europe 18.6% Latin American trade demonstrates that economic conditions abroad can have a significant impact on U.S. trade patterns. In 1981 the U.S. ran a trade surplus of $1.3 billion with Latin America. 4 Yet in 1982, the United States suddenly faced a deficit totaling $6 billion when austerity measures required Latin America to slash imports to service its foreign debt. Continued cutbacks were accompanied by a jump in the U.S. trade deficit with Latin America to $17.9 billion in 1983 and $20.4 billion in 1984, before it began to "level off" to $18.1 billion in 1985.5 As a result, U.S. imports from Latin America have grown at 11 percent yearly overthe ten-year period (1975 to 1985), but U.S. exports to Latin America have remained at low levels with a mere 6.1 percent per annum growth rate. Other (Mainly Latin America) 17.1% 22.2% 1985 Canada 19.2% -s. Europe 22.6% \ y / ^fl^^Ha^""^ ' Developing Asia 16.9% Substantial accompanied geographic (Mainly Latin America) ZTr15.2% Oil Exporters 6.0% shifts the sharp rise in toward Asia have imports. Source: Calculated by Federal Reserve Bank of Atlanta from data in Directions of Trade Statistics 1986 Yearbook, International Monetary Fund. 40 I The key trend established with Latin America is clearly our export share decline (Chart 4). U.S. exports to this region rose from $17 billion in 1975 to $42 billion in 1981, then plummetted to a low of less than $26 billion in 1983, recovering to just $31 billion in 1985. Latin America absorbed about 16 percent of U.S. exports in 1974-75, and as much as 18 percent in 1981. This share is now 14.6 percent. Growth of U.S. imports from Latin America has also been slow, especially relative to growth of imports from Asia. Latin America now OCTOBER 1986, ECONOMIC REVIEW accounts for 13.6 percent of total U.S. imports versus 16.3 in 1975. Falling export and import shares have resulted in a decline in total U.S. trade with Latin America from 16 percent in 1975 to 14 percent in 1985. Chart 4. U.S. Exports to Selected Regions as a Percentage of Total U.S. Exports 1975 Oil Exporters: Boom and Bust U.S. imports from predominant oil exporters (the IMF defines these countries as O P E C minus Ecuador and Gabon, but with O m a n added) peaked at $57.1 billion in 1980, then receded to $21.6 billion in 1985, nearing its level of $19.9 billion a decade before. Products imported from the oil exporting group, as a percentage of U.S. imports, showed a slight increase in the mid- to late-seventies from 18.8 percent in 1975 to 22.2 percent in 1980. However, a steady and pronounced decline since 1980 reduced their percentage share to a mere 6 percent in 1985.6 U.S. exports to the oil-exporting countries as a share of total U.S. exports have followed the general pattern of U.S. imports from these countries, but with a slightly lagged relationship and less variation. In 1975, U.S. exports to oil exporters accounted for 9.6 percent of total U.S. exports. This figure climbed to 11.1 percent in 1977-78, slid to 7.7 percent in 1980, and rebounded to 10.4 percent in 1982 before dropping steadily. The ten-year growth rates of U.S. trade with the oil exporters is the exception to the general trend of moderate U.S. export growth and rapid growth of imports. Over the past decade, U.S. exports to the oil exporters have expanded at a rate of only 1.4 percent, while U.S. imports from them have increased on average at an even more meager 0.8 percent rate. In 1980 the U.S. trade balance with all regions except this oil exporter group measured a surplus of $4.0 billion, but this surplus was swamped by the $40.2 billion deficit with the oil exporters. From 