The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
•r V Vol. 74, No. 6 November/December 1992 3 Seasonal Accom m odation and the Financial Crises o f the Great Depression: Did the Fed "Furnish an Elastic C urrency?” • 19 Is the United States Losing its Dominance in H igh-Technology Industries? 35 An Extended Series o f Divisia M onetary Aggregates THE FEDERAL A RESERVE RANK of ST.IXHTS 1 F e d e ra l R e se rv e Hank o f St. L o u is R evie w November/December 1992 In T h is Is s u e . . . Banking reform ers o f the early 20th century often attributed financial crises to the failure o f the nation's currency and credit supplies to in crease sufficiently to meet seasonal and extraordinary demands. The Federal Reserve System was established to accommodate these demands, and from the System’s founding until the stock market crash in 1929, no crises or banking panics occurred. In the first article in this Review, "Seasonal Accommodation and the Financial Crises o f the Great Depression: Did the Fed ‘Furnish an Elastic Currency?’ ” David C. W heelock investigates a recent claim that the reappearance o f financial crises during the Depression was caused by a reduction in the Fed’s accommodation o f seasonal currency and credit demands. His review o f Fed procedures and statistical evidence about the Fed’s use o f its policy tools and about market outcomes suggests no change in seasonal policy. The Federal Reserve may rightly be criticized fo r failing to offset dramatic nonseasonal increases in currency and reserve demands during the Depression, he says, but it appears unlikely that the financial crises o f the Depression w ere caused by a change in the System’s seasonal policies. * * * Policymakers are currently debating the importance o f protecting hightechnology industries through trade and industrial policies. Some, such as Laura D'Andrea Tyson, the current chair o f the Council o f Economic Advisers, argue that protective policies are necessary because our trading partners are protecting their high-technology industries. Others argue that the market system will provide all the assistance high-technology in dustries need. Implicit in these arguments is the assumption that hightechnology industries are somehow special. In the second article in this Review, "Is the United States Losing Its Dominance in High-Technology Industries?” Alison Butler provides a careful analysis o f U.S. high-technology industries. She discusses w hy high-technology industries are considered valuable to an econom y and examines several perform ance indicators fo r high-technology industries in the United States. W hen possible, Butler compares U.S. indicators w ith similar indicators in Japan and Germany. She shows that hightechnology industries have a significant positive effect on economic grow th and that continued U.S. participation in these industries will help maintain high-wage/high-skill jobs and continued econom ic grow th in the United States. NOVEMBER/DECEMBER 1992 2 In the hope o f encouraging empirical w ork in m onetary economics— particularly w ork to determine the importance o f different methods of monetary aggregation—Daniel L. Thornton and Piyu Yue, in the third article in this Review, discuss monetary aggregation and present an ex tended data series o f Divisia m onetary aggregates. Data from January 1959 to the present are used in this article. In "An Extended Series o f Divisia M onetary Aggregates,” Thornton and Yue briefly analyze the behavior o f five Divisia aggregates—M IA , M l, M2, M3 and L. Their analysis suggests that the method o f aggrega tion is not likely to be empirically important fo r narrow m onetary aggregates like M IA and M l and that beyond some point, successively broader Divisia monetary aggregates are likely to behave similarly in applied work. Data on the simple-sum and Divisia m onetary aggregates presented in this article are available from the St. Louis Fed. For details, please see p. 52. FEDERAL http://fraser.stlouisfed.org/ RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis * * * 3 David C. Wheelock David C. Wheelock, assistant professor of economics at the University of Texas at Austin, is a visiting scholar at the Federal Reserve Bank of St. Louis. Kevin White provided research assistance. Seasonal Accommodation and the Financial Crises of the Great Depression: Did the Fed “Furnish an Elastic Currency?” It was not by accident that most o f the money panics in this country occurred in the fa ll o f the year; it was in the fa ll that the usual seasonal strain, added to an unusual credit and currency stringency, became the last straw that broke the camels back. —W. Randolph Burgess (1936), p. 206. GINNING W ITH THE stock market crash in October 1929, the United States suffered a series o f financial crises that mark the Great Depression. In each crisis, the number o f bank failures and the declines in bank reserves, the money stock and economic activity w ere greater than in the preceding episodes. Many researchers investigating the causes o f financial crises during the Great Depression have blamed the Federal Reserve, fo r either pursuing policies that led to crises or fo r failing to respond to them ap propriately.1 1See, for example, Friedman and Schwartz (1963), Chandler (1971), Miron (1986) and Wicker (1966). The continuing de bate about the role of Federal Reserve policy during the Great Depression generally is reviewed by Wheelock (1992). This article investigates a recent claim by M i ron (1986) that the reappearance o f financial crises in 1929 was caused by a reduction in the Fed’s accommodation o f seasonal currency and credit demands.2 The follow ing three types o f evidence are examined: the Fed’s procedures fo r supplying currency and bank reserves across seasons, the stability o f the seasonal behavior o f the Fed's policy tools and the seasonal behavior o f market interest rates. The Fed's accommoda tion o f seasonal demands was passive, suggest ing that a deliberate change in seasonal policy 2Miron (1986, p. 136) argues that “ the Fed accommodated the seasonal demands in financial markets to a lesser ex tent during the 1929-33 period than it had previously. This means that the frequency of panics should have increased as it did” (Miron’s emphasis). NOVEMBER/DECEMBER 1992 4 was unlikely. Statistical analysis o f the seasonal patterns o f the Fed’s policy tools and o f market outcomes suggests further that no change in seasonal policy occurred. Th e Federal Reserve may rightly be criticized fo r failing to offset dramatic nonseasonal increases in currency and reserve demand during the Depression. It ap pears unlikely, how ever, that the financial crises w ere caused by a change in the System’s seasonal policies. The first sections o f this article discuss the ob jectives o f the Fed’s founders, particularly with regard to seasonal accommodation, and describe how Fed officials implemented those objectives. A review o f how the Fed’s presence affected the seasonal pattern o f interest rates and the fr e quency o f financial crises follows. Finally the ar ticle examines w hether a change in seasonal policy was a likely cause o f the reappearance o f crises in 1929, first focusing on the Fed’s proce dures and then on statistical evidence pertaining to the seasonal patterns o f Fed tools and market outcomes. FINA N CIA L CRISES A N D THE FO UN D ING OF THE FEDERAL RESERVE SYSTEM The Federal Reserve was founded to correct banking system flaws that reform ers believed contributed to financial crises. The National Banking era, w hich began w ith the National Bank Act o f 1863 and ended w ith the opening o f the Federal Reserve Banks in N ovem ber 1914, was marked by recurrent crises. Often a crisis was touched o ff by a sudden international gold outflow or the failure o f a major financial insti tution. Occasionally such an event triggered a general run by bank depositors seeking to con vert deposits into currency. In extrem e cases, banks w ere forced to suspend currency pay ments and call loans to protect their reserves. National Banking era crises w ere generally characterized by high interest rates, many bank failures, and a slowing o f economic activity.3 3There is no generally accepted definition of financial crisis. Rather than defining the term, researchers often list characteristics of financial crises (for example, Bordo, 1986). According to Schwartz (1986, p. 11), “A financial cri sis is fueled by fears that means of payment will be unob tainable at any price and, in a fractional-reserve banking system, leads to a scramble for high-powered money. It is precipitated by actions of the public that suddenly squeeze the reserves of the banking system. In a futile attempt to restore reserves, the banks may call loans, refuse to roll over existing loans, or resort to selling assets.” Schwartz FEDERAL http://fraser.stlouisfed.org/ RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis Studies o f financial crises during the National Banking era noted that crises generally occurred at times o f the year w hen demands fo r currency and credit reached seasonal peaks. In his study fo r the National M onetary Commission, Kemm erer (1910, pp. 222-23) w rote the following: It has been found that the two periods of the year in which the money market is most likely to be strained are the periods of the 'spring revival,’ about March, April, and early May, and that of the crop-moving demand in the fall; and that the two periods of easiest money market are the 'readjustment’ period, extending from about the middle of January to nearly the 1st of March, and the period of the summer depression, extending through the three sum mer months. Of the eight panics [of the era], four occurred in the fall or early winter ... and these four included two of the three really se vere panics of the period (i.e., those of 1873 and 1907); three occurred in May ... and one, ... probably the least important, ... extended from March until well along in November. The evidence accordingly points to a tendency for panics to occur during the seasons normally characterized by stringent money markets. This does not mean that the seasonal stringencies are the causes of the panics; it does mean that the months in which they occur are the weakest links in the seasonal chain, and that in periods of extraordinary tension the chain breaks at these links. Reform ers attributed the crises o f the National Banking era to inelasticity in the nation’s cur rency supply. National bank notes, U.S. govern ment currency issued during the Civil W ar (Greenbacks), silver certificates and specie (gold coin) w ere the principal currency form s during the National Banking era. Federally chartered (national) banks w ere perm itted to issue notes valued at up to 90% (later 100%) o f the face value o f U.S. governm ent bonds they pledged w ith the Com ptroller o f the Currency. The sup ply o f national bank notes was thus tied to the volume o f governm ent bonds outstanding and the profits national banks could earn issuing notes using these bonds as security. The supdistinguishes between real crises, in which financial dis tress leads to a sudden decline in the money supply, and pseudo-crises, which do not have money supply conse quences. Kindleberger (1989), however, argues against such a distinction. This article uses the terms crisis and panic interchangeably. See Sprague (1910) for an overview of financial crises during the National Banking era and Dwyer and Gilbert (1989) for a study of the effects of bank ing panics in this era. 5 plies o f Greenbacks and silver certificates w ere fixed, as was the supply o f specie fo r short peri ods. The currency supply was thus relatively in flexible and could not be increased sufficiently to accommodate a sudden large-scale attempt by depositors to convert funds into cash. A means o f supplying large amounts o f currency rapidly was key to most banking reform proposals.4 Reform ers proposed a system in which the sup plies o f currency and credit fluctuated w ith the needs o f trade. The theoretical justification fo r such a system became known as the Real Bills Doctrine, and that doctrine was implemented with passage o f the Federal Reserve Act.5 SEASONAL AC C O M M O D A TIO N BY THE FEDERAL RESERVE The title o f the Federal Reserve Act states that one purpose o f the Federal Reserve System is "to furnish an elastic currency.”6 M em ber com mercial banks w ere required to hold reserve deposits w ith the Federal Reserve Banks instead o f holding specie or deposits with Central Reserve City and Reserve City banks, as they had under the National Banking System. The Federal Reserve Act also provided fo r a new currency fo rm —the Federal Reserve note. W. Randolph Burgess (1936, p. 150), a long-time official at the Federal Reserve Bank o f New York, argued the following: The fundamental change which the Federal Reserve System has made ... is to shift much of the burden of meeting the fluctuations in the demand for credit from the reserves o f the member banks in New York City to the twelve Reserve Banks, which through the strength of their holding of pooled reserves and through their pow er of note issue and deposit expansion can provide almost any extra funds required. 4A variety of currency substitutes were used during the banking panics of the National Banking era. Loan certifi cates issued by clearinghouses have been the most studied (see Dewald [1972], Timberlake [1984], Gorton [1985], and Dwyer and Gilbert [1989]). The Aldrich-Vreeland Act of 1908 permitted bank associations to issue “ emer gency currency” during panics, which in essence made le gal the earlier practice of issuing clearinghouse certificates. 5See Friedman and Schwartz (1963, pp. 168-73), West (1977) or Timberlake (1978, pp. 186-206) for discussions of the re form movement and analysis of various proposals. 6The title of The Federal Reserve Act reads, “An Act to pro vide for the establishment of Federal Reserve Banks, to furnish an elastic currency, to afford means of rediscount ing commercial paper, to establish a more effective super vision of banking in the United States, and for other purposes.” A m em ber commercial bank could accommo date an increase in loan demand or currency withdrawals by rediscounting eligible com m er cial paper w ith its Federal Reserve Bank. The Federal Reserve Bank w ould provide the com mercial bank w ith reserves or currency and charge its discount rate.7 The provision o f currency and reserves by the Federal Reserve Banks was intended to be largely automatic and self-regulating. Propo nents o f the Real Bills Doctrine believed that the quantities o f currency and reserves provided by the Fed w ould be sufficient but not inflationary if supplied on the basis o f short-term com m er cial loans.8 Th e Federal Reserve Banks w ere authorized to rediscount commercial, agricultur al and industrial paper, bankers acceptances used to finance foreign trade, and U.S. govern ment securities w ith maturities o f up to three months. Consistent with the Real Bills Doctrine, Federal Reserve Banks w ere not authorized to rediscount loans used to support purely finan cial activity, such as stock market call loans, be cause they w ere believed to be inflationary or speculative.9 The Federal Reserve Banks set dis count rates, subject to approval by the Federal Reserve Board, and generally supplied currency and reserves elastically through the discount w indow . The Federal Reserve Banks also sup plied reserves by purchasing bankers accep tances outright. They set buying rates and purchased all acceptances that met minimum quality standards, thus supplying reserves freely at the buying rates. Besides setting discount and acceptance buying rates, the Federal Reserve Banks w ere permitted to buy and sell U.S. governm ent securities. The Fed's founders did not envision use o f this authorization to conduct m onetary policy as w e H’his action was originally known as rediscounting because commercial bank loans were often made on a discount ba sis; hence when they were endorsed by a commercial bank and sent to a Federal Reserve Bank, they were redis counted by the Fed. 8Because the Fed was required to maintain gold reserves equal to a fraction of its deposit and note liabilities, the gold standard ultimately constrained the growth of reserves and currency. In practice, however, the Fed maintained ex cess gold reserves, so the reserve requirement had little effect on its operations before 1931. E lig ib ility requirements were somewhat broadened in 1916 to include acceptances arising from domestic trade, and Federal Reserve Banks were authorized to lend directly to member banks on their own notes secured by eligible paper. See Board of Governors of the Federal Reserve System (1943, pp. 325-26) for a summary of the types of paper eligible for rediscount and for significant changes in eligibility rules from 1914 to 1933. NOVEMBER/DECEMBER 1992 6 Figure 1 The Accommodation of Currency Demand M illions of dollars January 1919 to Decem ber 1933 know it today, how ever, but rather to provide the Federal Reserve Banks w ith an additional means o f generating revenue. The use o f m one tary policy to influence economic activity and the price level and fo r other general purposes evolved slowly, and open market operations w ere not important until the mid-1920s.1 0 In accommodating the demands o f commercial banks, Federal Reserve credit tended to increase in the spring and autumn months, w hen com mercial loan demand peaked, and again in De cember, with increased holiday demand fo r currency. A fter the holidays, currency returned to banks and credit demand declined as it did during the summer. Federal Reserve credit tended to decline in w inter and summer months. A fter the Fed’s establishment, the behavior o f short-term interest rates changed dramatically. Before 1914, rates had a distinct seasonal pattern—high in the autumn and spring and low during the w inter and summer. According to Burgess (1936, p. 204): 10See Chandler (1958), Friedman and Schwartz (1963), Wick er (1966) and Wheelock (1991, 1992) for discussion of the evolution of monetary policy and the Fed’s objectives dur ing the 1920s and early 1930s. FEDERAL http://fraser.stlouisfed.org/ RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis (M)oney rates fluctuated much more rapidly and widely before the Federal Reserve System was established. . . . In January and February money tended to be easy. In the early spring rates rose, as the demand for funds increased with the planting o f crops and spring trade. Towards summer rates fell, but rose again to the year’s high point in the autumn with har vesting and autumn trade. They continued generally high throughout the holiday period with its heavy currency requirements. . . . Since the establishment o f the Reserve System, such seasonal swings of interest rates have been almost though not quite eliminated. Burgess (p. 206) goes on to w rite, "The explana tion o f the change which has taken place is found largely in the credit elasticity provided by the Reserve System.” Other w riters have recog nized the Fed’s influence on the pattern o f rates. Friedman and Schwartz (1963, pp. 292-93), fo r example, conclude the following: The Federal Reserve 'sterilized,' as it were, seasonal withdrawals and returns of currency and thereby kept deposits of member banks at 7 Figure 2 The Seasonal Pattern of the Call Loan Rate M onthly Deviations from Annual Mean Percent the Reserve Banks largely, though not entirely, free of seasonal movements. The effect was to change the pre-1914 seasonal patterns notably. The seasonal pattern in currency outside the Treasury was widened, the seasonal pattern in call money rates narrowed. The System was almost entirely successful in the stated objective of eliminating seasonal strain. The contribution the Federal Reserve made to accommodating seasonal (and nonseasonal) cur rency demands is illustrated in figure l . 1 Feder 1 al Reserve credit and currency in circulation tended to m ove together. Federal Reserve credit is a source o f bank reserves, so by extending Fed credit w hen currency was w ithdraw n from 11The data plotted in figure 1 are monthly averages of daily figures. The source for these data is the Board of Gover nors of the Federal Reserve System (1943, pp. 369-71). 