1982 onwards the trade deficit with oil exporters has diminished, nearly returning to its 1975 level of $9.5 billion. The ratio of U.S. exports to imports from oil exporters has risen from threetenths in 1980 to nearly three-fifths in 1985. Soviet Union and East Europe: U.S. Trade Small and Declining U.S. trade with the Soviet Union and Eastern Europe has followed a pattern similar to that of oil exporters but on a much smaller scale. Our trade FEDERAL RESERVE BANK OF ATLANTA 10.2% Oil Exporters America) 9.6% 24-9% 1980 Europe 29.2% Canada 16 Japan 9.4% Developing Asia 13.0% Oil Exporters 7.7% Other (Mainly Latin America) 24.6% 1985 Canada 22.2% Europe 25.1% Japan 10.6% Developing Oil Exporters Asia 5.6% 13.6% Other (Mainly Latin America) 22.9% The share of U.S. imports to developing Asian nations has grown faster than the share of U.S. exports to Japan. Source: Calculated by Federal Reserve Bank of Atlanta from data in Directions of Trade Statistics 1986 Yearbook, International Monetary Fund. 41 Asian countries have assumed a large and growing role in U.S. trade, particularly in the U.S. import market. Developing nations as well as Japan have contributed to the increase. The share of our imports supplied by Asia alone is nearing two-fifths, while our ratio of export to import trade with this region has declined below two-fifths. Explosive import growth combined with moderate export growth to Asia results in the expanding U.S. trade deficit with Asia, which has climbed to over $80 billion—more than half of our total deficit. with these countries climbed to a peak in 1979 of $5.1 billion in exports and $1.5 billion in imports. The grain embargo imposed by former President Jimmy Carter caused U.S. exports to these countries to tumble more than $2 billion by 1980 (Table 1 ). U.S. exports have yet to regain their preembargo levels. U.S. exports to Eastern Bloc countries accounted for just 2.4 percent of U.S. exports in 1975, but the export share rose to 2.8 percent in 1979. In 1985, longafterthe grain embargo's initial effect, U.S. exports to the Soviet Union and Eastern Europe as a share of total U.S. exports remained a mere 1.4 percent. The Asian gain in U.S. trade share has not penalized our two other main trading partners, Canada and Western Europe. These two areas roughly maintain their respective shares of just over one-fifth and just under one-quarter of the U.S. market. The significant loss of share has been incurred by oil exporters (primarily in the Middle East), who saw their share of U.S. trade fall by over two-thirds from 1977 to 1985. Latin America and the Soviet bloc also lost U.S. trade share, but the absolute magnitude of decline was relatively small. U.S. imports from the Soviet bloc as a share of total U.S. imports have always been negligible, less than one percent of total imports. Following the same trend as export share, the Soviet bloc import share peaked in 1979, then dropped so sharply that from 1982 through 1985 it averaged less than half of its 1977-79 average. Conclusion The composition of U.S. trade by world region has shifted substantially during the last decade. The authors are, respectively, an international economist and student interns in the Atlanta Fed's Research Department. NOTES 1 Capital markets worldwide and the international trade sectors of many other nations show significantly greater evidence of international integration than the U.S. trade sector. For a report on the trade sectors of other countries see Christopher Paul Beshouri, "The Global Economy: A Closer Look," Federal Reserve Bank of Atlanta, Economic Review, vol. 70 (August 1985), pp. 49-51. 