12As discussed later, this study begins its analysis in 1919 because there is controversy about the Fed’s effects in its early years and because of the disruptions caused by the Fed’s contribution to financing World War I. 13The plotted seasonal patterns are simply the monthly aver ages of the rate less the annual mean, after subtracting a time trend. The patterns are calculated from a linear regression of the interest rate on monthly dummy variables and a time trend, which included an AR(2) error process banks, the Fed reduced fluctuations in m em ber bank reserves.1 2 The seasonal pattern o f interest rates changed after the Fed opened in N ovem ber 1914. Figure 2 shows the estimated seasonal patterns o f the call loan renewal interest rate betw een January 1890 and October 1914 and betw een Novem ber 1914 and Decem ber 1933.1 In the earlier period 3 the interest rate fluctuated much m ore w idely during the year than it did after the Fed’s founding. Many studies, using a variety o f statis tical techniques, confirm the apparent decrease in seasonal amplitude illustrated here.1 Scholars 4 debate w hether the Federal Reserve was responand was estimated using maximum likelihood. It might be argued that it would be more appropriate to examine the behavior of real interest rates because the underlying sources of seasonal credit and currency demands are real phenomena. Miron (1986) argues, however, that nominal rates should reflect the extent of the Fed’s seasonal ac commodation, and most studies have focused exclusively on the behavior of nominal rates. ’ “ See Shiller (1980), Clark (1986), Miron (1986), Mankiw, Mi ron, and Weil (1987), Fishe and Wohar (1990), Canova (1991), Fishe (1991), and Holland and Toma (1991). NOVEMBER/DECEMBER 1992 8 sible fo r this change, particularly before the end o f W orld W ar I.1 Researchers generally agree, 5 how ever, that the evidence fo r a significant Fed role after W orld W ar I is strong. Clearly the Fed intended to accommodate seasonal currency and credit demands, as the follow ing statement o f Reserve Bank policy in the first Annual Report o f the Federal Reserve Board (1914, p. 17) indicates: "T h e m ore complete adaptation o f the credit mechanism and facilities o f the country to the needs o f industry, commerce, and agriculture—w ith all their seasonal fluctuations and contingencies—should be the constant aim o f a Reserve Bank’s management.”1 6 SEASONAL A C C O M M O D ATIO N A N D FINA N CIA L CRISES The elasticity o f currency and reserves sup plied by the Fed and the near elimination o f seasonal fluctuations in interest rates successful ly eliminated financial crises — or so it appeared until 1929. From the Fed's founding until the stock market crash, no crises or banking panics occurred despite a significant recession in 1921, m inor recessions in 1924 and 1927, and 5,700 bank failures during the 1920s. In the follow ing passage Burgess (1927, p. 122) noted that Fed operations apparently eliminated financial panics: In the old days there were rigid and not far distant limits to the reserves available; now the mechanism of the Reserve System provides for a much larger possible expansion. It gives much greater elasticity . . . . This elasticity results in much more stability of rates and practically eliminates the fear of money panic . . . . Burgess argued further that, "T h e presence o f ’ 5Miron (1986) and Mankiw, Miron and Weil (1987) argue strongly that Federal Reserve operations caused the seasonal fluctuations in short-term nominal interest rates to decline after 1914. Clark (1986), however, contends that a lowering of reserve requirements, which occurred simul taneously with the Fed’s opening, and gold inflows accom panying the start of World War I are more likely causes of the reduced seasonal pattern of interest rates in the United States between 1914 and 1917. Fishe (1991) also reaches this conclusion, finding little seasonal behavior before 1917 in any variable under the Fed’s control. Holland and Toma (1991) argue, however, that the Fed’s presence as lender of last resort may have made banks more willing to lend in seasons during which credit demand was high and thus might have caused a reduction in interest rate seasonality even though there was little seasonal behavior in Federal Reserve credit until the 1920s. 16The Federal Reserve continues to accommodate seasonal variation in money and credit demands. The Fed uses open market operations to remove seasonal changes in the FEDERAL http://fraser.stlouisfed.org/ RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis the Reserve System gives greater elasticity to the supply o f funds and stability to the money market and removes the fear o f money panics” (p. 125).1 Miron (1986, p. 136) also concludes 7 "that the Fed successfully eliminated financial panics from 1915 to 1928.” Based on historical experience, he calculates that the probability o f there being no financial crises in a 14-year peri od (for example, from 1915 to 1928) is .005. Like M iron (1986), other researchers have concluded that the crises o f the Great Depres sion resulted from a distinct change in Federal Reserve policy. Following Friedman and Schwartz (1963), M iron attributes the change in policy to the death o f Benjamin Strong, governor o f the Federal Reserve Bank o f N ew York and the Fed’s leading policymaker, in O ctober 1928. Other researchers, such as Trescott (1982) and Hamilton (1987), also conclude that m onetary policy changed significantly with, or just before, the onset o f the Depression.1 The rem ainder of 8 this article examines w hether there was a change in Federal Reserve accommodation o f seasonal currency and credit demands that could explain the reappearance o f financial crises beginning in October 1929. FEDERAL RESERVE M ETHODS OF SEASONAL AC C O M M O D AT IO N H ow did the Federal Reserve accommodate seasonal currency and credit demands during the Great Depression? A review o f the Fed’s methods indicates that the Fed was largely pas sive in this accommodation, suggesting that any apparent change in the seasonal pattern o f Fed eral Reserve credit was m ore likely due to changes in demand than to a deliberate policy decision. money stock, and its seasonal borrowing program permits special discount window access to banks that experience large seasonal fluctuations in loan demand. See Clark (1992) for a discussion of this program. 17ln the second edition of Burgess’ book (1936, p. 156), pub lished after the financial crises of the Great Depression, both of these sections were changed. The latter was modi fied to read "The presence of the Reserve System gives greater elasticity to the supply of funds — and the control of that elasticity is the central problem of Federal Reserve policy.” 1Specifically, Trescott dates the change in policy to early 1930, when the Fed’s open market committee was reor ganized; Hamilton dates the policy change to December 1927, when the Fed adopted a restrictive policy to combat stock market speculation. Other studies, however, have concluded that there was no fundamental change in Feder al Reserve policy at this time. This debate is examined in Wheelock (1991, 1992). 9 Figure 3 Federal Reserve Discount and Acceptance Buying Rates Percent January 1919 to Decem ber 1933 The Fed had the follow ing three main policy tools during the 1920s and early 1930s: the dis count rate, the acceptance buying rate and open market operations in U.S. governm ent securities.1 The Fed was established to provide 9 currency and bank reserves to accommodate the needs o f commerce. By specifying eligibility requirements fo r the paper that could be redis counted or purchased, the Fed's founders in tended to limit policymakers’ discretion, as w ell as to accommodate currency and credit de mands without fueling inflation or speculation. Only in open market operations in govern ment securities did the Fed determine the specific volume o f its operations. The Fed was generally passive in supplying currency and bank reserves through the discount w indow 19The Fed did not have the power to alter reserve require ments until 1933. 2°Burgess (1936, p. 42). 211 regressed each rate on a time trend and monthly dummy variables, and each regression included an AR(2) error process. Neither rate nor its first difference had a statisti cally significant seasonal pattern. An F-test of the null and by purchasing acceptances.2 The Federal 0 Reserve could o f course affect the volume o f discount loans and acceptance purchases by al tering the discount and acceptance buying rates, but neither rate had a seasonal pattern (see figure 3).2 Apparently the Fed did not alter its 1 rates to influence seasonal changes in discountw indow borrow in g or in its acceptance hold ings, although it may have altered rates fo r other reasons. The three panels o f figure 4 plot Federal Reserve credit outstanding and its prin cipal components. Each component had a statistically significant seasonal pattern, but changes in the Fed's acceptance portfolio and in discount loans w ere the principal causes o f seasonal variation in total Fed credit outstand ing. The only component whose volum e the Fed controlled directly—the governm ent security hypothesis that the coefficients on the monthly dummies equal zero cannot be rejected for either rate or its first difference. The source of these data is Board of Governors of the Federal Reserve System (1943, pp. 439-45). For months in which a rate changed, I computed a monthly average by weighting the rate by the number of days it was in effect. NOVEMBER/DECEMBER 1992 10 Figure 4 Fed Credit Outstanding and Its Principal Components January 1924 to December 1931 Millions of dollars portfolio—had the least seasonal pattern. The absence o f seasonal changes in the discount and acceptance buying rates and the minimal contri bution o f open market operations in govern ment securities to the seasonal pattern o f Federal Reserve credit indicate that the Fed’s ac commodation o f seasonal currency and credit demands was largely passive. It seems likely therefore that a shift in demand, rather than a deliberate policy action affecting supply, ex plains any change in the seasonal behavior o f Federal Reserve credit outstanding.2 2 EM PIRICAL EVIDENCE ON THE ST A B ILIT Y OF SEASONAL PATTER N S The reappearance o f financial crises in 1929 at times o f the year w hen credit and currency demands reached seasonal peaks suggests that the Fed might have reduced its accommodation o f seasonal demands. M iron (1986) contends that this was true because he finds that the seasonal fluctuations o f Federal Reserve credit w e re some what less pronounced betw een 1929 and 1933 than they had been betw een 1922 and 1928. Millions of dollars 1924 25 26 27 28 29 Millions of dollars FEDERAL http://fraser.stlouisfed.org/ RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis 30 1931 Miron estimates the seasonal pattern o f Feder al Reserve credit by calculating the uncondition al mean o f Fed credit in each month after subtracting a time trend. A convenient w ay o f doing this is to regress the series on monthly dummy variables and a time trend. Miron finds that the range and standard deviation o f the es timated dummy variable coefficients are smaller in the 1929-33 period than in the 1922-28 peri od and infers that Federal Reserve credit had less seasonal variation during the Depression than it had previously.2 He concludes therefore 3 that the Fed was less accommodative o f seasonal demands after 1928 than it was before 1928. Using M iron’s m ethodology but estimating the pre-Depression seasonal pattern over a some what longer period (1919-28), I find that Feder al Reserve credit actually had greater seasonal amplitude during the Depression years. Figure 5 plots the seasonal patterns o f Federal Reserve 22Federal Reserve credit also included a miscellaneous com ponent that was mainly float, which averaged about 4 per cent of total Fed credit outstanding. This component was also somewhat seasonal and, like discount loans and ac ceptances, was influenced more by the level of economic activity than by Fed policy. 23Miron does not test whether the changes are statistically significant. 11 F ig u re 5 The Seasonal Pattern of Federal Reserve Credit Monthly Deviations from Annual Mean Millions of dollars credit betw een 1919 and 1928 and betw een 1929 and 1933.2 These data do not suggest a 4 decline in Fed accommodation after 1928. Table 1 presents evidence on the statistical significance o f the seasonal patterns o f Federal Reserve credit and its chief components—the Fed’s acceptance holdings, discount w indow loans and governm ent security holdings. Con tinuing w ith Miron's methodology, I estimate the seasonal patterns o f each series and the change (first difference) o f each series as the average values fo r each month after rem oving any time trend. These averages are simply the estimated coefficients from a regression o f each variable on monthly dummy variables and, fo r the non-differenced data, a time trend.2 The 5 seasonal pattern is statistically significant if the 24The figure plots the estimated dummy coefficient for each month less the average of the 12 estimated coefficients. In addition to the dummy variables, the model includes a time trend and an AR(2) error process (which was suggest ed by standard model selection criterion) and was estimat ed using maximum likelihood. 25Model selection criterion suggested the use of an AR(2) er ror process in modeling Federal Reserve credit and an AR(3) process for each component. I used AR(1) for the change in Federal Reserve credit and AR(2) for the null hypothesis that the estimated monthly dum m y coefficients equal zero can be rejected. Ta ble 1 reports the test statistics fo r this hypothe sis. Between 1919 and 1928, Federal Reserve credit and each component had statistically sig nificant seasonal patterns.2 Between 1929 and 6 1933, how ever, only the Fed’s governm ent secu rity portfolio had a statistically significant seasonal pattern. These tests appear to suggest that the Fed was less accommodative o f seasonal demands during the Depression. The smaller F-statistic values fo r Fed accommodation o f seasonal demands during the Depression are not, how ever, necessarily evidence o f less seasonality because w ith fe w e r observations the seasonal patterns are estimated less precisely. Thus the conclusions drawn from them are less certain. changes in each component. I also estimated models for the difference in the logs of each variable. The results for these models are identical to those of the first-difference models. 26Because researchers who argue that Fed policy changed do not agree on the date of this change, I chose to follow Miron (1986) and break the sample at December 1928. Breaking the sample at December 1927, however, does not qualitatively change the results. NOVEMBER/DECEMBER 1992 12 Table 1 Significance of Seasonal Patterns in Federal Reserve Credit and its Components (F-Test Statistics)______________________ January 1919D ecem ber 1928 January 1 9 2 9 Decem ber 1933 Federal Reserve credit Acceptances Discount loans Government securities 23.48** 4.10** 5.42** 4.02** 0.99 1.79 1.04 2.02* 3.47’ 1.90 1.24 2.54’ Change Change Change Change 25.00** 4.26** 5.70** 3.08** 1.15 1.47 0.90 1.71 4.59’ 2.29 1.55 2.07 in in in in Federal Reserve credit acceptances discount loans government securities January 1 9 2 9Septem ber 1931 * Statistically significant at the .05 level. ** Statistically significant at the .01 level. Data source: Board of Governors of the Federal Reserve System (1943), pp. 369-71. The data are monthly averages of daily figures. Table 2 Stability of Seasonal Patterns in Federal Reserve Credit and its Components, December 1928 Breakpoint (F-Test Statistics) January 1919D ecem ber 1933 January 1919Septem ber 1931 Federal Reserve credit Acceptances Discount loans Government securities 1.02 1.68 1.24 0.70 1.55 2.61** 1.02 0.79 Change Change Change Change 1.17 2.04* 1.33 1.13 1.53 2.84** 0.92 0.98 in in in in Federal Reserve credit acceptances discount loans government securities * Statistically significant at the .05 level. ** Statistically significant at the .01 level. FEDERAL http://fraser.stlouisfed.org/ RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis 13 Table 3 Stability of Seasonal Patterns in Federal Reserve Credit and its Components (Standard Deviations of Monthly Dummy Coefficients) January 1919D ecem ber 1928 January 1 9 2 9 Decem ber 1933 January 1 9 2 9 Septem ber 1931 Federal Reserve credit Acceptances Discount loans Government securities 75.41 47.52 34.58 10.88 116.69 65.38 47.91 47.94 139.02 72.74 64.39 41.34 Change Change Change Change 50.19 27.79 22.13 16.04 90.84 47.01 56.16 29.97 78.63 48.62 24.95 33.17 in in in in Federal Reserve credit acceptances discount loans government securities The results o f further tests fo r a change in the Fed’s seasonal accommodation are presented in table 2 and table 3. Table 2 summarizes tests to determine w hether seasonal patterns o f Fed credit and its components before Decem ber 1928 w ere significantly different from seasonal patterns after Decem ber 1928. Except fo r the change in the Fed's acceptance holdings, these tests indicate no change in seasonal patterns.2 7 Table 3 reports the standard deviations o f the monthly dummy coefficients fo r each variable in each period.2 M iron (1986) concluded that 8 the Fed was less accommodative during the Depression in part because the standard devia tion o f the seasonal pattern o f Fed credit was smaller betw een 1929 and 1933 than it had been betw een 1922 and 1928. Although com par ing the standard deviations o f the seasonal pat terns is not a statistically rigorous test, it does provide evidence that the seasonal variability o f each series increased or decreased over time. The standard deviation o f the seasonal co effi cients fo r Fed credit and each component is larger fo r the Depression years than it had been betw een 1919 and 1928. The differences 27lf December 1927 is used as the breakpoint, the results are identical to those reported in table 2, except that the first difference of the Fed’s acceptance holdings had a sta ble seasonal pattern. 28These are simply the standard deviations of the point esti mates of the seasonal dummy coefficients and should not be confused with the standard errors of the coefficient es timates. are, how ever, not statistically significant.2 Thus 9 these data do not support the hypothesis that the Fed was less concerned w ith seasonal credit and currency demands during the Depression than it had been earlier. Federal Reserve credit outstanding was sub ject to large and erratic fluctuations during the Great Depression (see figure 1). In particular, a pronounced change in its behavior appears to have occurred in late 1931, suggesting that in ferences drawn from analysis o f Fed credit over the entire 1929-33 period might be misleading. On September 21, 1931, Great Britain stopped converting pounds sterling into gold. Fears that the United States w ould soon follow Britain o ff the gold standard led to a large w ithdraw al o f foreign deposits from U.S. banks and a conse quent gold outflow. The U.S. m onetary gold stock declined 15 percent in the six w eeks after Britain’s action.3 Domestic depositors also 0 panicked and converted deposits into currency. M em ber banks partially offset the reserve drains caused by the gold and currency out flows by selling acceptances to the Fed and by borrow in g at the discount w indow . Federal 29The standard deviations for 1928-33 are also larger than those for 1919-27, but the differences are again not statisti cally significant. The specific test employed is a Wald test of the equality of the variances of the monthly dummy coefficients across the two periods. I thank Joe Ritter for suggesting this test. 30Board of Governors of the Federal Reserve System (1943, p. 386). NOVEMBER/DECEMBER 1992 14 Reserve credit outstanding rose $959 million (75 percent) betw een September 16 and October 21.3 The increase in Fed credit outstanding was 1 not enough, how ever, to prevent a substantial decline in bank reserves.3 2 Federal Reserve credit outstanding declined somewhat in early 1932 but began to rise in March 1932 with what was then the Fed’s lar gest program o f open-market purchases o f governm ent securities.3 From March 1932 to 3 August 1932 the Fed bought m ore than $1 bil lion o f securities, and these purchases probably explain the atypical mid-year rise in Federal Reserve credit in 1932. The events o f late 1931 and 1932 appear to have altered the pattern o f Federal Reserve credit. The evidence reported in table 1 indi cates that although Fed credit did not have a statistically significant seasonal pattern betw een 1929 and 1933 as a whole, it did have one b e tw een January 1929 and September 1931. Thus if a change in the seasonal behavior o f Federal Reserve credit occurred during the Depression, it was m ore likely to have occurred after Sep tem ber 1931 than before September 1931. This is important because three o f the five financial crises o f the Depression occurred before this date and thus cannot be attributed to a possible change in Federal Reserve seasonal policy in September 1931. Statistical tests to determine w hether the sea sonal patterns o f Federal Reserve credit and its components changed betw een the January 1919Decem ber 1928 and January 1929-September 1931 periods also suggest no shift in seasonal policy b efore September 1931. Only in the case o f the Fed’s acceptance holdings is it possible to reject the null hypothesis that no change in seasonal patterns occurred (see table 2). Com parison o f the standard deviations o f the seasonal patterns also casts doubt on the view that the Fed was less accommodative of seasonal demands before late 1931 (see table 3).3 4 3 Board of Governors of the Federal Reserve System (1943, 1 p. 386). 