2These figures are calculated using the IMF's Direction of Trade Statistics. The IMF classification includes essentially the 13 members of O P E C ; however, Ecuador and Gabon are excluded, while Oman is added for a total of 12 oil exporting countries. 3 The X/M ratios used throughout this article to evaluate the status of U.S. trade with different geographic regions are not meant to imply that bilateral trade balance is desirable. Clearly, it is a country's total trade balance that is ultimately important. However, these ratios help demonstrate changing patterns of U.S. trade from a disaggregated, geographical perspective. As such they can help identify the main source of our current deficit and thereby guide the analysis of policy options. 42 4 Latin America is broadly defined here as all western hemisphere developing nations. This includes Caribbean as well as Central and South American nations. 5 The deficit with Latin America reached one-fourth of the total U.S. trade deficit in 1983, but by 1985 this share had fallen to one-eighth of the U.S. deficit. The trade deficit with Latin America has held steady in contrast to other regions where it has increased dramatically. 6 These percentages, like those throughout this article, are based on nominal rather than real changes. In the present case of U.S. imports from oil-exporting countries, much of the nominal change reflects fluctuations in oil prices rather than volume changes. While it is useful for some purposes to look at real changes, the present, simpler approach is appropriate in light of the large size of the U.S. trade deficit measured in nominal terms. O C T O B E R 1986, E C O N O M I C R E V I E W FINANCE SEPT 1985 ANN. % CH3. 1,622,380 1,615,298 1,471,404 349,739 356,052 312,638 127,906 126,032 101,863 479,607 473,303 416,014 703,674 701,765 675,031 +10 +12 +26 +15 + 4 170.,699 3b,,554 13., ¡>8b 46.,637 78.,827 SEPT 1986 ALO 1986 ANN. SEPT % 1985 Q C . SEPT 1986 ALIJ 1986 SScLs Total Deposits NCW Savings Tin« Credit Union Deposits Share Drafts Tin*; 702,192 29,732 159,054 511,197 56,440 7,623 48,473 700,805 29,705 158,458 511,003 56,197 7,793 48,207 N.A. N.A. N.A. N.A. 41,081 5,822 34,170 +37 +31 +42 +11 + 7 +28 +15 + 7 WcLs lotal Deposits NCW Sav i ngs Time Credit Union Deposits Share Drafts Time 93,096 4,878 20,759 66,904 6,561 745 5,595 93,101 4,921 20,696 67,076 6,534 779 5,571 N.A. N.A. N.A. N.A. 5,056 528 4,249 +30 +41 +32 $ millions Corrmeroial Bank Deposits Demand NCW Savings Time Ccnmercial Bank Deposits Demand NCW Sav i ngs Tirre 189.1264 38.,195 17.,344 53,,404 84.,396 189,,151 39.,257 17,,034 53.,076 84.,2b4 Ccnmercial Bank Deposits Demand Musavi ngs Tine 19,,184 3,,997 1.,682 4.,084 9.,912 19, 191 4, 130 1, 664 4,,068 9.,843 16,,8 BY 3,,706 1,,316 3,,559 8,,698 + 14 + 8 +28 +15 +14 i*xbs rotai u e p o s n s NOW Sav i ngs Time Credit Union Deposits Share Drafts Time b,OOZ 303 1,113 4,482 881 147 740 0,0 »( 304 1,099 4,529 868 162 727 IN.rt. N.A. N.A. N.A. 731 110 600 Ccrrmercial Bank Deposits Demand NCW Sav i ngs Time 71,,161 14,,187 7,,393 24.,389 26.,680 70.,986 14,,572 7.,332 24.,156 26.,622 62. 