32The Fed argued that it lacked sufficient gold reserves to purchase government securities in sufficient quantities to prevent reserves from declining. Friedman and Schwartz (1963) contend that the Fed did have sufficient reserves and in any event had the authority to suspend its reserve requirement temporarily. Wicker (1966), however, argues that Fed officials feared that security purchases would indi cate an unwillingness to defend the dollar’s value and ex acerbate the gold outflow. See Wheelock (1991, 1992) for further discussion. 33The Fed’s gold reserve requirement was relaxed signifi cantly by the Glass-Steagall Act of 1932, which permitted FEDERAL http://fraser.stlouisfed.org/ RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis The period from September 1931 to December 1933 is too short and was too volatile to deter mine seasonal m onetary policy, and the data af ter 1933 contain no information about accomm o dation o f seasonal demands. The crisis follow ing Britain's departure from the gold standard, the Fed’s large open-market purchases in 1932, the collapse o f the banking system and the bank holi day in March 1933 all had unusually large effects on Federal Reserve credit. A fter 1933 gold and currency inflows allowed banks to accumulate large excess reserve holdings, which virtually eliminated the demand fo r Fed credit. A ccord ingly, Federal Reserve credit and its components varied little from 1934 until W orld W ar II. THE SEASONAL BE H A V IO R OF INTEREST RATES This section examines the extent o f change in the seasonal behavior o f short-term interest rates before and during the Great Depression. By supplying currency and reserves in response to seasonal demands, the Federal Reserve, at least by 1919, had substantially reduced the seasonal amplitude o f interest rates. If the Fed reduced its accommodation o f seasonal demands during the Depression, it seems likely that in terest rates w ould have becom e m ore seasonal. Table 4 summarizes tests to determine whether three short-term interest rates had statistically significant seasonal patterns.3 The 5 commercial paper rate had a statistically signifi cant seasonal pattern betw een 1919 and 1928. The bankers acceptance and call loan rates, however, did not have statistically significant seasonal patterns. None o f the rates had a statistically significant seasonal pattern during the Depression, neither during the January 1929-December 1933 period nor during the January 1929-September 1931 period.3 6 Table 5 reports tests o f the stability o f the seasonal patterns o f each interest rate. Although the Fed to use government securities to partially back its liabilities. 34The differences in the standard deviations between the January 1919-December 1928 and January 1929-September 1931 periods are not statistically significant. The findings are not affected if December 1927 is used as the break point. 35Specifically, I test the null hypothesis that the coefficients on the monthly dummy variables all equal zero. 36The call loan rate did have a statistically significant seasonal pattern between January 1919 and December 1927. The regressions for the level of each rate included an AR(2) error process, and those for the first difference of each rate included an AR(1) error process. 15 Table 4 Significance of Seasonal Patterns in Short-Term Interest Rates (F-Test Statistics)____________________________________ January 1919D ecem ber 1928 January 1 9 2 9 Decem ber 1933 January 1 9 2 9Septem ber 1931 Commercial paper rate Call loan rate Bankers acceptance rate 2.94** 1.44 1.19 1.29 1.35 1.46 1.23 1.16 1.29 Change in commercial paper rate Change in call loan rate Change in bankers acceptance rate 2.78** 1.33 0.86 0.82 1.16 1.10 0.88 1.01 0.96 * Statistically significant at the .05 level. ** Statistically significant at the .01 level. Data source: Board of Governors of the Federal Reserve System (1943), pp. 450-51. The data are monthly averages of daily figures. Table 5 Stability of Seasonal Patterns in Short-Term Interest Rates, December 1928 Breakpoint (F-Test Statistics) January 1919D ecem ber 1933 January 1919Septem ber 1931 Commercial paper rate Call loan rate Bankers acceptance rate 0.92 1.46 1.21 1.47 1.23 1.20 Change in commercial paper rate Change in call loan rate Change in bankers acceptance rate 1.05 1.64 1.41 1.73 1.68 1.47 * Statistically significant at the .05 level. ** Statistically significant at the .01 level. NOVEMBER/DECEMBER 1992 16 Table 6 Stability of Seasonal Patterns in Short-Term Interest Rates (Standard Deviations of Monthly Dummy Coefficients) January 1919D ecem ber 1928 January 1 9 2 9 Decem ber 1933 January 1 9 2 9 Septem ber 1931 Commercial paper Call loan rate Bankers acceptance rate 0.06 0.24 0.09 0.16 0.26 0.26 0.12 0.47 0.13 Change in commercial paper rate Change in call loan rate Change in bankers acceptance rate 0.08 0.20 0.06 0.14 0.32 0.22 0.12 0.43 0.15 no rate had a statistically significant seasonal pattern during the Depression, the null hypothe sis that the seasonal pattern was stable is ac cepted in all cases at conventional levels o f significance. Finally table 6 reports the standard deviations o f the seasonal dummy coefficients from each regression. For each rate and the change in each rate, the standard deviations are larger fo r the 1929-33 period than fo r the 1919-28 period. The standard deviations o f the seasonal dummy coefficients are also higher during the January 1929-September 1931 peri od than during the January 1919-December 1928 period. Thus, as with Federal Reserve credit and its components, comparison o f the standard deviations suggests that the interest rates might have fluctuated somewhat m ore w idely across seasons during the Depression than b efore it. How ever, none o f the changes in the standard deviations is statistically significant. T o put the possible increase in the seasonal pattern o f interest rates during the Depression in perspective, figure 6 plots the estimated coefficients on the monthly dummy variables fo r the call loan renewal rate betw een 1919 and 1928 and betw een 1929 and 1933. I use the same scale as in figure 2, w h ere these co effi cients are plotted fo r the 1890-1914 and 1914-33 periods. Comparison o f the tw o figures shows that the increase in the seasonal ampli tude o f the call loan rate during the Depression 37The months in which the call loan rate reached seasonal peaks between 1929 and 1933 do not coincide with those between 1919 and 1928. The behavior of the call loan rate probably changed markedly after the stock market crash in 1929. Similar plots for the commercial paper and bankers acceptance rates show that the seasonal high and low months for these rates remained the same across periods. http://fraser.stlouisfed.org/ RESERVE BANK OF ST. LOUIS FEDERAL Federal Reserve Bank of St. Louis was small relative to the decline in amplitude in 1914.3 For comparison, the standard deviation 7 o f the seasonal pattern o f the call loan rate b e tw een January 1890 and O ctober 1914 is 1.02 (0.67 fo r the first difference o f the call loan rate). For the Novem ber 1914-Decem ber 1933 period the standard deviation is 0.16 (0.13 fo r the first difference). Th e increases in the stan dard deviations after Decem ber 1928 are thus quite small in comparison w ith the size o f the changes in N ovem ber 1914. The levels o f the standard deviations during the Depression are also small in comparison with those in the pre-N ovem b er 1914 period.3 Th e various evi 8 dence that interest rates w ere m ore seasonal during the Depression is ambiguous. The possi ble increase in the seasonal pattern o f interest rates after 1929, how ever, seems too small to conclude that unaccommodated seasonal curren cy and credit demands caused the return of financial crises during the Depression. CONCLUSION Proponents o f banking reform in the United States in the early 20th century noted that financial crises tended to occur in months w hen currency and credit demands reached seasonal peaks. Eliminating crises was a principal goal o f the founders o f the Federal Reserve System, and they designed methods o f accommodating 38Because the United States did not have an active bankers acceptance market before 1914, it is impossible to make a similar comparison for the bankers acceptance rate, 17 Figure 6 The Seasonal Pattern of the Call Loan Rate Monthly Deviations from Annual Mean Percent January 1919 to December 1928 and January 1929 to December 1933. currency and credit demands to accomplish that objective. Federal Reserve credit outstanding rose and fell with fluctuations in currency and loan demands, and the Fed’s presence appears to have substantially reduced seasonal fluctua tions in bank reserves and interest rates. Most important, until 1929 it appeared that the Fed’s presence had eliminated financial crises. The reappearance o f financial crises during the Great Depression suggests the possibility that the Fed accommodated seasonal currency and credit demands less in those years than it had betw een 1914 and 1928. This article shows, how ever, that the Fed’s accommodation o f seasonal demands was generally passive, occur ring mainly through discount loans and accep tance purchases. Th ere was no seasonal pattern in the Fed’s discount and acceptance buying rates. Commercial banks initiated most seasonal extensions o f Fed credit, suggesting that changes in demand, rather than deliberate policy deci sions, caused any apparent changes in the seasonal pattern o f Fed credit. In addition, Fed eral Reserve credit and its components do not appear to have been less seasonal during the Depression, at least not before September 1931, than they had been during the 1920s. And although the seasonal pattern o f interest rates during the Great Depression may have in creased slightly, the seasonal fluctuations re mained trivial compared w ith those that occurred before 1914. The Fed’s failure to p re vent banking panics and declines in bank reserves and the money supply during the Great Depression was a serious error. It ap pears unlikely, how ever, that the financial crises o f the Great Depression w ere caused by a change in the seasonal policies o f the Federal Reserve System. REFERENCES Burgess, W. Randolph. The Reserve Banks and the Money Market (Harper and Brothers, 1927). ________The Reserve Banks and the Money Market, revised edition (Harper and Brothers, 1936). Board of Governors of the Federal Reserve System. Banking and Monetary Statistics, 1914-1941 (1943). Bordo, Michael D. “ Financial Crises, Banking Crises, Stock Market Crashes and the Money Supply: Some International Evidence, 1870-1933,” in Forrest Capie and Geoffrey E. Wood, eds. Financial Crises and the World Banking System (St. Martin’s Press, 1986), pp. 190-248. NOVEMBER/DECEMBER 1992 18 Chandler, Lester V. Benjamin Strong, Central Banker (Brook ings Institution, 1958). ________American Monetary Policy, 1928-1941 (Harper and Row, 1971). Canova, Fabio. “ The Sources of Financial Crisis: Pre- and Post-Fed Evidence,” International Economic Review (August 1991), pp. 689-713. Clark, Michelle A. “Are Small Rural Banks Credit- Con strained? A Look at the Seasonal Borrowing Privilege in the Eighth Federal Reserve District,” this Review (May/June 1992), pp. 52-66. Clark, Truman A. “ Interest Rate Seasonals and the Federal Reserve,” Journal of Political Economy (February 1986), pp. 76-125. Kemmerer, Edwin W. Seasonal Variations in the Relative De mand for Money and Capital in the United States, National Monetary Commission, Senate Document No. 588, 61 Cong. 2 Sess. Washington: Government Printing Office, 1910. Kindleberger, Charles P Manias, Panics, and Crashes, revised . ed. (Basic Books, 1989). Mankiw, N. Gregory, Jeffrey A. Miron and David N. Weil. “ The Adjustment of Expectations to a Change in Regime: A Study of the Founding of the Federal Reserve,” American Economic Review (June 1987), pp. 358-74. Miron, Jeffrey A. “ Financial Panics, the Seasonality of the Nominal Interest Rate, and the Founding of the Fed," American Economic Review (March 1986), pp. 125-40. Dewald, William G. “ The National Monetary Commission: A Look Back,” Journal of Money, Credit, and Banking (Novem ber 1972), pp. 930-56. Schwartz, Anna J. “ Real and Pseudo-financial Crises,” in For rest Capie and Geoffrey E. Wood, eds. Financial Crises and the World Banking System (St. Martin’s Press, 1986), pp. 11-31. Dwyer, Gerald P. Jr. and R. Alton Gilbert, “ Bank Runs and Private Remedies,” this Review (May/June 1989), pp. 43-61. Shiller, Robert. “ Can the Fed Control Real Interest Rates?” in Stanley Fisher, ed. Rational Expectations and Economic Policy (University of Chicago Press, 1980), pp. 117-68. Federal Reserve Board. Annual Report of the Federal Reserve Board (1914). Sprague, O.M.W. History of Crises under the National Banking System, U.S. National Monetary Commission, Senate Docu ment No. 538, 61 Cong. 2 Sess. Washington: Government Printing Office, 1910. Fishe, Raymond P. H. “ The Federal Reserve Amendments of 1917: The Beginning of a Seasonal Note Issue Policy,” Journal of Money, Credit, and Banking (August 1991), pp. 308-26. Fishe, Raymond P. H. and Mark Wohar. “ The Adjustment of Expectations to a Change in Regime: Comment,” American Economic Review (September 1990), pp. 968-76. Friedman, Milton and Anna J. Schwartz. A Monetary History of the United States, 1867-1960 (Princeton University Press, 1963). Gorton, Gary. “ Clearinghouses and the Origin of Central Banking in the United States,” The Journal of Economic History (June 1985), pp. 277-83. Hamilton, James D. “ Monetary Factors in the Great Depres sion,” Journal of Monetary Economics (March 1987), pp. 145-69. Holland, A. Stephen and Mark Toma. “ The Role of the Feder al Reserve as ‘Lender of Last Resort' and the Seasonal Fluctuation of Interest Rates,” Journal of Money, Credit, and Banking (November 1991), pp. 659-76. FEDERAL RESERVE BANK OF ST. LOUIS Timberlake, Richard. The Origins of Central Banking in the United States (Harvard University Press, 1978). ________“ The Central Banking Role of Clearinghouse As sociations,” Journal of Money, Credit, and Banking (Febru ary 1984), pp. 1-15. Trescott, Paul B. “ Federal Reserve Policy in the Great Depression: A Counterfactual Assessment,” Explorations in Economic History (July 1982), pp. 211-20. West, Robert Craig. Banking Reform and the Federal Reserve 1863-1923 (Cornell University Press, 1977). Wheelock, David C. The Strategy of Consistency of Federal Reserve Monetary Policy, 1924-1933 (Cambridge University Press, 1991). ________“ Monetary Policy in the Great Depression: What the Fed Did, and Why,” this Review (March/April 1992), pp. 3-28. Wicker, Elmus. Federal Reserve Monetary Policy, 1917-1933 (Random House, 1966). 19 Alison Butler Alison Butler is an economist at the Federal Reserve Bank of St. Louis. Lora Holman and Leslie Sanazaro provided research assistance. Is The United States Losing Its Dominance in HighTechnology Industries? T h e RECENT PO LITIC AL SEASON once again focused attention on high-technology industries and U.S. competitiveness. Many politicians bemoan the loss o f dominance in high-technology industries by the United States.1 The statistics they use to support their argument include the loss o f U.S. global market share in high-technology products, the declining U.S. balance-of-payments surplus in high-technology industries and the persistent balance-of-payments deficit w ith Japan. Others argue that the U.S. demise has been greatly exaggerated. They point out that labor productivity in the United States remains greater than in other industrialized countries and that the United States spent m ore than tw ice as much in absolute terms as other countries on research and developm ent (R&D). In fact, the evidence is mixed. Although the United States no longer dominates high-technology industries as it did in the 1950s and 1960s, much o f that is due to the economic grow th o f Japan ’This attitude can be seen, for example, in the hearing Factors Affecting U.S. Competitiveness (see U.S. Congress, 1992) and in articles such as “ America’s High-Tech Decline,” in Foreign Policy (see Ferguson, 1989). and Germany rather than to a decline in U.S. high-technology industries. As per capita output in these countries converges, one w ould also expect indicators in high-technology industries to also begin converging. Some indicators, however, suggest that the United States places a relatively smaller emphasis on R&D and education than do Germany and Japan, the main U.S. competitors in the high-tech arena.2 As a result, fe w clear conclusions can be drawn. Th e goal o f this article is to provide a careful, albeit not comprehensive, analysis o f U.S. hightechnology industries. First, the paper discusses w h y high-technology industries are considered valuable to an econom y and presents evidence to support these arguments. Next, several perform ance indicators fo r high-technology industries in the United States are examined. W hen possible, these indicators are compared w ith similar indicators in Japan and Germany. The paper concludes w ith a discussion o f what these indicators predict fo r the future. 2AII statistics for Germany refer to the former Federal Republic of Germany. NOVEMBER/DECEMBER 1992 20 W H A T ARE HIGH-TECHNOLOGY INDUSTRIES? The term high tech is often used, but rarely defined. The Organisation fo r Economic CoOperation and Developm ent (OECD) defines high-technology industries as having the fo llo w ing characteristics: • the need fo r a strong R&D effort; • strategic importance fo r governments; • very rapid product and process obsoles cence; • high-risk and large capital investments; and • a high degree o f international cooperation and competition in R&D production and w orldw id e marketing.3 Unfortunately, although this definition is important in isolating the general industries, these characteristics are too general to be used to classify firm s fo r statistical purposes. Th e OECD uses the ratio o f R&D expenditure to production costs (the R&D intensity) o f an industry, which is the defining characteristic o f high-technology fo r which data are available.4 According to this criterion, the top six R&Dintensive industries in the main 11 countries are aerospace, office machines and computers, electronics and components, drugs, instruments and electrical machinery.5 These industries had an average R&D intensity o f 11.4 percent in 1980, com pared w ith an average o f approximately 4.0 percent fo r all manufacturing industries. ARE HIGH-TECHNOLOGY INDUSTRIES MORE IM P O R T A N T TH A N O THER INDUSTRIES? The special concern expressed about hightechnology industries suggests that these industries provide unique benefits absent from other manufac 3OECD (1986). In this paper we consider only high-technology manufacturing industries. Other sectors, such as banking services and insurance, could be considered high tech. 4Many variations of this definition are used, in part because some prefer to define high technology in terms of product classes, whereas others (including the OECD) define them in terms of industry classes. Because of data limitations, this paper uses the OECD classification unless stated otherwise. ^The main 11 countries, as classified by the OECD, are the United States, Japan, Germany, France, United Kingdom, Italy, Canada, Australia, Netherlands, Sweden and Belgium. The industry classifications differ somewhat from those used by the National Science Board (1991), which classifies the http://fraser.stlouisfed.org/ RESERVE BANK OF ST. LOUIS FEDERAL Federal Reserve Bank of St. Louis turing industries. These benefits result from the relatively higher amount o f innovation in these industries and the subsequent effect on em ploy ment, wages, productivity and econom ic growth. High-Technology Industries and Econom ic Innovation As discussed in the preceding section, hightechnology industries are R&D intensive by definition. Innovation, w hich takes an invention and transforms it into a product or process that a firm can sell or use, generally results from R&D expenditures. Innovations can be broadly divided into three types: process, final product and intermediate product innovations. A process innovation is one that im proves the production technique o f a product—fo r example, Henry Ford’s use o f the m oving assembly line to massproduce automobiles. This innovation dramatically low ered the cost o f producing automobiles and significantly increased auto production. An important distinguishing feature o f process innovation is that it directly increases the productivity o f one or more factors o f production (capital, labor, energy and materials).6 In fact, experts argue the following: process innovation tends to “ have a bigger effect on an industry’s ow n rate o f productivity increase than does product R&D.”7 A final product innovation, in contrast, does not increase productivity directly; instead, it introduces a new product or a variation o f an existing product that individuals consume. An example o f a final product innovation is the refrigerator, which replaced the icebox. Final product innovations generally have a positive effect on the quality o f life—for example, refrigerators involve much less maintenance than iceboxes, leaving m ore fre e time fo r other activities—and have a stimulative effect on output. following as high-technology industries: industrial chemicals; drugs and medicines; engines and turbines; office and computing machinery; communication equipment; aerospace; and scientific instruments. This difference is likely due to the fact that the OECD measure is an average over 11 countries in 1980. If a seventh industry were included by the OECD, it would be the automobile industry. The industries classified as high technology according to the OECD classifications have not changed during the course of the sample. 