12.,770 5,,692 21. ,675 23 ,874 +14 +11 +30 +13 +12 sacks rotai Deposits tew Savings Time Credit Union Deposits Share Drafts Time Dl, 3,,147 14.,341 42.,968 3.,503 381 2,871 61,068 3,202 14,307 43,035 3,493 398 N.A. N.A. N.A. N.A. 2,487 2,868 1,992 +41 +47 +44 +19 +48 +21 260 Cenrnercial Bank Deposits Demand NOV Savings Tirre 29,894 7,960 2,472 8,633 12,211 30,158 8,196 2,385 8,696 12,303 26,633 7,233 1,854 7,240 11,570 +12 +10 +33 +19 + 6 SfcLs Total Deposits NOW Savings Time Credit Union Deposits Share Drafts Time 7,428 647 1,615 5,231 1,260 124 1,157 7,416 634 1,622 5,250 1,254 125 1,150 N.A. N.A. N.A. N.A. 1,060 84 958 Ccnmercial Bank Deposits Demand tov Savings Time 28, 638 5,,107 1,,972 7.,859 14.,180 28.,534 5,,167 1.,933 7.,749 14.,142 27,,469 Ò,,U9b 1.,69b 6.,547 14.,005 + 4 + 0 +16 +20 +1 SücLs Total Deposits NJV Savings Time Credit Union Deposits Share Drafts Time 10.,163 360 2.,161 7.,660 10.,197 368 2.,167 7,,713 N.A. N.A. N.A. N.A. * * Comnereial Bank Deposits Demand fOV Savings Time 13,,490 2,,315 1 , ,211 2.,826 7.,360 13,,445 2,,414 1 , ,153 2.,797 7.,380 12,,698 2.,295 932 2.,526 7.,115 + 6 + 1 +30 +12 + 3 SStLs Total Deposits Musavi ngs Time Credit Union Deposits Share Drafts Time 2,048 120 286 1,554 Conmercial Bank Deposits Demand NOW Savings Time 26.,897 4,,629 2.,614 5,,613 14.,053 2b.,837 4.,778 2.,567 5.,610 13.,964 24,,479 4.,260 2.,096 5.,090 13.,09b +10 + 9 +25 +10 + 7 S3<Ls Total Deposits NOV Sav i ngs Time Credit Union Deposits Share Drafts Time * * 301 1.,243 5.,009 917 93 827 1,982 112 275 1,514 N.A. N.A. N.A. N.A. * ,541 301 1,,226 5.,035 919 94 826 +19 N.A. N.A. N.A. 778 74 699 +18 +26 +18 Notes: All deposit data are extracted fron the Federal Reserve Report of Transaction Accounts, other Deposits and Vault Cash (FR2900), and are reported for the average of the week ending the 1st Monday of the month. This data, reported by institutions with over $26.8 million in deposits and $2.6 million of reserve requirements as of June 1986, represents 95fc of deposits in the six state area. The annual rate of change is based on mast recent data over comparable year ago data. The major differences between this report and the "call report" are size, the treatrrent of interbank deposits, and the treatment of float. The data generated frcrn the Report of Transaction Accounts is for banks over $26.8 million in deposits as of June 1986. The total deposit data generated frcm the Report of Transaction Accounts eliminates interbank deposits by reporting the net of deposits "due to" and "due from" other depository institutions. The Report of Transaction Accounts subtracts cash in process of collection frcm demand deposits, while the call report does not. The Southeast data represent the total of the six states. Subcategories were chosen on a selective basis and do not add to total. * = fewer than four institutions reporting. N.A. = Not available at this time. Series being revised to reflect reporting changes. FEDERAL RESERVE BANK O F ATLANTA 43 CONSTRUCTION ANN. % CH6. SEPT 1986 SEPT 1986 AUG 1986 SEPT 1985 Mil. 53,213 8,696 14,955 11,939 2,478 1,171 55,031 8,758 15,093 11,952 2,526 -22 1,181 67,822 8,897 16,803 10,671 2,252 1,210 -11 +12 +10 - 3 esidenti al BuiIding Permits Value - $ M i l . Residential Permits - Thous. Single-family units Multifamily units Total Building Permits Value - $ M i l . ; Mil. 8,596 1,105 2,172 2,304 396 145 8.,903 1 . ,197 2.,196 2 ,328 404 138 11,,317 1 . ,183 2.,525 2.,207 437 161 -24 - 7 -14 + 4 - 9 -10 Residential Building Value - S M i l . Residential Permits - Thous. Single-family units Multifamily units Total Building Permits Value - S M i l . Nonresidential Building Permits Total Nonresidential Industrial Bldgs. Offices Stores Hospitals Schools 574 62 142 158 24 19 581 62 145 162 22 12 654 68 131 152 47 13 -12 - 9 + 8 + 4 -49 +46 Residential Building Permits Value - $ M i l . Residential Permits - Thous. Single-family units Multifamily units Total Building Permits Value - S M i l . Nonresidential Building Permits Total Nonresidential Industrial Bldgs. Offices Stores Hospitals Schools Mil. 4,,314 453 1.,093 1.,195 218 40 4,,397 454 1.,089 1.,214 227 42 5:,817 565 1.,123 1.,204 236 54 -26 -20 - 3 - 1 - 8 -26 Residential Building Permits Value - $ M i l . Residential Permits - T h o u s . Single-family units Multifamily units Total Building Permits Value - $ M i l . Nonresidential Building Permits - $ M i l . Total Nonresidential 1,816 Industrial Bldgs. 355 Offices 387 Stores 455 Hospitals 39 Schools 37 1,868 362 392 446 39 36 1,999 296 546 318 26 20 - 9 +20 -29 +43 +50 +85 Residential Building Permits Value - $ Mil. Residential Permits - Thous. Single-family units Multifamily units Total Building Permits Value - $ M i l . 648 26 210 165 41 31 703 27 233 174 42 30 1,399 52 410 256 65 56 -54 -50 -49 -36 -37 -45 Residential Building Permits Value - S M i l . Residential Permits - T h o u s . Single-family units Multifamily units Total Building Permits Value - $ M i l . Mil. 258 25 75 79 12 6 266 26 71 83 12 7 293 23 50 59 16 8 -12 + 9 +50 +34 -25 -25 Residential Building Permits Value - $ Mil. Residential Permits - Thous. Single-family units Multifamily units Total Building Permits Value - $ M i l . 986 184 264 252 62 11 1,087 267 265 248 62 11 1,155 179 265 219 47 11 -15 + 3 - O +lb +32 0 Residential Building Permits Value - $ Mil. Residential Permits - T h o u s . Single-family units Multifamily units Total Building Permits Value - $ M i l . AUG 1986 SEPT 1985 ANN. % CHG. (12-month cumulative rate) Nonresidential Buildin Total Nonresidential Industrial Bldgs. Offices Stores Hospitals Schools Nonresidential Building Total Nonresidential Industrial Bldgs. Offices Stores Hospitals Schools Nonresidential Building Total Nonresidential Industrial Bldgs. Offices Stores Hospitals Schools Nonresidential Building Permits Total Nonresidential Industrial Bldgs. Offices Stores Hospitals Schools Nonresidential Building Total Nonresidential Industrial Bldgs. Offices Stores Hospitals Schools - 2 92,398 91,586 79,881 +16 1,052.0 711.6 1,040.3 744.3 930.5 758.1 +13 - 6 145,620 146,626 147,702 - 1 15,823 15,917 14,286 +11 205.8 150.4 204.6 157.4 192.9 161.2 + 7 - 7 24,629 25,030 25,602 - 4 663 668 523 +27 10.7 8.7 10.5 9.7 9.7 7.4 +10 +18 1,237 1,249 1,176 + 5 8,687 8,806 8,105 + 7 106.0 93.4 105.2 97.8 101.7 98.1 + 4 - 5 13,001 13,203 13,922 - 7 3,722 3,708 3,031 +23 51.7 26.4 51.7 27.7 46.5 24.1 +11 +10 5,539 5,577 5,029 +10 623 626 805 -23 9.4 3.2 9.6 3.0 11.9 7.9 -21 -60 ,271 1,330 2,205 -42 365 360 333 +10 5.8 3.0 5.8 2.9 6.0 2.1 - 3 +43 624 626 626 - 0 1,761 1,749 1,489 +18 22.2 15.7 21.8 16.3 17.1 21.7 +29 -28 2,957 3,046 2,643 +12 NOTESData supplied by the U . S . Bureau of the C e n s u s , Housing Units Authorized By Building Permits and Public Contracts C - 4 0 . Nonresidential data exclude the cost of construction for publicly owned buildings. The"Southeast data represent the total of the six states. 44 OCTOBER 1986, E C O N O M I C REVIEW GENERAL Personal Income ($ b i l . - SAAR) Taxable Sales - $ bil. Plane Pass. A r r . (thous.) Petroleum P r o d , (thous.) Consumer Price Index 1967=100 Kilowatt Hours - m i l s . Personal Income ($ bil. - SAAR) Taxable Sales - $ bil. Plane P a s s . A r r . (thous.) Petroleum P r o d , (thous.) Consumer Price Index 1967=100 Kilowatt Hours - m i l s . ($ bil. - SAAR) Taxable Sales - $ bil. Plane P a s s . A r r . (thous.) Petroleum Prod, (thous.) Consumer Price Index 1967=100 Kilowatt Hours - m i l s . ersonal Income ($ bil. - SAAR) Taxable Sales - $ bil. Plane P a s s . A r r . (thous.) Petroleum P r o d , (thous.) Consumer Price Index 1977=100 MIAMI Kilowatt Hours - m i l s . Personal Income ($ bil. - SAAR) Taxable Sales - $ bil. Plane P a s s . A r r . (thous.) Petroleum P r o d , (thous.) Consumer Price Index 1967=100 