6Mansfield proposes using total factor productivity, which is the most general measure of productivity. For a discussion of this measure, see Mansfield (1990). 7See Mansfield (1988). Rosenberg (1982) also stresses the im p ortan ce of im p rovem e nts to an in itia l in n o va tio n . 21 Final product innovations, how ever, do not increase factor productivity directly.8 An intermediate product innovation results in a new product that is used to produce another good. In other words, the new product is not consumed by individuals, but rather is used by firms. Productivity increases in industries that use intermediate product innovation. For example, a new machine tool that significantly reduces the time it takes to produce furniture w ould be considered an intermediate product innovation. Output increases both because the tool industry has a new product and because productivity increases in the furniture industry. Other industries may also benefit if they can adapt the innovation fo r their ow n use or if the innovation leads to other innovations. For example, a w ood lathe might suggest the possibility o f a metal lathe. Innovations that enable a country to produce m ore output w ith the same amount o f input increase productivity and therefore aggregate output. Social Benefits f r o m Innovation Economists have long believed that innovation has a positive effect on economic growth. Schumpeter (1950) argued that the process o f creative destruction (the creation o f new products that replaced existing products) drives economic growth. Others, such as Solow (1957), estimated the effects o f technological change on economic growth. Recently, new grow th theorists, such as Romer (1990), Aghion and Howitt (1990) and Grossman and Helpman (1991), have explored the determinants of technological change and its effects on economic growth. They identify endogenous technological spillovers as the prim ary deter minant o f economic grow th.9 Productivity grow th is a prim ary determinant o f a country’s standard o f living. As labor 8Because the measurement techniques currently used do not measure changes in quality or nonmarket activities (such as household work), output may or may not increase as a result of the innovation. For a discussion of the problems in growth accounting and measuring the value of innovation, see Griliches (1979) or Grossman and Helpman (1991). 9ln this context, endogenous technical spillovers refer to the gains in knowledge associated with the process of innovation. For a more comprehensive discussion, see the sources cited in the text. ' “Situations exist that could make some workers worse off, however. An innovation that substitutes capital for labor may reduce employment in the industry adopting the innovation. Al though displaced workers may be worse off in the short run, the lower prices that result from an increase in productivity becomes m ore productive, wages rise.1 Some 0 economists have argued that technological innovation has a negative effect on employment. The United States, how ever, has had continued improvements in productivity over the last 100 years, w hereas its average unemployment rate has remained essentially unchanged. This suggests that if productivity increases have a negative effect on employment, the effect is not permanent. Firms invest in R&D because they hope to earn an above-average rate o f return on any innovation. The amount o f innovation (if any) that results from R&D is always uncertain, so there is no guarantee o f a return. As a result, firms expect a greater-than-average rate o f return to compensate them fo r the risk associated with R&D. This return is the cash flo w earned over time from an innovation, which includes revenue from product sales, as w ell as earnings from the leasing or sale o f the new technology. The greater the expected return, the greater the incentive to invest in R&D. The benefits to society from innovation, how ever, can be substantially larger than the return earned by the innovating firm. The social rate o f return measures this benefit. The social rate o f return equals the private rate o f return plus any technological spillovers, that is, any benefits from an innovation that are not appro priated by the innovator.1 Because the social 1 rate o f return usually exceeds the rate o f return the innovator earns, there tends to be a lessthan-optimal level o f investment in R&D and new technologies.1 As a result, most countries enact 2 policies that encourage innovation. The most common w ay is through patent and copyright protection, w hich im prove the likelihood that the innovator w ill earn an above-normal return on an innovation.1 3 could increase consumers’ purchasing power and increase demand in other industries. As a result, employment could rise in those other industries, leaving aggregate employment unchanged. Of course, there are likely adjustment costs associated with the shift in employment. For a discussion of this issue, see Baumol and McLennan (1985). 11For a discussion of alternative measures of spillovers, see Griliches (1979). 12The possibility of a significant difference between the social and private rate of return exists because it is impossible to control the flow of information generated by an innovation. See Arrow (1962) for a careful discussion of this problem. 13See Butler (1990) for a discussion of the relationship between property rights and innovation. NOVEMBER/DECEMBER 1992 22 Governments often provide other incentives, both explicit and implicit, to innovate. Tax cred its, fo r example, are offered fo r R&D in many industrial countries.1 Many governm ents also 4 provide funds fo r R&D, both directly and indirectly, by subsidizing education. In fact, the United States, though it has no explicit industrial policy, publicly finances nearly half o f all R&D in the country (an estimated 43.5 percent in 1991). International Effects o f Innovation The benefits o f an innovation, particularly its indirect benefits, are not restricted geographically. T o the extent that knowledge generated from innovation is internationally available, countries benefit from all innovation, regardless o f w here it originates. In general, the gains from innovation are greater in the innovating country than in countries that im port the technology because o f the increased jobs and higher wages associated w ith high-technology industries. In addition, the innovating country benefits from earnings on the sale or lease o f new technology to other countries. The extent to which a nation benefits from domestic innovation depends greatly on the degree to which that nation is compensated fo r the innovation-related knowledge and technologies that flo w abroad. Technological innovation, particularly process innovation, has historically traveled slowly because o f international capital and labor immobility, as well as linguistic and cultural differences. Mansfield (1984) discusses another reason w h y process innovation disseminates slowly across borders: Firms are often unwilling to license new tech nology abroad because it is difficult to control the diffusion o f the technology in other countries. This licensing argument should not apply within firms, however. The recent growth o f multinational corporations, such as IBM and Toyota, in hightechnology industries has significantly increased the pace o f technological diffusion internationally. Regardless o f w hether impediments to the flo w o f information or technology exist, h ow ever, an innovating country still benefits from ,4Some empirical evidence of the effect of tax credits for R&D suggests that their importance in the United States may be modest. See Mansfield (1986) and Cordes (1989). A recent study by Hall (1992) finds stronger support for the effectiveness of R&D tax credits. ,5See Minasian (1962, p. 94). Recall that increasing the rate of productivity increases economic growth. Output may not necessarily increase if the result of an innovation is a substantial increase in leisure relative to hours worked. FEDERAL http://fraser.stlouisfed.org/ RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis innovation through both the private and social returns generated. I f process im provem ents to an initial innovation are made in the innovating country, the benefits o f the initial innovation are even greater over time fo r that country because o f increases in productivity. Evidence Regarding High-Technology Industries A 1960 National Bureau o f Economic Research conference was specifically designed to examine inventive activity, the activity that generates in novation. One o f the conference papers discussed the chemical, allied products and pharmaceutical industries betw een 1947 and 1957 and found that “ productivity increases are associated w ith invest ment in the im provem ent o f technology and the greater the expenditures fo r research and developm ent the greater the rate o f grow th o f productivity.” 1 M ore recent studies also found R&D 5 to be an important determinant o f innovation and productivity and th erefore econom ic grow th .1 6 In addition, researchers have found significant differences betw een the social and private rates o f return earned on innovations, supporting the view that the benefits to society from innovation are greater than those appropriated by firm s.1 7 Unfortunately, these results must be view ed w ith some skepticism because o f measurement and data problem s.1 8 Another w ay high-tech industries benefit a country directly is through their effect on wages and employment. In general, wages might be expected to be higher in innovating industries because producing and developing new products or implementing n ew processes initially requires a higher skill level. Existing wage and employment data in U.S. high-technology industries support this theory. In 1972, wages in high-technology industries w ere 16.7 percent higher than wages in all other manufacturing industries. By 1989, wages in high-technology industries w ere 24.7 percent higher.1 9 U.S. compounded annual em ploym ent grow th 16See, for example, Leonard (1971), Mansfield (1980) and Scherer (1982). 17See, for example, Mansfield (1981) and Bernstein and Nadiri (1988, 1989). ,8See Griliches (1979) and Grossman and Helpman (1991) for a discussion of these problems. 19This rise in wages could be due to an increase in the demand for skilled labor that exceeds the supply. See Katz and Murphy (1991). 23 rates betw een 1970 and 1989 w ere among the highest in the pharmaceuticals (3.3 percent) and aircraft (3.2 percent) sectors. Employment declined in most lower-technology U.S. industries; the largest declines w ere in ferrous metals (-2.4 percent) and other transport equipment (-2.5 percent).2 On average, the compounded annual 0 employment grow th rate fo r all manufacturing industries was 0.2 percent during this period. W age differences in high- and low-technology industries can be seen in many industrialized countries. In 1988, wages in high-technology industries fo r the Group o f Seven countries w ere on average 26.5 percent higher than wages in low-technology industries.2 1 TRENDS IN HIGH-TECHNOLOGY INDUSTRIES Both the theoretical and empirical evidence o f the benefits associated with innovation suggest that R&D-intensive industries are particularly valuable to a country. U.S. high-technology industries dominated the w orld market fo r most o f the postwar period. In the last tw o decades, how ever, this dominance appears to have deteriorated, as reflected by the declining U.S. share o f high-technology manufacturing output in the OECD since 1970. This section looks at several indicators o f current and future perform ance in high-technology industries for the United States.2 2 The next section compares some o f these indicators with those o f Japan and Germany. High-Technology Indicators f o r the United States Table 1 shows various statistics on U.S. R&D. High-technology manufacturing output increased by m ore than 50 percent in 10 years—from 20.0 20The compounded annual growth rate for the aircraft and other transport equipment industry was calculated for the period 1972-89. The other transport equipment category is transport equipment less shipbuilding, automotive and aircraft. 2,The Group of Seven countries are Canada, France, Germany, Italy, Japan, United Kingdom and United States. Because of data limitations, this wage comparison uses a broader definition of high technology. 22Some analysts use patent statistics as an indicator of inventive activity. According to Cockburn and Griliches (1988), however, "Data on R&D expenditures... are stronger measures of input to the process by which firms produce technical innovation than patents are of its ‘output.’ ” In addition, cross-country comparisons of patent statistics are often invalid because of varying standards across countries. percent o f total manufacturing output in 1980 to 30.4 percent by 1990.2 This marked increase 3 occurred at the expense o f other manufacturing industries. Manufacturing output as a percent o f gross domestic product (GDP) remained fairly constant over this period. Statistics on gross expenditures on R&D (GERD) and business expenditures on R&D (BERD), which are available over a longer period, provide mixed evidence on the behavior o f U.S. R&D. Figure 1 shows the components o f GERD fo r 1991, with BERD clearly being the largest component. BERD is divided to show the percent o f business R&D that is governm ent funded. Both GERD and BERD have risen in real (constant-dollar) terms since 1975. As a percent o f GDP, how ever, both GERD and BERD have fluctuated since 1970, falling until 1978, rising from 1978 to 1985 and declining since then.2 4 Many o f the fluctuations have been in defenserelated expenditures on R&D. Nondefense spending on R&D as a percent o f gross national product (GNP) increased slightly, from 1.6 percent in 1972 to 1.9 percent in 1989. An important caveat to these numbers results from the problem s associated w ith using an aggregate deflator (such as the GDP deflator) fo r R&D expenditure. One study found that because o f the inadequacies in the deflator used, real R&D expenditures in the period 1969-79 rose only 1 percent—not the 7 percent reported using a standard deflator.2 In fact, most evidence 5 suggests that R&D costs increase m ore rapidly than the R&D deflator.2 6 Another indicator o f R&D activity can be obtained by examining its components—basic research, applied research and development. According to the National Science Foundation, basic research is "research that advances scientific knowledge but does not have specific commercial objectives.”2 7 Thus patent statistics can be used to compare changes within a country over time, but not across countries. 23Table 1 uses the National Science Board definition of high technology. 24To get a sense of the magnitudes being examined, a 0.1 percent change in GDP in 1990 equals $5.5 billion. 25See Mansfield (1984). 26See Bureau of Labor Statistics (1989). Unfortunately, no standard series on an R&D deflator for the United States is available. Available estimates are not consistent with the series shown above. These series were used because they are internationally comparable. 27See National Science Board (1991). NOVEMBER/DECEMBER 1992 24 Table 1 High-Technology Indicators for the United States Year GERD (billions of 1982 dollars) 1970 1975 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 65.27 61.93 74.90 78.38 81.69 87.51 95.86 104.62 106.85 108.62 112.28 114.66 114.65 N.A. GERD as a percent of GDP BERD (billions of 1982 dollars) BERD as a percent o f GDP 2.72 2.32 2.39 2.45 2.62 2.71 2.78 2.93 2.91 2.87 2.83 2.82 2.80 2.82 43.02 40.79 51.93 55.12 58.65 62.82 69.45 75.96 76.78 78.43 80.63 80.44 79.24 N.A. 1.79 1.53 1.66 1.72 1.88 1.95 2.01 2.13 2.09 2.07 2.04 1.98 1.93 1.95 Hi-tech m anufactures as a percent of total m anufacturing output N.A. N.A. 20.0 20.7 22.4 22.9 24.5 25.5 27.0 27.9 28.7 29.6 30.4 N.A. M anufacturing as a percent of GDP in 1982 dollars1 21.12 20.54 21.52 21.25 20.37 20.87 21.76 21.75 21.78 22.24 23.25 22.73 N.A. N.A. DEFINITIONS: GERD—Gross expenditure on research and development BERD— Business expenditure on research and development GDP—Gross domestic product ’ Two different deflators were used for this calculation. SOURCE: OECD, Science and Technology Statistics (1992); National Science Board; Economic Report of the President. Applied research is the application o f new scien tific knowledge to determine how a specific problem or need can be met. For industry this includes specific commercial objectives. Development, on the other hand, is the “ systematic use o f the knowledge or understanding gained from research directed tow ard the production o f useful materials [and] devices... including design and developm ent o f prototypes and processes.”2 8 Thus research is necessary fo r invention, but developm ent is required to bring an invention to market. That is, developm ent is required fo r innovation. Between 1960 and 1990, the allocation o f R&.D expenditures within these categories remained essentially unchanged. Some commentators have expressed concern about the lack o f relative increase in development expenditures. They believe that such expenditures are critical fo r fu ture technological progress. 28See National Science Board (1991). 29Cited in Mansfield (1988). FEDERAL RESERVE BANK OF ST. LOUIS This argument is particularly relevant concerning n ew and im proved production processes, which have a m ore direct effect on productivity. For example, the President’s Commission on Industrial Competitiveness states the following: "It does [the United States] little good to design state-of-theart products, if w ithin a short time our foreign competitors can manufacture them more cheaply.”2 9 A ccording to Mansfield (1988), despite these criticisms there is nothing to "indicate that there was any perceptible increase betw een 1976 and 1985 in the proportion o f [U.S. firms'] R&D expenditures devoted to new or im proved processes.” Overall, these statistics appear to contradict the idea that R&D expenditures by U.S. hightechnology industries declined in the 1980s. Since 1985, how ever, R&D as a percent o f GDP has been declining. This relative decline may be 25 Figure 1 Components of GERD, 1991 Government R&D Private Nonprofit R&D 12.1% 3.0 % Higher Education R&D 15.9% Privately Funded (48.4%) Government Funded (20.6%) a cause fo r concern, fo r both productivity grow th and perform ance in high-technology industries.3 0 AN INTERNATIONAL COM PARISON IN HIGH TECHNOLOGY Much o f the concern about U.S. high-technology perform ance has focused on U.S. indicators relative to those o f other countries. This section compares U.S. high-technology perform ance w ith that o f Japan and Germany.3 Because o f 1 the sheer size o f the United States, its total R&D expenditures are much greater than those o f Germany or Japan. For example, using OECD purchasing-power parities to convert to dollars, GERD in 1990 was $66 billion in Japan, $28 billion in Germany and $151 billion in the United States.3 2 30The effects of R&D are estimated to have about a two-year lag on productivity. A longer lag is associated with basic research (Bureau of Labor Statistics, 1989). 31ln 1989, the average expenditure on GERD as a percent of GDP for these three countries was 2.9 percent, compared with the OECD average for reporting countries, which was 1.7 percent. This statistic excludes Australia, Belgium and Portugal. If they were included, the number would likely be slightly lower. 32Purchasing-power parities measure the number of U.S. dollars required in each country to buy the same represen tative basket of final goods and services that cost $100 in the United States. 