ATLANTA Kilowatt Hours - m i l s . Personal Income ($ b i l . - SAAR) Taxable Sales - $ b i l . Plane P a s s . A r r . (thous.) Petroleum Prod, (thous.) Consumer Price Index 1967=100 Kilowatt Hours - m i l s . Personal Income ($ b i l . - SAAR) Taxable Sales - $ oil. Plane Pass. A r r . (thous.; Petroleum Prod, (thous.) Consumer Price Index 1967=100 Kilowatt Hours - m i l s . Personal Income ($ bil. - SAAR) Taxable Sales - $ bil. Plane P a s s . A r r . (thous.) Petroleum P r o d , (thous.) Consumer Price Index 1967=100 Kilowatt Hours - mi Is. ANN. % CHG. LATEST CURR. DATA PERIOD PREV. PERIOD YEAR AGO 3,430.0 N.A. N.A. 8,653.1 3,294.9 N.A. N.A. 8,978.1 + 6 SEP 3,479.6 N.A. N.A. 8,594.2 SFP JUL 330.2 217.5 328.6 193.7 324.5 204.0 + 2 + Ì 417.5 N.A. 5,561.7 1,427.0 399.7 N.A. 4,945.4 1,527.5 + 5 AUG SEP 421.2 N.A. 5,805.3 1,482.0 JUL N.A. 37.3 N.A. 32.8 N.A. 28.5 AUG SEP 44.4 N.A. 159.9 57.0 44.4 N.A. 158.7 59.0 42.7 N.A. 147.8 58.5 JUL N.A. 5.0 N.A. 4.3 N.A. 4.5 + 11 Q2 164.9 162.3 155.4 + 6 2,806.4 2,270.6 35.0 SEP 173.5 10.2 +24 SEP 174.3 10.7 2,677.6 29.0 JUL 171.2 9.9 79.4 N.A. 2,182.5 N.A. AUG 338.9 6.9 78.6 N.A. 2,086.8 N.A. JUN 338.5 6.1 74.0 N.A. 1,980.7 N.A. AUG 331.4 5.8 + 7 ANN. SEPT. % 1985 CHG. SEPT. 1986 AUG. 1986 Agriculture Prices Rec'd by Farmers 122 Index (1977=100) 80,839 Broiler Placements (thous. 64.10 Calf Prices ($ per cwt.) 37.80 Broiler Prices (t per lb.) 4.74 Soybean Prices ($ per bu.) Broiler Feed Cost ($ per ton) (Q3)190 (Q3)190 125 81,200 61.10 45.90 4.98 (Q2)189 120 77,561 58.30 31.60 4.99 (Q3)196 + 2 + 4 +10 +20 - b - 3 119 34,639 60.88 36.78 4.89 189 122 34,450 59.04 45.13 5.13 181 113 32,996 55.24 28.50 5.12 190 + 5 + 5 +10 +29 - 4 - 1 Agriculture Farm Cash Receipts - $ m i l . 824 Dates: JUN., JUN. 12,196 Broiler Placements (thous.) 59.90 Calf Prices ($ per cwt.) 35.00 Broiler Prices (t per lb.) 4.84 Soybean Prices ($ per bu.) Broiler Feed Cost ($ per ton) 189 11,911 57.70 43.00 5.17 181 903 11,268 53.90 29.00 5.16 191 - 9 + 8 +11 +21 - b - 1 2,582 2,041 63.00 37.00 4.84 189 2,139 61.40 46.00 5.17 181 3,087 1,982 57.10 30.00 5.16 230 -16 + 3 +10 +23 - 6 -18 Agriculture Farm Cash Receipts - $ m i l . 1,165 Dates: JUN., JUN. 13,969 Broiler Placements (thous.) 60.40 Calf Prices ($ per cwt.) Broiler Prices (4 per lb.) 36.00 Soybean Prices ($ per bu.) 4.78 Broiler Feed Cost ($ per ton) 189 13,854 57.20 46.00 5.11 181 1,228 13,226 51.50 29.50 5.09 195 - 5 + 6 + 1/ +/Z - 6 - 3 Agriculture Farm Cash Receipts - $ m i l . 455 Dates: J U N . , JUN. N.A. Broiler Placements (thous.) 63.00 Calf Prices ($ per cwt.) 37.00 Broiler Prices (t per lb.) 5.00 Soybean Prices (S per bu.) Broiler Feed Cost ($ per ton) 189 61.40 47.00 5.32 189 556 N.A. 57.00 31.50 5.17 250 + 11 +1/ - 3 -24 6,547 60.40 46.30 4.89 181 831 6,519 56.70 31.50 5.17 154 -17 - 1 + 7 +31 - 7 +23 -12 56.50 44.50 5.24 189 860 N.A. 54.80 28.50 5.05 173 . , Q2 Q2 02 AUG SEP 28.0 JUL Q2 AUG JUL m » • - * - I - 4 +17 - 3 +31 + 4 + 8 - 3 -20 AUG SEP 50.5 N.A. 301.7 1,349.0 JUL N.A. 5.7 N.A. 5.3 N.A. 5.6 AUG SEP 25.8 N.A. 46.1 82.0 25.2 N.A. 43.4 84.0 23.8 N.A. 41.2 85.0 JUL N.A. 2.7 N.A. 2.2 N.A. 2.4 56.2 N.A. 306.4 N.A. 55.8 N.A. 286.1 N.A. 53.3 N.A. 203.3 N.A. N.A. 6.5 N.A. 5.1 N.A. 6.1 02 AUG JUL +10 + 2 +19 « 51.2 N.A. 309.1 1,255.0 Q2 Agriculture Farm Cash Receipts - $ m i l . Dates: JUN., JUN Broiler Placements (thous.) Calf Prices ($ per cwt.) Broiler Prices (t per lb.) Soybean Prices ($ per bu.) Broiler Feed Cost ($ per ton) + 0 + 5 50.5 N.A. 304.0 1,315.0 Q2 Agriculture Prices Rec'd by Farmers Index (1977=100) Broiler Placements (thous.) Calf Prices ($ per cwt.) Broiler Prices (t per lb.) Soybean Prices ($ per bu.) Broiler Feed Cost ($ per ton) h + 0 + 1 - 3 + 2 + 8 +12 - 4 +12 + 5 +51 + 6 -18 ~~~~ Agriculture Farm Cash Receipts - $ m i l . Dates: J U N . , JUN. Broiler Placements (thous.) Calf Prices ($ per cwt.) Broiler Prices (4 per lb.) Soybean Prices ($ per bu.) Broiler Feed Cost ($ per ton) 690 6,433 60.40 41.30 4.83 189 Agriculture Farm Cash Receipts - $ m i l . Dates: J U N . , JUN. Broiler Placements (thous.) Calf Prices ($ per cwt.) Broiler Prices (4 per lb.) Soybean Prices ($ per bu.) Broiler Feed Cost ($ per ton) 756 N.A. 59.20 35.50 4.97 189 + 8 +25 - 2 + 9 NOTES: Personal Income data supplied by U . S . Department of Commerce. Taxable Sales are reported as a 12-month cumulative total. Plane Passenger Arrivals are collected from 26 airports. Petroleum Production data supplied by U. S . Bureau of Mines. Consumer Price Index data supplied by Bureau of Labor Statistics. Agriculture data supplied by U . S . Department of Agriculture. Farm Cash Receipts data are reported as cumulative for the calendar year through the month shown. Broiler placements are an average weekly rate. The Southeast data represent the total of the six states. N . A . = not available. The annual percent change calculation is based on most recent data over prior y e a r . R = revised. FEDERAL R E S E R V E BANK O F ATLANTA 45 EMPLOYMENT SOUTHEAST REGIONAL ECONOMIC ANN. % CHG SEPT 1986 AUG 1986 SEPT 1985 Civilian Labor Force - t h o u s . Total Employed - t h o u s . Total Unemployed - t h o u s . U n e m p l o y m e n t R a t e - % SA Insured Unemployment - t h o u s . Insured U n e m p l . R a t e - % M f g . A v g . Wkly. Hours Mfg. Avg. Wkly. Earn. - $ 118,244 110,229 8,015 7.0 N.A. N.A. 41.0 400 119,471 111,515 7,955 6.8 N.A. N.A. 40.7 394 115,850 107,867 7,984 6.9 N.A. N.A. 40.8 391 ;ivi M a n Labor Force - t h o u s . Total Employed - t h o u s . Total Unemployed - thous. Unemployment Rate - % SA Insured Unemployment - t h o u s . Insured U n e m p l . R a t e - % M f g . A v g . W k l y . Hours Mfg. Avg. Wkly. Earn. - $ 14,808 1,256 8.1 N.A. N.A. 41.4 357 14,847 1,279 8.1 N.A. N.A. 41.0 351 14,357 1,185 7.9 N.A. N.A. 41.3 348 Civilian Labor Force - t h o u s . Total Employed - t h o u s . Total Unemployed - thous. U n e m p l o y m e n t R a t e - % SA Insured Unemployment - t h o u s . Insured U n e m p l . Rate - % M f g . A v g . W k l y . Hours Mfg. Avg. Wkly. Earn. - $ 1,913 1,728 185 10.3 N.A. N.A. 41.7 361 1,905 1,713 192 10.3 N.A. N.A. 41.2 351 1,807 1,660 147 8.7 N.A. N.A. 41.2 351 + 6 + 4 +26 Civilian Labor Force - t h o u s . Total Employed - t h o u s . Total Unemployed - t h o u s . Unemployment Rate - % SA Insured U n e m p l o y m e n t - t h o u s . Insured U n e m p l . R a t e - % M f g . A v g . Wkly. Hours Mfg. Avg. Wkly. Earn. - S 5,603 5,251 352 6.1 N.A. N.A. 40.8 330 5,698 5,361 337 6.0 N.A. N.A. 40.7 328 5,386 5,038 348 6.2 N.A. N.A. 41.8 332 + 4 + 4 + 1 Civilian Labor Force - t h o u s . Total Employed - t h o u s . Total Unemployed - t h o u s . Unemployment R a t e - % SA Insured Unemployment - t h o u s . Insured U n e m p l . R a t e - % M f g . A v g . Wkly. Hours Mfg. Avg. Wkly. Earn. - $ 2',862 175 6.0 N.A. N.A. 41.5 342 2^875 183 6.0 N.A. N.A. 41.0 334 2', 726 188 6.7 N.A. N.A. 40.7 328 ;ivilian Labor Force - t h o u s . Total Employed - t h o u s . Total Unemployed - thous. Unemployment Rate - % SA Insured Unemployment - t h o u s . Insured U n e m p l . R a t e - % M f g . A v g . W k l y . Hours Mfg. Avg. Wkly. Earn. - $ 2,005 1,754 251 12.9 N.A. N.A. 42.3 446 1,986 1,739 247 12.5 N.A. N.A. 41.8 441 2,017 1,787 230 11.8 N.A. N.A. 42.0 440 + Civilian Labor Force - t h