33See Scherer (1982) and Mansfield (1988). Mohnen, Nadiri and Prucha (1986) compared the rate of return on R&D in the three countries and also found that the return was lowest in the United States. BERD 69.0% As a result, the United States can benefit from the additional resources it can spend on R&D, to the extent that its R&D is at least as productive as R&D in the other tw o countries. Several researchers have expressed concern regarding the productivity o f U.S. R&D, particularly in regard to other countries.3 One reason fo r this 3 is the high percentage o f R&D funded by the government. Studies have found that governmentfunded private R&D is less productive than privately funded business R&D.3 4 Figures 2 and 3 show GERD and BERD as a percent o f GDP fo r the three countries. Through out most o f the last 20 years, the United States has had higher GERD/GDP and BERD/GDP ratios than Germany and Japan.3 From 1964 to 1990, 5 Japan and Germany each increased its GERD as a percent o f GDP by approximately 100 percent. From 1980 to 1990 (1981 for Germany), however, 35Because of the problems associated with using standard deflators for R&D, these numbers could be somewhat misleading. If, for example, R&D costs rose faster in Japan relative to GDP than in the other two countries, the actual ratio for Japan would be relatively lower. The only attempt to calculate R&D deflators across countries was done by the OECD (1979) for the period 1967-75. The results suggested that the R&D deflator moved together for these countries. Unfortunately, the deflators for Japan and Germany were not directly comparable, and a deflator was not calculated for the United States. As a result, it is difficult to predict whether the price of R&D would move differently across countries. A worldwide program has attempted to produce international comparisons of variables in the National Income and Product Accounts across countries. See Kravis and Lipsey (1990) for a summary and update on this program. ^S ee Griliches (1986, 1987). NOVEMBER/DECEMBER 1992 26 Figure 2 GERD as a Percent of GDP Percent 3.3 2.7 2.3 1.9 1.7 1970 72 74 76 78 80 82 84 86 88 U.S. data for 1971 and German data for 1980 were not available. Figure 3 BERD as a Percent of GDP Percent German data for 1976,1978, and 1980 were not available. FEDERAL RESERVE BANK OF ST. LOUIS 1990 27 Figure 4 Scientists and Engineers Working in R&D per 10,000 Labor Force Numbers of workers The figure for Germany increased in 1979 because the 1979 survey includes small and medium enterprises not surveyed in 1977. Data for Germany in 1978,1980,1982,1984,1986, and 1987 are estimated. SOURCE: National Science Board, 1991. Japan’s increase in GERD as a percent o f GDP was 40.8 percent, much higher than the 15.6 percent increase in Germany or the 17.2 percent increase in the United States. BERD as a percent o f GDP also increased steadily in Japan and Germany over the last 23 years, w hereas U.S. spending fluctuated during the same period. By 1990 the ratios w ere essentially equal in the three countries. The lack o f variability in other countries could be attributed to the low level o f military spending in these other countries. For example, in 1989, 28.9 percent o f U.S. R&D was defense related, compared with 4.6 percent o f Germany’s R&D and less than 1.0 percent o f Japan’s R&D. Another measure o f innovative activity is the number o f science and engineering (S&E) personnel relative to the total w ork fo rce (see figure 4). Throughout the sample, the United States has em ployed m ore S&E personnel per 10,000 w orkers than either Germany or Japan. Although the number o f S&E workers relative to the labor force has risen on average in all three countries, the increase has been substantially greater in Japan and Germany. As a result, this difference among the three countries has narrow ed considerably. The technological balance o f payments (shown in table 2) measures the difference betw een receipts and payments related to earnings on technology and is an indicator o f the degree to which a country is an exporter or im porter o f technology. This measure includes revenues associated w ith the use o f patents, licenses, trademarks, designs, inventions, know-how and closely related technical services. This balance has been steadily increasing for the United States since 1969 (the first year data are available), showing that earnings on U.S. technological exports continue to significantly exceed U.S. NOVEMBER/DECEMBER 1992 28 Table 2 International Comparisons in High-Technology Indicators High-tech m an ufactures’ share of total m anufacturing o utpu t1 (percent) Technology balance of paym ents2 (billions of U.S. dollars) Year U.S. Japan Germ any U.S. Japan Germ any 1971 1975 1980 1985 1986 1987 1988 1989 1990 N.A. N.A. 20.0 25.5 27.0 27.9 28.7 29.6 30.4 N.A. N.A. 16.3 24.6 26.4 29.5 32.9 34.5 35.1 N.A. N.A. 16.1 20.4 20.8 20.9 21.3 20.6 20.3 2.30 3.83 6.36 5.10 6.19 7.70 8.80 9.57 12.65 -0 .4 4 -0 .3 7 -0 .3 2 -0 .2 7 -0 .1 7 -0 .3 2 -0 .3 2 0.00 -0 .1 7 -0 .3 2 -0 .4 5 -0 .5 6 -0 .6 5 -0 .5 5 -0 .6 0 -0 .6 5 -1 .0 5 -0 .9 3 'National Science Board definition. 2OECD purchasing-power parties are used to convert yen and deutsche marks to dollars. SOURCE: National Science Board, Science and Engineering Indicators (1991); OECD, Science and Technology Statistics. payments fo r technological information. Japan and Germany have both, on average, increased their exports o f technology during this period. By 1989 Japan exported as much technology as it imported, suggesting that, contrary to popular perception, Japan is becoming an innovator in its ow n right. The increased importance o f high-technology industries in these countries can also be seen by looking at international comparisons in hightechnology indicators (see table 2).3 High6 technology manufactures as a percent o f total domestic manufacturing output rose by more than 100 percent in Japan betw een 1980 and 1990. This ratio increased in the United States and Germany, although by significantly less— 52.0 percent and 26.1 percent, respectively. In the last 10 years, the market fo r hightechnology products in OECD countries has increased by 117 percent in constant (1980) dollars. Output in high-technology industries rose in Japan, Germany and the United States during this period. H ow ever, despite the 92.9 percent increase in the value o f its high-technology output from 1980 to 1990, the U.S. share o f 36The rest of this section uses the National Science Board definition of high-technology industries. Because the board changed from the OECD definition in its 1991 report, a longer consistent time series for these variables is not available. FEDERAL http://fraser.stlouisfed.org/ RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis global high-technology m anufacturing declined by 11.1 percent. Germany’s share declined 20.3 percent during this period as well, and the share o f the remaining OECD countries as a w hole declined 13.6 percent (see table 3). As these countries lost market share, Japan’s market share increased 58.7 percent. Thus although the United States remains the major producer o f high-technology goods, it no longer dominates all high-technology industries. The composition o f high-technology goods production also changed markedly during these 10 years. For example, the U.S. share o f global production o f office and computing machinery fell by 15.2 percentage points, whereas Japan’s share rose an offsetting 15.5 percentage points. Similarly, the U.S. share o f radio, television and communications equipment declined 6.0 percentage points from 1980 to 1990, whereas Japan’s share o f this global market increased by 15.6 percentage points. This suggests that the United States has faced increased competition in these industries. On the other hand, its position in industrial chem icals, drugs and m edicines rem ains essentially unchanged, and its market share o f scientific instruments increased somewhat. 29 Table 3 Country Share of Global Market for High-Tech Manufactures , by Industry: 1980-90 (in percent) 1980 1981 1982 1983 1984 1985 1986 1987 1988 (est.) 1989 (est.) 1990 (est.) 40.4 18.4 11.8 39.5 19.7 11.7 38.9 20.4 11.8 37.8 21.6 11.8 37.9 23.3 11.3 36.3 23.6 12.0 36.9 23.4 11.5 37.5 25.1 10.5 37.0 26.5 10.1 36.0 28.4 9.5 35.9 29.2 9.4 32.7 16.1 16.2 33.1 14.4 16.9 29.8 15.3 17.9 29.2 14.0 19.1 28.0 14.1 19.5 25.8 13.4 20.4 28.5 12.1 20.4 31.4 13.1 18.5 31.2 12.7 18.7 32.2 13.4 18.8 32.5 14.1 18.4 29.6 21.2 13.1 29.6 21.7 13.1 30.3 22.1 12.5 30.3 22.0 12.5 30.4 21.2 12.7 30.0 20.7 12.3 30.4 20.4 12.1 31.4 19.9 11.4 31.4 20.1 11.5 30.8 20.1 11.4 29.2 20.3 10.9 44.2 18.4 11.3 37.9 16.1 9.9 35.0 17.9 9.0 33.0 18.8 9.4 35.4 18.0 10.3 34.8 17.0 11.2 35.4 14.9 10.9 35.4 15.7 11.2 35.8 15.5 10.7 35.2 15.8 10.8 34.9 15.3 11.6 50.0 22.0 6.5 49.0 23.0 7.4 49.1 24.0 7.0 45.2 27.2 7.0 44.0 27.5 7.4 39.6 30.2 8.3 37.8 30.8 8.0 38.1 31.8 7.1 37.3 33.3 6.6 35.6 34.6 5.5 34.8 37.5 5.4 36.6 26.4 12.0 34.8 30.5 11.4 35.0 30.7 11.4 34.0 32.2 11.1 33.8 35.5 9.8 32.9 34.0 11.3 32.8 33.0 11.6 32.3 36.5 10.3 31.5 39.3 9.6 29.9 42.9 9.5 30.6 42.0 10.0 57.6 2.2 4.8 56.4 2.4 5.3 56.6 2.3 6.0 55.8 2.4 5.4 58.7 2.5 5.0 57.9 2.9 5.0 59.5 2.5 4.4 58.7 2.8 4.6 59.2 3.2 4.7 56.4 3.6 4.6 55.9 3.6 4.8 49.1 17.6 11.4 49.0 19.2 10.8 50.5 18.1 10.2 50.0 19.0 9.8 50.4 19.0 9.8 48.4 19.7 10.8 48.4 18.9 11.1 50.8 18.1 11.1 51.5 16.2 11.4 52.7 16.1 10.8 53.4 15.4 11.1 High-tech m anufactures United States Japan Germany Industrial chem icals United States Japan Germany Drugs and m edicines United States Japan Germany Engines and turbines United States Japan Germany O ffice and com puting machinery United States Japan Germany Radio, TV and com m unication equipm ent United States Japan Germany Aircraft United States Japan Germany Scientific instrum ents United States Japan Germany NOTES: Total shipments by OECD countries are used as a proxy for global output. Shares represent each country’s shipments as a percentage of OECD shipments. Germany refers to the former Federal Republic of Germany. SOURCE: See National Science Board (1991). NOVEMBER/DECEMBER 1992 30 Figure 5 International Com parisons in Education Average mathematics test scores for eighth-grade students, 1981-1982 8069 70- 605040- 30 60 60 III ilk Test Average Arithmetic Algebra Geometry Measurement ■ Students in Japan and Hong Kong were attending the seventh grade. W heth er the United States w ill continue to be a w orld leader in high-technology manufacturing is unclear. Although its relative position in hightech manufacturing has slipped in the past decade, so have those o f many other industrialized countries, w ith Japan gaining most o f the lost market share.3 On average, the U.S. decline was less 7 than those o f European countries. Furthermore, it seems unlikely that any country could maintain the degree o f dominance that the United States enjoyed in the early postwar period. Even if the United States had continued to increase its hightech manufacturing at the same rate as in the postwar period, the entry o f other countries into high-technology industries would have guaranteed a loss o f market share fo r the United States. Hence the recent loss o f w orld market share itself is not cause fo r alarm, given the significant output increase in U.S. high-tech industries during the period. 37The United Kingdom substantially increased its share of several high-technology industries and its share overall. Data were not presented for the United Kingdom, however, because the R&D series is incomplete and its R&D expenditure as a percent of GNP is significantly lower than for the countries presented (2.0 in 1989). http://fraser.stlouisfed.org/ RESERVE BANK OF ST. LOUIS FEDERAL Federal Reserve Bank of St. Louis Japan □ United States 1! The overall test average is a weighted average. SOURCE: U.S. Department of Education. Statistics Other Countries W H A T IS LIKELY FOR THE FUTURE? Concern remains that hidden in these trends is the future decline o f U.S. high-technology industries. Given the higher skills necessary fo r both employment in high-technology industries and success in R&D, the education level and scholastic performance o f U.S. students (relative to those in other countries) is coming under increased scrutiny. A possible indicator o f future performance in hightechnology industries, educational performance comparisons, is presented in figures 5 and 6. International education comparisons are extremely difficult because o f the differences in educational systems. As a result, these statistics should be view ed only as suggestive.3 8 An international assessment study comparing students from 18 countries found significant 38For the complete comparisons, see original source. 31 Figure 6 international Comparison in Education Science test scores for fifth- and ninth grade students—1983-1986 2 2 -i ■ □ United States B Grade 5 Japan Other Countries Grade 9 Tests were administered between 1983 and 1986. The average age in years and months is 10:9 for fifth-grade students and 14:10 for ninth-grade students. SOURCE: International Association for the Evaluation of Educational Achievement. differences in the performances o f U.S. and Japanese students (Germany did not participate).3 9 In fact, the U.S. ranking in geometry was eleventh, and its ranking in measurement was tw elfth out o f the 12 industrialized countries that participated in the tests. In a different study o f science test scores fo r 10- and 14-year-olds, the U.S. students ranked significantly lower than Japanese students.4 0 On average, U.S. students generally did poorer than students from other countries.4 1 Several other studies that focus only on U.S. students have found a general decline in their 39The participating countries were Belgium, Canada, England and Wales, Finland, France, Hong Kong, Hungary, Israel, Japan, Luxembourg, Netherlands, New Zealand, Nigeria, Scotland, Swaziland, Sweden, Thailand and the United States. 40The participating countries were Austrailia, Canada (English), England, Finland, Hong Kong, Hungary, Italy, Japan, South Korea, Netherlands, Norway, Philippines, Poland, Singapore, Sweden, Thailand and the United States. perform ance in mathematics and science during the 1970s with some improvement in the 1980s.4 2 These statistics suggest that the United States may have difficulty m eeting the demand fo r the jobs associated with high-technology industries because these jobs require an increasingly high level o f skill. Table 4 shows the percentage o f higher education degrees awarded to US. citizens and permanent U.S. residents that w ere awarded in science and engineering. The percentage o f master’s degrees and doctorates in science and ■"Unfortunately, only a small amount of research has occurred in this area. For a discussion of several other comparative studies, which reached similar conclusions, see National Science Board (1991), OECD (1992) and the November 21, 1992, issue of The Economist. 42See, for example, National Science Board (1991, 1989). NOVEMBER/DECEMBER 1992 32 Table 4 Degrees Awarded to U.S. Citizens and U.S. Permanent Residents in the United States for Selected Years 1977 Baccalaureate degrees in science and engineering as a percent of total Science and engineering baccalaureate degrees awarded per 100,000 population Total baccalaureate degrees awarded 1979 1981 1985 1987 1989 1990 40.1 39.8 39.1 35.5 35.1 33.7 33.6 166.2 161.5 157.0 142.5 140.2 136.1 138.3 912,484 913,487 924,246 961,619 974,940 1,003,714 1,035,598 Master’s degrees in science and engineering as a percent of total 25.0 25.0 25.3 25.9 26.6 25.8 25.1 Per capita science and engineering master’s degrees awarded per 100,000 population 34.1 31.4 30.2 28.2 28.6 28.9 28.9 262,268 278,927 290,345 Total master’s degrees awarded Doctorate degrees in science and engineering as a percent of total Per capital science and engineering doctorate degrees awarded per 100,000 population Total doctorate degrees awarded 300,896 282,648 274,740 260,261 53.2 54.0 54.8 56.2 56.6 57.7 57.2 6.6 6.4 6.3 5.8 5.7 5.8 5.9 27,487 26,784 26,342 24,694 24,561 25,024 25,844 SOURCE: National Science Foundation. engineering being awarded has increased since 1977. On a per capita basis, however, the number of people getting bachelor's and advanced degrees, both in science and engineering and overall, has generally declined, although some improvement has occurred in the last three years. Another factor that could play a pivotal role in determining the future o f R&D investment in the United States is the recently proposed cuts in military spending. As previously discussed, defense-related expenditures on R&D in 1987 w ere responsible fo r 65.5 percent o f governmentfunded R&D and 28.9 percent o f total R&D in the United States, which is a significantly larger portion than allocated in oth er countries. Analysts are concerned that a loss o f these 43For a study that examined the effect of a cut in federally financed R&D in the energy sector, see Mansfield (1984). “''See, for example, the National Defense Authorization Act (1992). FEDERAL http://fraser.stlouisfed.org/ RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis funds could cause the U.S. share o f global output in high technology to continue its decline. O f course, firm s or the governm ent could replace all o f the military-funded R&D w ith other R&D funding.4 A significant decline 3 in R&D expenditure, however, would likely re duce U.S. innovation both absolutely and relative to other countries and could have an adverse effect on U.S. high-technology industries. Legislation has already been proposed in Congress to ensure governm ent’s comm itm ent to R&D; one proposal uses defense-funded scientists to develop commercial technologies.4 4 A t this point, determining either the magnitude o f any R&D cuts or the response o f the nondefense governm ent and private sectors to these cuts is essentially impossible. 33 CONCLUSION High-technology industries have a significant positive effect on economic grow th because o f their high rates o f innovation. During the 1980s, production o f high-technology products in OECD countries increased by 117 percent. The continued increase in resources devoted to R&D in Germany, Japan and the United States reflects the importance o f high-technology industries. Although high-technology output as a percent o f GDP has decreased somewhat in the United States during the last few years, it remains higher than it was in 1970. During this period, Japan and Germany, which initially spent a much smaller portion o f GDP on R&D than did the United States, had significant grow th in R&D expenditures and high-technology output. Thus although the commitment o f resources fo r R&D relative to the size o f each economy has essentially equalized in the three countries, the United States still spends the most in absolute terms on R&D and has the largest market share in high-technology industries. The extent to which the United States can exploit its size advantage depends on how productive U.S. R&D is relative to these other countries. Unfortunately, little research has been conducted on this topic, so although some experts have expressed concern about the productivity o f U.S. R&D, evidence remains inconclusive. This important area o f research has yet to be fully explored. Neverthe less, the increasing importance o f high-technology industries suggests that a continued presence in these industries w ill help maintain high-wage/ high-skill jobs and continued economic growth fo r the United States. REFERENCES Aghion, Philippe, and Peter Howitt. “ A Model of Growth Through Creative Destruction,” NBER Working Paper No. 3223 (January 1990). Arrow, Kenneth J. “ Economic Welfare and the Allocation of Resources for Invention,” The Rate and Direction of Inventive Activity: Economic and Social Factors (Princeton University Press, 1962), pp. 609-26. Baumol, William J., and Kenneth McLennan. “ U.S. Productivity Performance and Its Implications,” in William J. Baumol and Kenneth McLennan, eds. Productivity Growth and U.S. Competitiveness, (Oxford University Press, 1985), pp. 3-28. Bernstein, Jeffrey I., and M. Ishaq Nadiri. “ Research and Development and Intra-Industry Spillovers: An Empirical Application of Dynamic Quality,” Review of Economic Studies (April 1989), pp. 249-69. _______ . “ Interindustry R&D Spillovers, Rates of Return, and Production in High-Tech Industries,” American Economic Review, Papers and Proceedings (May 1988), pp. 429-34. Bureau of Labor Statistics. The Impact of Research and Development on Productivity Growth, Bulletin 2331 (GPO, September 1989). Butler, Alison. “ The Trade-Related Aspects of Intellectual Property Rights: What Is At Stake?” this Review (November/December 1990), pp. 34-46. Cockburn, lain, and Zvi Griliches. “ Industry Effects and Appropriability Measures in the Stock Market's Valuation of R&D and Patents,” American Economic Review, Papers and Proceedings (May 1988), pp. 419-23. Cordes, Joseph J. “ Tax Incentives and R&D Spending: A Review of the Evidence,” Research Policy (June 1989), pp. 119-33. Ferguson, Charles H. “ America’s High-Tech Decline,” Foreign Policy (Spring 1989), pp. 123-44. Griliches, Zvi. “ R&D and Productivity: Measurement Issues and Econometric Results,” Science (July 1987), pp. 31-35. _______ . “ Productivity, R&D, and Basic Research at the Firm Level in the 1970's,” American Economic Review (March 1986), pp. 141-54. _______ . “ Issues in Assessing the Contribution of Research and Development to Productivity Growth,” Bell Journal of Economics (Spring 1979), pp. 92-116. Grossman, Gene M., and Elhanan Helpman. Innovation and Growth in the Global Economy (MIT Press, 1991). Hall, Bronwyn. “ R and D Tax Policy During the Eighties: Success or Failure?” NBER Working Paper No. 4240 (December 1992). Katz, Lawrence F., and Kevin M. Murphy. “ Changes in Relative Wages, 1963-1987: Supply and Demand Factors,” NBER Working Paper No. 3927 (December 1991). Kravis, Irving B., and Robert E. Lipsey. “ The International Comparison Program: Current Status and Problems,” NBER Working Paper No. 3304 (March 1990). Leonard, William N. “ Research and Development in Industrial Growth,” Journal of Political Economy (March/April 1971), pp. 232-56. Mansfield, Edwin. “ Technological Change and Industrial Innovation,” Managerial Economics (W.W. Norton & Company, 1990), pp. 223-54. ________“ Industrial R&D in Japan and the United States: A Comparative Study,” American Economic Review, Papers and Proceedings (May 1988), pp. 223-28. _______ . “ The R&D Tax Credit and Other Technology Policy Issues,” American Economic Review, Papers and Proceedings (May 1986), pp. 190-94. _______ . “ R&D and Innovation: Some Empirical Findings,” in Zvi Griliches, ed. R&D, Patents, and Productivity (University of Chicago Press, 1984), pp. 127-54. _______ . “ How Economists See R&D,” Harvard Business Review (November/December 1981), pp. 98-106. _______ . “ Basic Research and Productivity Increase in Manufacturing,” American Economic Review (December 1980), pp. 863-73. Minasian, Jora R. “ The Economics of Research and Development,” The Rate and Direction of Inventive Activity: Economic and Social Factors (Princeton University Press, 1962), pp. 93-142. Mohnen, Pierre A., M. Ishaq Nadiri and Ingmar R. Prucha. “ R&D, Production Structure and Rates of Return in the U.S., Japanese and German Manufacturing Sectors: A Non-Separable Dynamic Factor Demand Model,” European Economic Review (August 1986), pp. 749-71. NOVEMBER/DECEMBER 1992 34 National Defense Authorization Act for Fiscal Year 1993, Report of the Committee on Armed Services, House of Representatives on H.R. 5006 (GPO, May 19, 1992). Rosenberg, Nathan. Inside the Black Box: Technology and Economics (Cambridge University Press, 1982). National Science Board. Science and Engineering Indicators (GPO, 1991). Scherer, Frederic M. “ Inter-Industry Technology Flows and Productivity Growth,” Review of Economics and Statistics (November 1982), pp. 627-34. _______ Science and Engineering Indicators (GPO, 1989). OECD. Education at a Glance, OECD Indicators (Paris: OECD, 1992). _______ . OECD Science and Technology Indicators, No. 2, R&D, Invention and Competitiveness (Paris: OECD, 1986). ________Trends in Industrial R&D in Selected OECD Member Countries, 1967-1975 (Paris: OECD, 1979). Romer, Paul M. “ Endogenous Technological Change,” Journal of Political Economy (October 1990), pp. 71-102. FEDERAL http://fraser.stlouisfed.org/ RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis Schumpeter, Joseph A. Capitalism, Socialism and Democracy, 3d ed. (Harper & Brothers, 1950). Solow, Robert M. “ Technical Change and Aggregate Production Function,” Review of Economics and Statistics (August 1957), pp. 312-20. U.S. Congress, House of Representatives, Committee on Ways and Means. Factors Affecting U.S. International Competitiveness. Hearing, 102 Cong. 1 Sess. (GPO, 1992). 