o u s . Total Employed - t h o u s . Total Unemployed - t h o u s . Unemployment R a t e - % SA Insured Unemployment - t h o u s . Insured U n e m p l . R a t e - % M f g . A v g . W k l y . Hours Mfg. Avg. Wkly. Earn. - $ 1,183 1,051 132 12.2 N.A. N.A. 41.0 310 1,164 1,017 147 12.7 N.A. N.A. 40.2 300 1,145 1,036 109 10.4 N.A. N.A. 40.8 296 + 3 + 1 +21 f m T T a n Labor Force - t h o u s . Total Employed - t h o u s . Total Unemployed - t h o u s . Unemployment R a t e - % SA Insured Unemployment - t h o u s . Insured U n e m p l . Rate - % M f g . A v g . W k l y . Hours Mfg. Avg. Wkly. Earn. - $ 2,323 2,161 162 7.9 N.A. N.A. 41.0 352 2,314 2,142 172 7.9 N.A. N.A. 41.2 352 2,274 2,110 164 7.9 N.A. N.A. 41.1 339 + 2 + 2 - 1 NOTES: INDICATORS + 2 + 2 + 0 + 0 + 2 + + 3 + 6 + 0 + 3 + 1 + 3 - 2 - 1 + 4 + 5 - 7 + 2 + 4 1 2 9 + 1 + 2 + 0 + 5 - 0 + 4 SEPT 1985 ANN. % CHG SEPT 1986 AUG 1986 Nonfarm Employment - t h o u s . Manufacturing Construction Trade Government Services F i n . , I n s . & Real E s t . Trans., Com. & Pub. Util. 899 19,280 5,325 24,073 16,354 23,389 6,396 5,332 19,236 5,363 24,034 15,687 23,381 6,439 5,267 Nonfarm EmiHoyment _ t h o u s . Manufacturing Construction Trade Government Services F i n . , I n s . & Real E s t . Trans., Com. & Pub. Util. 3, 2 ,,318 797 3,,282 2.,268 2.,786 763 719 12,878 2,304 800 3,269 2,157 2,761 765 719 12,772 2,318 794 3,164 2,230 2,670 740 724 + 2 0 + 3 + 4 + 2 + 4 + 3 - 1 Nonfarm Employment - t h o u s . Manufacturing Construction Trade Government Services F i n . , I n s . & Real E s t . Trans., Com. & Pub. Util. 442 354 74 317 294 249 70 72 1,437 352 74 317 291 247 70 72 1,423 357 74 303 290 244 66 73 + 1 - 1 0 + 5 + 1 + 2 + 6 - 1 Nonfarm E m p l o y m e n t - t h o u s . Manufacturing Construction Trade Government Services F i n . , I n s . & Real E s t . Trans., Com. & Pub. Util. 525 340 1,246 695 1,185 334 243 4,508 523 341 1,236 640 1,182 334 243 4,415 514 335 1,187 672 1,134 321 242 + + + + + + + + 4 2 1 5 3 4 4 0 Nonfarm E m p l o y m e n t - t h o u s . Manufacturing Construction Trade Government Services F i n . , I n s . & Real E s t . Trans., Com. & Pub. Util. 560 164 685 450 495 146 167 2,650 553 163 682 443 488 146 167 2,591 558 152 657 442 471 140 164 + + + + + + + + 3 0 8 4 2 5 4 2 Nonfarm E m p l o y m e n t - t h o u s . Manufacturing Construction Trade Government Services F i n . , I n s . & Real E s t . T r a n s . , C o m . & P u b . Util 516 167 93 375 315 318 85 105 1,504 166 94 377 305 314 85 104 1,604 177 107 387 325 325 86 115 - 5 - 6 -13 - 3 - 3 - 2 - 1 - 9 Nonfarm E m p l o y m e n t - t h o u s . Manufacturing Construction Trade Government Services F i n . , I n s . & Real E s t . Trans., Com. & Pub. Util. 865 222 37 185 192 136 37 40 835 221 37 184 176 132 37 40 848 222 38 180 193 131 36 40 + 2 0 - 3 + 3 - 1 + 4 + 3 0 Nonfarm E m p l o i e n t - t h o u s . Manufacturing Construction Trade Government Services F i n . , I n s . & Real E s t . Trans., Com. & Pub. Util. 1,966 490 89 474 321 402 91 92 1,943 488 90 473 301 399 92 93 1,891 490 87 449 308 365 90 92 + 4 0 + 2 + 6 + 8 +10 + 1 0 98,643 19,402 5,022 23,393 16,260 22,310 6,024 5,308 All labor force data are from Bureau of Labor Statistics reports supplied by state a g e n c i e s . Only the unemployment rate data are seasonally a d j u s t e d . The Southeast data represent the total of the six s t a t e s . N . A . = Not A v a i l a b l e . 46 O C T O B E R 1986, E C O N O M I C R E V I E W Federal Reserve Bank of Atlanta 104 Marietta St, N.W. Atlanta, Georgia 30303-2713 Bulk Rate U.S. Postage Address Correction Requested Atlanta, Ga. Permit 292 PAID
@FedFRASER