35 Daniel L. Thornton and Piyu Yue Daniel L. Thornton is an assistant vice president at the Federal Reserve Bank of St. Louis. Piyu Yue, a research associate at the 1C2 Institute, University of Texas at Austin, was a visiting scholar at the Federal Reserve Bank o f St. Louis when this article was written. Lynn Dietrich and Kevin White provided research assistance. An Extended Series of Divisia Monetary Aggregates J . HE CONVENTION IN monetary economics has been to create m onetary aggregates by simply adding together the dollar amounts o f the various financial assets included in them. This is the simple-sum method o f aggregation. This procedure has been criticized because such m onetary aggregates are essentially indexes that weight each component financial asset equally, a practice that is economically meaningful only under special circumstances. A num ber o f alternative indexes o f m onetary aggregates have been developed recently. The most w ell known are the Divisia m onetary aggregates developed by Barnett (1980). This article review s the theoretical basis fo r m onetary aggregation and presents series o f Divisia m onetary aggregates fo r an extended sample period. The behavior o f the simple-sum aggregates and their Divisia counterparts are compared over this period. THE TH EO RETICAL BASIS FOR M O N E T AR Y A G G R E G A T IO N 1 Simple-sum aggregation stemmed directly from the classical economists’ notion that the essential function o f m oney is to facilitate transactions, that is, to serve as a medium o f exchange. Assets that served as media o f exchange w ere consid ered money and those that did not, w ere not. By this definition only tw o assets, currency and demand deposits, w ere considered money. Both assets w ere non-interest bearing, and individuals 'The discussion in this section is based on consumer demand theory. This may not be a serious limitation. For example, Feenstra (1986) has shown that money in the utility function is equivalent to other approaches. These approaches assume, however, that all of the costs and benefits of money are internalized, and it is commonly believed that there are externalities to the use of money in exchange (see Laidler [1990]). NOVEMBER/DECEMBER 1992 36 w ere fre e to alter the composition o f their m oney holdings betw een currency and demand deposits at a fixed one-to-one ratio. Consequently the m onetary value o f transactions was exactly equal to the sum o f the tw o monies.2 Simplesum aggregation was a natural extension o f both restricting the definition o f money to noninterest-bearing medium-of-exchange assets and o f the fixed unitary exchange rate betw een the tw o alternative monies.3 asset could not be used directly to facilitate transactions was no longer a sufficient condition fo r excluding it from the definition o f money. Instead, the asset approach to m oney emphasized m oney’s role as a tem porary abode o f purchasing p o w er that bridges the gap betw een the sale o f one item and the purchase o f another. Currency and checking accounts are m oney because they are both media o f exchange and tem porary abodes o f purchasing pow er. Non-medium o f exchange assets are superior to currency and non-interest-bearing checking accounts as stores o f value because they earn explicit interest. This superiority typically increases with the length o f time betw een the sale o f one item and subsequent purchase o f another because the cost o f getting into and out o f such assets and the medium o f exchange assets is thought to be small and not proportional to the size o f the transaction. In consumer demand theory, simple-sum aggregation is tantamount to treating currency and demand deposits as if they are perfect sub stitutes. Currency and demand deposits, however, are not equally useful fo r all transactions, so this assumption was clearly inappropriate. But, simple-sum aggregation o f those tw o m onetary assets was still appropriate because the assets w ere non-interest bearing and exchanged at a fixed one-to-one ratio. Consequently individuals w ould allocate their portfolio o f m oney betw een the tw o assets until they equalized the marginal utilities o f the last dollar held o f each. Under these conditions, simple-sum aggregation is ap propriate if it is also assumed that each agent is holding his equilibrium portfolio. This shift in emphasis in m onetary theory dramatically expanded the num ber o f assets that w ere considered m oney and the number o f alternative monetary aggregates proliferated.5 Nonetheless, the method o f aggregation remained the same—simple-sum aggregation. The recognition that non-interest-bearing demand deposits may have paid an implicit interest weakened the theoretical justification for simple-sum aggregation. A m ore serious blow to simple-sum aggregation, how ever, was dealt by a shift in m onetary theory to emphasizing the store-of-value function o f m oney.4 That an As m ore financial assets came to be considered money, it became increasingly clear that it was inappropriate to treat these assets as perfect substitutes. Some financial assets have m ore “moneyness” than others, and hence they should receive larger weights. In w hat appears to be the first attempt at constructing a theoretically 2This need not be true for the economy as a whole when measured over a sufficiently long time interval. In this case the amount of each form of money multiplied by its turnover velocity will equal total expenditures. This is the basis for the velocity of the demand for money. Fisher (1911) explicitly recognized that turnover velocities of currency and checkable deposits would likely be different. He circumvented this problem by assuming that there was an optimal currency-todeposit ratio that would be a function of economic variables. Given these variables, the demand for the two alternative monetary assets was taken to be strictly proportional. More over, because individuals were free to adjust their money holdings between currency and checkable deposits quickly and at low cost, Fisher argued that the actual ratio would deviate from the desired ratio for only short periods. For some recent evidence that the actual currency-to-deposit ratio might be determined by the policy actions of the Federal Reserve, see Garfinkel and Thornton (1991). The possibility that currency and checkable deposits have different turnover velocities is the basis for Spindt’s (1985) weighted monetary aggregate, MQ. (1967) argued that the one-to-one exchange rate was a natural outcome of competitive pressures in the banking industry. Whether the fixed one-to-one ratio is the endogenous outcome of a free market economy or is simply due to legal restrictions remains controversial. Of course today some checkable deposits earn explicit interest. Consequently such deposits are a better store of wealth than currency. They are also a preferable medium of exchange for some, but not all, transactions. 3There is an issue of whether the fixed ratio was endogenous, from either the perspective of supply or demand, or the result of arbitrary legal restrictions. From the demand side, this would require that these assets be perfect substitutes for all transactions. From the supply side, Pesek and Saving FEDERAL http://fraser.stlouisfed.org/ RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis 4There has been a difference of opinion about the degree of emphasis that should be placed on the asset and transac tions motives for holding money. Indeed, Laidler (1990, pp. 105-6) has noted that “ ...th e most extraordinary development in monetary theory over the past fifty years is the way in which money's means-of-exchange and unit-ofaccount roles have vanished from what is widely regarded as the mainstream of monetary theory.” Broaching the medium-of-exchange line of demarcation between money and non-money assets also gave rise to an extensive literature on the empirical definition of money. For a critique of this literature and the idea of distinguishing between monetary and non-monetary assets based on the concept of the temporary abode of purchasing power, see Mason (1976). 5 one point the Federal Reserve published data on five At alternative monetary aggregates. 37 preferable alternative to the simple-sum monetary aggregate, Chetty (1969) added various savingstype deposits, weighted by estimates o f the degree o f substitution betw een them and the pure medium o f exchange assets, to currency and demand deposits. Larger weights w ere given to assets with a higher estimated degree o f substitution.6 Divisia aggregation, which also relies on consumer demand theory and the theory o f economic aggregation, treats m onetary assets as consumer durables such as cars, televisions and houses. They are held fo r the flo w o f utilitygenerating m onetary services they provide. In theory, the service flo w is given by the utility level. Consequently the marginal service flo w o f a m onetary asset is its marginal utility. In equilibrium, the marginal service flo w o f a monetary asset is proportional to its rental rate, so the change in the value o f a m onetary asset’s service flo w per dollar o f the asset held can be approximated by its user cost. The marginal m onetary services o f the components o f Divisia aggregates are likewise proxied by the user costs o f the component assets. The user cost o f each component is proportional to the interest income foregone by holding it rather than a pure store-of-wealth asset—an asset that yields a high rate o f return but provides no m onetary services. Currency and non-interest-bearing demand deposits have the highest user cost because they earn no explicit interest income. Consequently they get the largest weights in the Divisia measure. On the other hand, pure storeof-wealth assets get zero weights.7 Th e object o f a Divisia measure is to construct an index o f the flo w o f m onetary services from 6Chetty’s work was motivated by the Gurley/Shaw hypothesis and the general lack of agreement in the empirical findings of Feige (1964) and others about the degree of substitutability between money and near-money assets. Gurley and Shaw (1960) suggested that the effectiveness of monetary policy was limited because of the high degree of substitutability between money (currency and demand deposits) and near money (various bank and nonbank savings-type accounts) assets. Subsequent research has tended to support Feige’s finding of a relatively low degree of substitutability between transactions media and liquid, non-medium-of-exchange assets. See Fisher (1989) for a survey of much of this literature. 'There does not appear to be agreement about what consti tutes the best proxy measure for the theoretical pure store-ofwealth asset. Barnett, Fisher and Serletis (1992, p. 2,093) state the following, “ The benchmark asset is specifically assumed to provide no liquidity or other monetary services and is held solely to transfer wealth intertemporally. In theory, R (the benchmark rate) is the maximum expected holding period yield in the economy. It is usually defined in practice in such a way that the user costs for the monetary a group o f m onetary assets, w h ere the m onetary service flo w per dollar o f the asset held can vary from asset to asset.8 Applying an appropriate index number to a group o f assets is not sufficient, how ever, to get a correct measure o f the flo w o f m onetary services. The index must also be constructed from a set o f assets that can be aggregated under conditions set by consumer demand theory. The objective o f economic aggregation is to identify a group o f goods that behave as if they w ere a single commodity. A necessary condition fo r this is block-wise weak separability. Block-wise weak separability requires that consumers’ decisions about goods that are outside the group do not influence their p re ferences over the goods in the group whatso ever.9 I f this condition is satisfied, consumers behave just as though they w ere allocating their incomes over a single aggregate measure o f m onetary services and all other commodities to maximize their utility. Th eir total expenditure on m onetary services is subsequently allocated over the various financial assets that provide such services. The Divisia index generates such a m onetary aggregate. M oreover, in continuous time it has been shown to be consistent w ith any unknown utility function implied by the data. In discrete time the Divisia index is in the class o f superlative index numbers. Simple-sum indexes, on the other hand, do not have this desirable property. Thus they have no basis in either consumer demand theory or aggregation theory.1 0 In principle, all financial assets other than pure store-of-wealth assets provide some monetary services. W hich assets can be combined into a meaningful m onetary aggregate is an empirical assets are (always) positive.” Parentheses added. The Baa bond rate, or the highest rate paid on any of the component assets when the yield curve becomes inverted, has frequently been used to construct Divisia aggregates. 8See Barnett, Fisher and Serletis (1992) and Yue (1991a and b) for more detailed analyses of issues in monetary aggregation. te c h n ic a lly the marginal rates of substitution between any two goods inside the group must be independent of the quantities of the goods consumed that are outside of the group. 10Fisher (1922) was especially critical of the simple-sum index in his extensive analysis of index numbers. In parti cular, Fisher argued that simple-sum aggregates cannot internalize pure substitution effects associated with relative price changes. Thus changes in utility, which should occur only as a result of the income effect associated with relative price changes, occur in simple-sum aggregates because of both income and substitution effects. NOVEMBER/DECEMBER 1992 38 issue because economic theory does not tell us w hich group o f assets satisfies the condition o f block-wise weak separability. Unfortunately, the most w idely used test fo r weak separability is not pow erful.1 Consequently, it has been common 1 simply to create Divisia indexes under the maintained hypothesis that the assets that compose the aggregate satisfy this condition. Thus the issues o f the appropriate method o f aggregation and the appropriate aggregate have been treated separately.1 2 SIMPLE-SUM A N D D IVISIA M O NE TAR Y INDEXES A simple-sum m onetary aggregate is a measure o f the stock o f financial assets that compose it, whereas a Divisia monetary aggregate is a measure o f the flo w o f m onetary services from the stocks o f financial assets that compose it.1 3 For this reason alone, the methods o f measure ment are quite different. Simple-sum aggregates are obtained by simply adding the dollar amounts o f the component assets. On the other hand, Divisia m onetary aggregates are obtained by multiplying each component asset’s grow th rate by its share weights and adding the products. A component’s share w eight depends on the user costs and the quantities o f all component assets.1 Specifically, the share w eight given to 4 the jlh component asset at time t is its share o f total expenditures on m onetary services; that is, n s,< = u,< v ( E U‘ J i- i w h ere q denotes the nominal quantity o f the j ,h component asset, a denotes the j th component’s 11The most widely used test, developed by Varian (1982, 1983), is not statistical. The null hypothesis of weak separability is rejected if a single violation of the so-called regularity conditions is found. Because tests for weak separability lack power, Barnett, Fisher and Serletis (1992, p. 2,095) argue that “ existing methods of conducting such tests are n o t.. .very effective tools of analysis.” See Barnett and Choi (1989) for evidence indicating that available tests of block-wise weak separability are not very dependable. For results of tests for weak separability, see Belongia and Chalfant (1989) and Swofford and Whitney (1986, 1987). 12A common practice both in the United States and abroad is to construct Divisia monetary aggregates for collections of assets that are reported by the country’s central bank. For example, see Yue and Fluri (1991), Belongia and Chrystal (1991) and Ishida (1984). 13lt should be noted that the accounting stock, that is, the sum of the dollar amounts of all assets that are considered money, is not necessrily equal to the capital stock of money. The accounting stock is the present value of both service flow of money and the interest income (the service as a store of value). The economic capital stock of money FEDERAL RESERVE BANK OF ST. LOUIS user cost and n denotes the num ber o f component financial assets. Th e user cost is equal to (R-r.)p/ (1 + R), w h ere R is the benchmark rate (that is, the rate on the pure store-of-wealth asset), r. is the ow n rate on the j ,hcomponent, and p is the true cost-of-living price index that cancels out o f the num erator and denominator o f the shares. The grow th rate o f the ithDivisia m onetary aggre gate, GDMj, is given by GDM= } ] [(S., + Sjt_t)/2]gjt, j- J w h ere g)tis the grow th rate o f q it.1 5 A Comparison o f Simple-Sum and Divisia M onetary Aggregates Because the Divisia aggregates are an alterna tive to the conventional simple-sum aggregates, it is instructive to compare them. W hen con structing data in this section, the authors used an extension o f the Farr and Johnson (1985) method. Th e Appendix presents details o f the construction o f the Divisia m onetary aggregates used here. A Divisia m onetary index is an approximation to a nonlinear utility function. Because it is an index, the level o f utility is an arbitrary unit o f measure; the level o f the index has no particular meaning.1 Nevertheless, because they are alter 6 native measures o f money, the Divisia and simplesum aggregates are frequently com pared to see how any analysis o f the effects o f m onetary policy or other issues might be affected by the m ethod o f aggregation. The comparison o f the levels o f comprises only the present value of the flow of monetary services. See Barnett (1991) for the formula for the economic capital stock of money. 14For the Divisia monetary aggregates, the share weight of each component’s growth rate is its expenditure share of total expenditures on monetary services. Theoretically the share weights for the Divisia monetary aggregates are not a function of prices or user costs, but of quantities. The observable user costs are substituted for the unobservable marginal utilities under the implicit assumption of marketclearing equilibrium, where each consumer holds an optimal portfolio of monetary and nonmonetary assets. For the simple-sum monetary aggregates, the share weights are the components’ share of the aggregate. 15GDMj = In Dlt - In Dlt _ w h e r e Dl denotes the Divisia index. The index is initialized at 100, that is, Dl0 = 100. See Farr and Johnson (1985) for more details. 16Ftotemberg (1991) derives a weighted monetary aggregate stock under conditions of risk neutrality and stationarity expectations; however, Barnett (1991) shows that this measure is the discounted value of future Divisia monetary service flows. 39 i Figure 1 Year-Over-Year Growth of SSM1a and D IVM Ia, and Levels of SSM1a and D IVM Ia Percent Index 16-|— — 600 SS Growth « - —■ ■ ■■ Div Growth ■ ■ ■ ■ -400 -300 -200 -100 Div Level - - - 1 1960 I I— I 62 I 64 I l 66 l l 68 l 70 l l the simple-sum and Divisia measures is made by norm alizing both measures so that they equal 100 at some point in the series, usually the first observa tion.1 Comparisons o f the levels and grow th rates 7 o f the Divisia and simple-sum measures are p re sented in figures 1-5 fo r four monetary aggregates, M IA , M l, M2 and M3, and for total liquid assets, L.1 The figures have tw o scales. The left-hand 8 scale indicates the grow th rate, and the righthand scale indicates the level o f the series. Both indexes equal 100 in January 1960. 17An alternative justification for comparing the Divisia and simple-sum aggregates might come from noting that the appropriate Divisia monetary aggregate would be the simple-sum aggregate if all of the component assets had identical own rates. Such a comparison is tenuous, however, because the actual level of the simple-sum aggregate might have been different from the observed level had the user costs actually been equal. It is common to compare the levels and growth rates of simple-sum and Divisia monetary aggregates. For example, see Barnett, Fisher and Serletis (1992). Because Divisia indexes involve logarithms, the growth rate of a component asset is plus or minus infinity, respectively, when a com ponent is introduced or eliminated. To circumvent this problem, the Divisia index is replaced by Fisher’s ideal index at these times and the user cost is measured by its reservation price during the period that precedes the intro duction or follows the elimination of the asset. See Farr i r 82 72 i i 84 i i 86 i i 88 i i r 90 1992 M IA M IA comprises currency and non-interestbearing demand deposits held by households and businesses. Although neither household nor business demand deposits earn explicit interest, business demand deposits are assumed to earn an implicit ow n rate o f return proportional to the rate paid on one-month commercial paper.1 9 Consequently, additional units o f business demand deposits are assumed to yield a smaller and Johnson (1985) for a discussion of this procedure. 18Note that the simple-sum aggregates presented here are not identical to the official published series. The official series are obtained by adding the non-seasonally adjusted components and seasonally adjusting the aggregate as a whole or by adding large subgroups of component assets that have been seasonally adjusted as a whole. The simple-sum aggregates presented here are obtained by adding the components after each component (that has a distinctive seasonal) has been seasonally adjusted. See the Appendix for details. A comparison of the series used here and the official series shows that the differences are small. ^Alternatively, estimates of the own rate on household demand deposits could also be used. However, such a series was not available for the entire sample period. Moreover, the desire was to follow the procedure used by Farr and Johnson (1985) as closely as possible. NOVEMBER/DECEMBER 1992 40 Figure 2 Year-Over-Year Growth of SSM1 and DIVM1, and Levels of SSM1 and DIVM1 Percent Index Figure 3 Year-Over-Year Growth of SSM2 and DIVM2, and Levels of SSM2 and DIVM2 Percent 1 4 -| SS Growth Div G rowth SS Level Div L evel--------2 - l 1960 i i 62 i i 64 I i 66 i i 68 i r 70 http://fraser.stlouisfed.org/ RESERVE BANK OF ST. LOUIS FEDERAL Federal Reserve Bank of St. Louis i i 72 i i 74 i i 76 i i 78 i i 80 i i 82 i I 84 i i 86 i i 88 i i 90 r ^0 1992 41 Figure 4 Year-Over-Year Growth of SSM3 and DIVM3, and Levels of SSM3 and DIVM3 Percent Index Figure 5 Year-Over-Year Growth of SSL and DIVL, and Levels of SSL and DIVL Percent Index NOVEMBER/DECEMBER 1992 42 flo w o f m onetary services than are additional units o f household demand deposits. On the other hand, the simple-sum measure implicitly assumes that each unit o f each component provides the same flo w o f m onetary services. Hence the Divisia aggregate gives m ore w eight to the grow th rates o f currency and household demand deposits than does the simple-sum aggregate.2 0 The average differences in the grow th rates o f the simple-sum and Divisia measures o f M IA fo r the entire sample period, January 1960 to Decem ber 1992, and fo r selected sub-periods are presented in table 1. Because currency generally g rew m ore rapidly than demand deposits over the sample period, the grow th rate o f Divisia M IA averaged about half a percentage point higher than the grow th rate o f simple-sum M IA over the entire period.2 Much 1 o f this difference occurs during the latter part o f the 1980s, when the grow th rate o f demand deposits generally slowed relative to the grow th rate o f currency.2 This m ore rapid grow th o f 2 the Divisia measure is reflected in a generally widening gap betw een the levels o f the indexes. Ml The behavior o f simple-sum and Divisia M l is similar to that o f M IA . Indeed, the grow th rates o f simple-sum and Divisia M l w ere similar until the late 1970s, w hen the grow th o f interestbearing N O W accounts began to accelerate. The sharp rise in N O W accounts after their nationwide introduction on January 1, 1981, tended to increase the grow th rate o f the simple-sum measure relative to the Divisia measure because the grow th rate o f NO W accounts gets a smaller w eight in the Divisia measure. As a result, the Divisia measure g rew m ore slowly on average 20ln both cases, the sum of the weights must equal unity. 21Currency grew at an annual rate of 7 percent during the entire period, whereas household and business demand deposits both grew at a 3.2 percent annual rate. 22This is a period of very slow reserve growth. Because reserves and checkable deposits are tied closely together under the present system of reserve requirements, it is not surprising that this is also a period of slow growth in check able deposits, including household and business demand deposits. See Garfinkel and Thornton (1991) for a discussion of the relationship between reserves and checkable deposits under the present system of reserve requirements. 23We have refrained from using the phrase ‘‘statistically significant” because these observations are clearly dis tributed identically and independently, so the “ t-statistics” reported in table 1 are biased and neither the direction nor extent of the bias is known. These statistics are presented FEDERAL http://fraser.stlouisfed.org/ RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis than the simple-sum measure from the late 1970s until the mid-1980s, after grow in g m ore rapidly previously. H ow ever, in neither period is the average difference in the grow th rate o f the alternative measures large.2 3 A fter the late 1980s the Divisia measure g re w m ore rapidly than the simple-sum measure, reflecting the rise in the growth rate o f currency relative to the grow th rate o f checkable deposits. O f course, the smaller average difference in the grow th rates o f the alternative M l aggregates compared w ith M IA is reflected in a smaller difference in the levels o f the tw o indexes as well. M2, M3 and L Not surprisingly, larger differences arise w hen the m onetary measures are broadened to include savings-type deposits because their explicit ow n rates o f return are higher than those o f trans actions deposits. Th e higher ow n rate reduces the share weights o f these component assets relative to the weights they receive in the simple-sum measures. During the sample period the grow th rates o f the broader simple-sum aggregates tend to be substantially larger than those o f the corresponding Divisia measures. For the broader measures, the average grow th rates o f the simple-sum measures are about 2 percentage points greater than the corresponding Divisia measures over the entire sample period. Much o f this difference arises from the late 1970s to the mid-1980s and is likely due to financial innovation and deregulation in the period . T h e late 1970s w itn essed a m arked acceleration in the grow th o f money market mutual funds. These accounts paid relatively high interest rates and had limited transactions capabilities. A number o f new deposit instruments that paid higher market interest rates w ere to give the reader a rough approximation of the magnitude of the differences in the growth rates. Correlegrams of the difference in the growth rates of simple-sum and Divisia M1A and M1 show some lower level persistence through the sample period and some large spikes at seasonal frequencies after 1969. Correlegrams for the difference in the growth rates of the broader monetary aggregates reveal some higher level persistence. In any event, differences that are small in absolute value tend to be small relative to the estimated standard errors, and differences that are large in absolute value tend to be large in relative terms. Another measure of the distance between the growth rates is the square root of the sum of the squared differ ences in the growth rates. These measures for the entire sample period are 58.5, 52.1, 69.6, 81.4 and 77.6 for M1A, M1, M2, M3 and L, respectively. These data are broadly comparable with those presented in table 1. 43 Table 1 Average Percentage Point Difference in the Annual Growth Rate of Simple-Sum and Divisia Aggregates M ean1 Standard Deviation M1A M1 M2 M3 L -0 .5 1 4 -0 .1 5 3 1.889 2.317 2.223 3.16 2.83 3.49 3.88 3.66 3.24* 1.07 10.75* 11.88* 12.07* M1A M1 M2 M3 L -0 .2 8 5 -0 .2 5 3 1.660 2.134 1.897 2.04 2.03 1.54 2.08 1.73 2.05* 1.82 15.81* 15.02* 16.10* M1A M1 M2 M3 L -0 .3 2 4 0.420 3.526 4.334 4.303 3.83 3.65 5.59 5.75 5.50 0.88 1.20 6.55* 7.84* 8.12* M1A M1 M2 M3 L Period -1 .4 8 5 -0 .7 1 4 0.116 -0 .1 6 3 0.076 4.41 3.32 2.45 2.82 2.84 2.85* 1.82 0.40 0.49 0.23 Aggregate t-statistic 1 9 6 0 .0 1 -1 9 9 2 .0 7 1 9 6 0 .0 1 -1 9 7 7 .1 2 1 9 7 8 .0 1 -1 9 8 6 .1 2 1 9 8 7 .0 1 -1 9 9 2 .1 2 'The growth rate of the simple-sum aggregate less the growth rate of the Divisia aggregate. 'Indicates a t-statistic greater than 2. See footnote 23. introduced in the early 1980s and Regulation Q interest rate ceilings w ere being phased out.2 4 M oreover, short-term interest rates reached very high levels in the early 1980s. W ith share weights sensitive to the spread betw een an asset’s ow n rate o f return and the return on the benchmark asset, it is not surprising that the Divisia measures g rew markedly slower than the corresponding simple-sum measures during this period. Nevertheless, the signi ficantly slower grow th o f the broader Divisia measures during this period is m ore consistent with the disinflation o f the period than is the grow th o f the simple-sum aggregates, whose grow th remained fairly rapid. Although the grow th rates o f the broader Divisia and simplesum aggregates have been essentially the same, on average, since about the mid-1980s, the pattern o f grow th o f these alternative measures is somewhat different. 24For a discussion of the financial innovations of this period see Gilbert (1986) and Stone and Thornton (1991). NOVEMBER/DECEMBER 1992 44 Figure 6 Levels of Divisia M2, M3, and L Index A Comparison o f Broader Divisia Aggregates That Divisia aggregation gives relatively small w eight to less liquid assets that yield high rates o f return suggests that differences in the grow th rates o f successively broader Divisia m onetary aggregates w ill tend to get smaller.2 Th e levels 5 o f Divisia M2, M3 and L presented in figure 6 and simple correlations o f the compounded annual grow th rates o f these Divisia aggregates presented in table 2 confirm this. The grow th rates o f Divisia M3 and L d iffer little from the grow th rate o f Divisia M2. This implies that adding successively less liquid assets to those in M2 adds little to the flo w o f m onetary services.2 That the 6 average difference in the grow th rates o f Divisia M2 and L is nearly zero over the entire sample 250 f course this tendency also exists for the simple-sum aggregates. For the simple-sum aggregate, the growth rate of each component is weighted by the component’s share of the total asset. Hence the growth rates of successively broader monetary aggregates could diverge if the marginal components were successively larger. For example, this is what happens from M1 to M2. The growth rates tend to converge, however, because the marginal components are smaller. This tendency is exacerbated in the Divisia FEDERAL http://fraser.stlouisfed.org/ RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis period is reflected by the levels o f the tw o Divisia aggregates, which are essentially equal by the end o f the sample. Divisia M3, how ever, has grow n m ore rapidly than the other measures, so the spread betw een its level and the levels o f Divisia M2 and L has w idened over the sample period. CONCLUDING REM ARKS Despite their theoretical advantage, Divisia and other weighted m onetary aggregates have garnered relatively little attention outside o f academe, and the official U.S. monetary aggregates remain simple-sum aggregates. The official reliance on simple-sum aggregates w ill probably continue unless the Divisia aggregates or other alternative w eighted aggregates are shown to be superior in economic and policy analysis. measures because of smaller weights associated with higher own rates of return on successively less liquid assets. 26The average differences in the growth rates of Divisia M2, M3 and L over the sample period are small (less than 0.12 percentage points in absolute value). The absolute values of the average differences in the growth rates of simplesum M2, M3 and L are larger than those of the corresponding Divisia measures; the standard errors are also much larger. 45 Table 2 Correlations of the Annual Growth Rates of the Divisia Monetary Aggregates_______________________ Aggregate M 1A Ml M2 M1 .7920 M2 .6540 .6914 M3 .6015 .6346 .9568 L .5754 .6126 .8863 M3 .9216 Although nothing definitive can be said about this issue from the simple analysis o f the data presented here, a fe w observations are offered. First, that the grow th rates o f the narrow simplesum and Divisia monetary aggregates are quite similar suggests that the method o f aggregation may not be important at low levels o f aggregation.2 7 For example, it does not appear that conclusions about the long-run effects o f m oney grow th on inflation would be much different using either simple-sum or Divisia M l or M IA . The average difference in the grow th rates o f narrow simplesum and Divisia m onetary aggregates is small. This observation is consistent w ith the empirical w ork o f Barnett, Offenbacher and Spindt (1984) who, using a broad array o f criteria, found that the difference in the perform ance o f simple-sum and Divisia m onetary aggregates was small at low levels o f aggregation. Second, the method o f aggregation is likely to be m ore important fo r broader m onetary aggregates. Beyond some point, however, a further broadening o f the m onetary aggregate makes little difference. For the United States, the differences in the average grow th rates o f Divisia M2, M3 and L are small. Consequently, long-run analysis using the grow th rates o f any o f these Divisia aggregates is likely to produce similar results. Monthly grow th rates o f these Divisia aggregates are also highly correlated. Hence it would not be too surprising to find that the broader Divisia aggregates perform similarly to one another in many short-run analyses as well. These observations point to the critical need fo r m ore w ork to determine which financial 27There may be some differences in the levels, however, because the levels of the simple-sum and Divisia assets should be included in the appropriate monetary aggregate. In consumer demand theory, these assets must satisfy the condition o f weak separability. If analysis suggests a relatively narrow m onetary aggregate such as M l, policymakers may be reluctant to adopt the theoretically superior index measure because, as a practical matter, the m ethod o f aggregation may not be em pirically important. If such tests point to an aggregate that includes a much broader array o f financial assets, the practical case fo r the w eighted aggregates w ill be enhanced. Even casual analysis o f simplesum and Divisia m onetary aggregate data show differences in both the levels and grow th rates o f these aggregates that are large, suggesting that the method o f aggregation is important. Consequently, the method o f aggregation should also be a concern fo r those w ho favor broader m onetary aggregates on other grounds. The objective o f the present article in publishing Divisia m onetary statistics is to stimulate further empirical research both on the importance o f m onetary aggregation and on the role o f money in the economy. REFERENCES Barnett, William A., Douglas Fisher and Apostolos Serletis. “ Consumer Theory and the Demand for Money,” Journal of Economic Literature (December 1992), pp. 2,086-119. Barnett, William A. “ Reply,” in Michael T. Belongia, ed., Monetary Policy on the 75th Anniversary of the Federal Reserve System (Kluwer Academic Publishers, 1991), pp. 232-42. Barnett, William A., and Seungmook Choi, “ A Monte Carlo Study of Tests of Blockwise Weak Separability,” Journal of Business and Economic Statistics (July 1989), pp. 363-79. Barnett, William A., Edward K. Offenbacher and Paul A. Spindt. “ The New Divisia Monetary Aggregates,” Journal of Political Economy (December 1984), pp. 1,049-85. Barnett, William A. “ Economic Monetary Aggregates: An Application of Index Number and Aggregation Theory,” Journal of Econometrics (September 1980), pp. 11-48. Belongia, Michael T., and K. Alec Chrystal. “ An Admissible Monetary Aggregate for the United Kingdom,” Review of Economics and Statistics (August 1991), pp. 491-502. Belongia, Michael T., and James A. Chalfant. “ The Changing Empirical Definition of Money: Some Estimates from a Model of the Demand for Money Substitutes,” Journal of Political Economy (April 1989), pp. 387-97. Chetty, V. Karuppan. “ On Measuring the Nearness of NearMoneys,” American Economic Review (June 1969), pp. 270-81. Farr, Helen T., and Deborah Johnson. “ Revisions in the Monetary Services (Divisia) Indexes of Monetary Aggregates,” Staff Study 147, Board of Governors of the Federal Reserve System (1985). measures do not appear to be cointegrated at any level of aggregation. NOVEMBER/DECEMBER 1992 46 Feenstra, Robert C. “ Functional Equivalence between Liquidity Costs and the Utility of Money,” Journal of Monetary Economics (March 1986), pp. 271-91. Feige, Edward L. The Demand for Liquid Assets: A Temporal Cross-Section Analysis, (Englewood Cliffs, 1964). Fisher, Douglas. Money Demand and Monetary Policy, (University of Michigan Press, 1989). Fisher, Irving. The Making of Index Numbers: A Study of Their Varieties, Tests, and Reliability, (Houghton Mifflin, 1922). ________The Purchasing Power of Money (Augustus M. Kelley, 1911). Garfinkel, Michelle R., and Daniel L. Thornton. “ The Multiplier Approach to the Money Supply Process: A Precautionary Note,” this Review (July/August 1991), pp. 47-64. Gilbert, R. Alton. “ Requiem For Regulation Q: What It Did and Why It Passed Away,” this Review (February 1986), pp. 22-37. Gurley, John G. and Edward S. Shaw. Money in a Theory of Finance, (Brookings Institution, 1960). Rotemberg, Julio J., “ Monetary Aggregates and Their Uses,” in Michael T. Belongia, ed., Monetary Policy on the 75th Anniversary of the Federal Reserve System (Kluwer Academic Publishers, 1991), pp. 223-31. Spindt, Paul A. “ Money is What Money Does: Monetary Aggregation and the Equation of Exchange,” Journal of Political Economy (February 1985), pp. 175-204. Stone, Courtenay C., and Daniel L. Thornton. “ Financial Innovation: Causes and Consequences,” in Kevin Dowd and Mervyn K. Lewis, eds., Current Issues in Monetary Analysis and Policy, (MacMillan Publishers, 1991). Swofford, James L., and Gerald A. Whitney. “ Nonparametric Tests of Utility Maximization and Weak Separability for Consumption, Leisure and Money,” Review of Economics and Statistics (August 1987), pp. 458-64. ________“ Flexible Functional Forms and the Utility Approach to the Demand for Money: A Nonparametric Analysis,” Journal of Money, Credit and Banking (August 1986), pp. 383-89. Varian, Hal R. “ Non-Parametric Tests of Consumer Behavior,” Review of Economic Studies (January 1983), pp. 99-110. ________“ The Nonparametic Approach to Demand Analysis,” Econometrica (July 1982), pp. 945-73. Ishida, Kazuhiko. “ Divisia Monetary Aggregates and Demand for Money: A Japanese Case,” Bank of Japan Monetary and Economic Studies (June 1984), pp. 49-80. Yue, Piyu. “ A Microeconomic Approach to Estimating Money Demand: The Asymptotically Ideal Model,” this Review (November/December 1991a), pp. 36-51. Laidler, David. Taking Money Seriously and Other Essays, (MIT Press, 1990). ________Theoretic Monetary Aggregation Under Risk Averse Preferences, IC2 Institute (University of Texas at Austin, 1991b). Mason, Will E. “ The Empirical Definition of Money: A Critique,” Economic Inquiry (December 1976), pp. 525-38. Pesek, P. Boris, and Thomas R. Saving. Money, Wealth and Economic Theory (The MacMillan Company, 1967). Yue, Piyu, and Robert Fluri. “ Divisia Monetary Services Indexes for Switzerland: Are they Useful for Monetary Targeting?” this Review (September/October 1991), pp. 19-33. Appendix Constructing Divisia Monetary Aggregates: 1 9 6 0 -1 9 9 2 1 The assets used to calculate Divisia monetary aggregates are the same as those used by the Board o f Governors to calculate the official simple-sum aggregates M lA through L. The only major dif ference is that demand deposits are broken into household demand deposits (HDD) and business demand deposits (BDD). W e assume that house holds receive a zero rate o f return on demand deposits and that businesses receive an implicit, nonzero rate o f return. HDD and BDD are com puted using seasonally adjusted monthly data for total demand deposits and non-seasonally adjusted quarterly data fo r consumer, foreign, financial, nonfinancial and other demand deposits. These can be found in Table 1.31 o f the Federal Reserve Bulletin. Using the non-seasonally adjusted quarterly data, w e calculate tw o ratios and use them to par 1 The Divisia monetary aggregates data presented here were constructed under the direction of Piyu Yue. Lynn D. http://fraser.stlouisfed.org/ FEDERAL RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis tition the seasonally adjusted monthly data. The ratio fo r BDD is the sum o f financial and nonfinancial demand deposits divided by the sum o f all five non-seasonally adjusted series, whereas the ratio fo r HDD is one minus the BDD ratio. The nonseasonally adjusted series go back only to 1970.01 and are discontinued after 1990.06. For data before 1970.01 and after 1990.06, the means o f the re spective ratio series over the available sample w ere used. The means used w ere 62.33 percent fo r BDD and 37.67 percent fo r HDD. To get the final HDD and BDD series, these quarterly ratios w ere multiplied by the seasonally adjusted monthly data fo r total demand deposits. Each quarterly observation was multiplied by the three months o f data fo r that particular quarter. All assets are seasonally adjusted and in millions o f dollars. Dietrich and Kevin L. Kliesen gathered the data, wrote the computer code and wrote this appendix. 47 M o n e y Com pon en ts CUR — Sum o f seasonally adjusted currency and traveler’s checks. DEMDPS — Total demand deposits. HDD — Demand deposits fo r households as described in the preceding section. BDD — Demand deposits fo r businesses as described in the preceding section. OCD — Other checkable deposits less super NO W account balances. OCD includes ATS and NO W balances, credit union share draft balances and demand deposits at thrift insti tutions. SNOWC — Super NO W accounts at comm ercial banks. SNOWC data begin in 1983.01 and end in 1986.03. A fter 1986.03 there is no distinction betw een NOW s and super NOWs. SNOW T — Super NO W accounts at thrifts. SNO W T begin in 1983.01 and end in 1986.03. A fter 1986.03 there is no distinction betw een NOWs and super NOWs. ONRP — Overnight repurchase agreements. ONRP includes overnight and continuing contract repurchase agreements issued by commercial banks to organizations other than depository institutions and money market mutual funds (MMMFs) (general purpose and broker/dealer organizations). ONED — Overnight eurodollars. ONEDs are issued by foreign (principally Caribbean and London) branches o f U.S. banks to U.S. residents and organizations other than depository institutions and m oney market mutual funds. MMMF — M oney market mutual funds. MMMF is general purpose and broker/dealer money market mutual fund balances including taxable and tax-exempt funds and excluding IRA/KEOGH accounts at money funds. MMDAC — M oney market deposit accounts at commercial banks. MM DAC initially had a minimum balance requirem ent o f $2,500 until Decem ber 31, 1984, and a $1,000 minimum balance requirem ent until Decem ber 31, 1985, w hen the requirem ent was removed. MMDACs w ere no longer reported after 1991.08. M M D AT — M oney market deposit accounts at thrifts. M M D AT initially had a minimum balance requirem ent o f $2,500 until Decem ber 31, 1984, and a $1,000 minimum balance requirem ent until Decem ber 31, 1985, w hen the minimum requirem ent was removed. M M DATs w ere no longer reported after 1991.08. SDCB — Savings deposits at commercial banks less m oney market deposit accounts at commercial banks. MMDACs are included after 1991.08. SDSL — Savings deposits at thrifts less money market deposit accounts at thrifts. M M D ATs are included after 1991.08. STDCB — Small time deposits (less than $100,000) at thrifts including retail repurchase agreements less IRA/KEOGH accounts. STDTH — Small time deposits (less than $100,000) at thrifts including retail repurchase agreements less IRA/KEOGH accounts. LTDCB — Large time deposits (m ore than $100,000) at commercial banks excluding international banking facilities (IBFs). LTD TH — Large time deposits (m ore than $100,000) at thrifts excluding IBFS. MMMFI — Institution only m oney market mutual funds. MMMFI includes taxable and tax-exempt funds and excludes IRA/KEOGH accounts at money funds. TRP — Term repurchase agreements. TRP consists o f RPs w ith original maturities greater than one day, excluding continuing contracts and retail RPs. NOVEMBER/DECEMBER 1992 48 TED Term eurodollars w ith original maturities greater than one day. TED includes those eurodollars issued to U.S. residents by foreign branches o f U.S. banks and by all banking offices in the United Kingdom and Canada. Eurodollars held by depository in stitutions and MMMFs are not included. SB Savings bonds. STTS Short-term Treasury securities. STTS comprises U.S. Treasury bills and coupons with remaining maturities o f less than 12 months not held by depository institutions, Fed eral Reserve Banks, MMMFs or foreign entities. BA Bankers acceptances. BA is the net o f bankers acceptances held by accepting banks, Federal Reserve Banks, foreign official institutions, federal home loan banks and MMMFs. CP Total commercial paper less commercial paper held by MMMFs. The interest rate data are m ore complicated than the asset data. The major concern w ith the interest rate data is the variety o f form s in which they are reported. Before including dif feren t rates in an aggregate, the characteristics o f all the rates should be as similar as possible. To this end, tw o problems need to be addressed. First, fo r composite asset stocks w h ere the total asset is the sum o f deposits with different maturities, such as small and large time de posits, the ow n rate is the maximum rate paid across the deposit categories at each point. Because there are a variety o f maturity lengths among the rates o f a given composite asset stock, an adjustment is needed to transform each rate to a common maturity before the final rate is computed. Given rates with differin g maturities and a typical upward-sloping yield curve, liquidity premiums keep rates on assets w ith longer maturities higher than rates on those with shorter maturities. T o adjust these rates to a common maturity, this liquidity premium must be rem oved using a yield curve adjustment as described in Farr and Johnson (1985). As Farr and Johnson did, all rates that are yield curve adjusted are adjusted to a one-month maturity: R' = R - (TBM- T B 1 w here ), R = the original rate on a bond basis (that is, a 365-day basis) basis R‘ = the yield curve adjusted rate TBM = the M-month Treasury bill rate IB, = the one-month Treasury bill rate A second adjustment is needed to convert all the rates to the same yield basis. Interest rates are quoted in various forms, including discount basis and annual percentage rate basis, and have various interest bases, including bond (365 day) and bank (360 day). T o the extent possible, the rates w ere transformed into annualized one-month investment yields on a bond-interest basis. For rates quoted on a discount basis fo r a 360-day year, the follow ing form ula can be used to convert them to an annualized yield fo r a 365-day year (see Farr and Johnson [1985]): R = ([(365»D)/100] / {360 —[(N*D)/100]}) * 100, w h ere R = the annualized rate D = a discount basis rate (360-day year) N = the number o f days to maturity Including the variable N ensures that the form u la is maturity independent. Interest Rate Series f o r the Monetary Components RZER — Rate on currency and traveler's checks. RZER is zero by definition. RDD1 — Rate on household demand deposits. RDD1 is zero by definition. RDD2 — Rate on business demand deposits. The basic formula fo r computing is as follows: RDD2 = (1-MRR)»RCP w h ere MRR = maximum reserve requirem ent on demand deposits RCP = one-month financial paper rate Before applying this formula, adjust RCP, which is quoted on a discount basis fo r a 360-day year, to an annualized one-month yield fo r a 365-day year. This is done by using the form ula described in the preceding text. http://fraser.stlouisfed.org/ FEDERAL RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis RCP‘ = ([(365*RCP)/100]/{360 -[(30*RCP)/100]}) * 100 Then RDD2 = (1 -M R R )*R C P ‘ For MRR and all ceiling rates used in the follow ing text, w e use the same conven tion as Farr and Johnson and assume that rates are quoted as annualized one-month yields. ROCD — Rate on other checkable deposits. 1960.01-1974.11 — Regulation Q ceiling rate on passbook savings accounts at commercial banks. From 1962.01-1964.12 the ceiling rate on savings deposits o f less than 12 months is used. 1974.12-1986.03 — Regulation Q ceiling rate on N O W accounts. 1986.04-present — W eighted average interest rate on NOW s and super NOWs. RSNOWC — Rate on super NOWs at commercial banks. RSNOWC is the average rate paid on super N O W accounts at insured commercial banks and is quoted on an effective annual yield in the monthly Survey o f Selected Deposits, a special supplementary table in the w eekly Federal Reserve Statistical Release H.6. RSNOW T — Rate on super NOWs at thrift institutions. RSNOW T is the average rate paid on super N O W accounts at FDIC-insured savings banks (both mutual and federal savings banks) and is quoted on an effective annual yield in the monthly Survey o f Selected Deposits, a special supplementary table in the w eekly Federal Reserve Statistical Release H.6. RONRP — Rate on overnight dealer financing in the repurchase market. Because RONRP is an overnight rate quoted on a bank-interest basis, it must be transform ed into an annualized one-month yield on a bond-interest basis using the follow ing formula: RONRP* = {[1 + (RONRP/36000)]3 - 1 } * (36500/30) 0 Data fo r RONRP goes back only to 1972.01, whereas asset data goes back to 1969.11. Farr and Johnson argue that the rate on overnight RPs has historically been five basis points below the federal funds rate, so w e use the follow ing formula to compute a rate before 1972: RONRP* = ( {[1 + (RFF/36000)]3 - 1 } *(36500/30)) - .05 0 Like RONRP, the fed funds rate is an overnight rate quoted on a discount basis and must be transform ed into an annualized one-month yield on a bond-interest basis. RONED — Rate on overnight eurodollars from London. The original series is weekly, and thus the monthly series is a simple average o f the w eekly observations fo r a particular month. Like RONRP, RONED is an overnight rate quoted on a bank-interest basis and must be con verted to an annualized one-month yield on a bond-interest basis using the follow ing formula: RONED* = {[1 + (RONED/36000)]30- l } *(36500/30) RMMMF — Average yield o f money market mutual funds. RMMMF comes from the Board, which in turn gets it from Donoghue’s Money Fund Report. Data fo r RMMMF is available only back to 1974.06. RMMMF data from before this date are set to the rate on large time deposits at commercial banks (RLTDCB) less 70 basis points (see Farr and Johnson [1985]). RMMDAC — Rate on money market deposit accounts at commercial banks. Before 1989.06 RMMDAC is the average rate paid at insured commercial banks. A fter 1989.07 it is the average o f the rates paid at insured commercial banks fo r personal and nonpersonal MMDAs, which are quoted as effective annual yields in the monthly Survey o f Selected Deposits, a special supplementary table in the w eekly Federal Reserve Statistical Release H.6. NOVEMBER/DECEMBER 1992 50 RM M DAT — Rate on m oney market deposit accounts at thrift institutions. Before 1989.06 RM M D AT is the average rate paid at FDIC-insured savings banks. A fter 1989.07 it is the average o f the rates paid at FDIC-insured savings banks (including both mutual and federal savings banks) fo r personal and nonpersonal MMDAs, which are quoted as effective annual yields in the monthly Survey o f Selected Deposits, a special supplementary ta ble in the w eekly Federal Reserve Statistical Release H.6. RSDCB — Rate on savings deposits at commercial banks less m oney market deposit accounts at commercial banks. RSDCB comes from the Board and is quoted as an effective annual yield. RSDSL — Rate on savings deposits at FDIC-insured savings banks (the thrift rate). 1966.10-1986.03 — Th e ceiling rate on NO W accounts at thrifts. 1986.04-present — Th e rate on savings deposits at thrifts published in the Board’s H.6 release. Th ere are tw o problems w ith data before 1966.10: 1) interest rates on savings deposits at thrifts w ere not regulated and 2) different states paid different rates on these accounts. One o f the fe w series published fo r this period is the average dividend paid on savings deposits at thrifts, w hich is w hat w e use here. This is an annual rate and includes passbook savings accounts and fixed-term certificates. FITSTCB — Rate on small time deposits and retail repurchase agreements at comm ercial banks. FITSTCB is the Fitzgerald-adjusted small time deposit rate that is calculated at the Board and quoted as an effective annual yield. RSTTH — Rate on small time deposits and retail repurchase agreements at thrifts. RSTTH is the Fitzgerald-adjusted small time rate that is calculated at the Board and quoted as an e f fective annual yield. RLTDCB — Rate on large time deposits at commercial banks. RLTDCB is a yield-curve-adjusted rate that is calculated using the one-, three- and six-month secondary CD rates (of deposits greater than $100,000) and the one-, three- and six-month Treasury bill rates. 1) Th e first step is to convert the Treasury bill rates, which are quoted on a discount basis fo r a 360-day year, to annualized yields fo r a 365-day yea r as follows: Y* = ([(365*Y) /100] /{360 —[(N * Y)/100]} ) *100 w h ere Y = one-, three- and six-month Treasury bill rates on a discount basis N = number o f days to maturity 2) Second, calculate the yield-curve-adjusted three- and six-month CD rates using the follow ing formula: RCD3YCA = RCD3 - (Y 3 -Y 1 ) RCD6YCA = RCD6 - (Y 6 -Y 1 ) w h ere RCD3 = three-month CD rate RCD6 = six-month CD rate Y1 = one-month Treasury bill rate Y3 = three-month Treasury bill rate 3) Finally, the interest rate fo r large time deposits at commercial banks is given as follows: RLTDCB — MAX (RCD1, RCD3YCA, RCD6YCA). 1) Data on CD rates w ere not available before 1964.06 so RLTDCB was set to the ceiling rate on savings deposits o f less than one year as set by Regulation Q. 2) Before entering any calculations, the CD rates w e re multiplied by (365/360) to convert them to a bond, or 365-day, basis. http://fraser.stlouisfed.org/ FEDERAL RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis 51 RLTDTH — Rate on large time deposits at thrifts. RLTDTH is simply the rate on large time deposits at commercial banks (RLTDCB) plus 30 basis points based on Farr and Johnson’s result that the rate on large time deposits at thrifts has been about 30 basis points above that on large time deposits at commercial banks. RMMMFI — Rate on institution-only mutual fund shares. RMMMFI is simply the rate on general purpose and dealer/broker mutual fund shares (RMMMF). RTRP — Rate on term repurchase agreements. RTRP is equal to the rate on overnight RPs plus the difference o f the rates on term eurodollars and overnight eurodollars (RONRP + [RTED-RONED]). Asset data fo r term RPs is available back to 1969.10, whereas data fo r RONED is available only back to 1971.01. From 1969.10 to 1970.12 the spread (RTED-RONED) is estimated as the average difference betw een the tw o rates fo r 1971. RTED — Rate on term eurodollars. RTED is a yield-curve-adjusted rate computed from the one,three- and six-month term eurodollar rates. It is calculated in the same w ay as the rate on large time deposits (RLTDCB). 1) First, use annualized rates on one-, three- and six-month Treasury bill rates to calculate the yield-curve-adjusted three- and six-month term eurodollar rates (see the formulas from RLTDCB). 2) The RTED rate w ill then be the maximum o f the one-month term eurodollar rate and the three- and six-month yield-curve-adjusted term eurodollar rates. NOTES: l)O n ly data fo r the three-month eurodollar rate is available back to 1960.01, so RTED is just equal to that yield-curve-adjusted rate until 1963.05. 2) Before entering any calculations, the eurodollar rates w ere first multiplied by (365/360) to convert them to a bond-interest, or 365-day, basis. RSB — Rate on savings bonds. RSB is a six-month average rate fo r the current month converted to a bond-interest basis by multiplying by (365/360). RSTTS — Rate on short-term Treasury securities. RSTTS is simply the rate on the one-month Treasury bill. Data fo r the one-month Treasury bill rate is available only back to 1968.01, so data before 1968.01 is set at the three-month rate less the average difference betw een the one- and three-month rates fo r 1968.01 to 1990.12. Because this rate is quoted as a discount rate fo r a 360-day year, it is converted to an annualized one-month yield using the follow ing formula: RSTTS = ([(365*TBl)/100]/{360 -[(30 *TB 1) /100]}) *100 RBA — Three-month bankers acceptances rate. Although this rate has a three-month maturity, it is not yield curve adjusted as RLTDCB and RTED w ere because only one rate is used in the calculation (compared with three used fo r each o f the others). Instead, it is converted from a discount basis fo r a 360-day year to an annualized yield fo r a 365-day year using the follow ing form ula (see Farr and Johnson [1985]): RBA' = ([(365*RBA)/100]/{360 -[(90*RB A)/100]}) *100 RCPL — Rate on comm ercial paper. RCPL is the rate on one-month financial paper, which is converted from a discount rate fo r a 360-day year to an annualized one-month yield fo r a 365-day year using the follow ing formula: RCPL- =([(365*RCPL)/100]/{360 - [(30*RCPL)/100]})*100 RBAA — Rate on M oody’s Baa corporate bonds. RBAA is taken directly from the Board’s G.13 release and is used only in the computation o f the benchmark interest rate. It was yield-curve adjusted to a one-month annualized yield on a bond basis. NOVEMBER/DECEMBER 1992 52 BENCH — BENCH is the highest yielding rate fo r the period among all 24 interest rate series and the Baa bond rate; that is, BENCH = MAX [RBAA, (Rit, i = l , 24)]. Simple-sum and Divisia m onetary aggregates presented in this article can be downloaded from the FRED electronic bulletin board w ith a personal computer and a modem. The m onetary aggregates are in a file called "DIVISIA.” T o access FRED, dial 314-621-1824. Parameters fo r communication softw are should be set to no parity, w o rd length = 8 bits, one stop bit, full duplex and the fastest baud rate you r modem supports, up to 9,600 bps. For m ore information, telephone Tom Pollmann at 314-444-8562. http://fraser.stlouisfed.org/ RESERVE BANK OF ST. LOUIS FEDERAL Federal Reserve Bank of St. Louis FED ER AL RESERVE BANK OF S T. LO UIS REVIEW INDEX 1992 JA N U A R Y /FE B R U A R Y R. Alton Gilbert, “ Implications of Netting Arrange ments for Bank Risk in Foreign Exchange Trans actions” Jodi G. Scarlata, “ Institutional Developments in the Globalization of Securities and Futures Markets” Piyu Vue, “ Data Envelopment Analysis and Com mercial Bank Performance: A Primer with Appli cations to Missouri Banks” JU L Y /A U G U S T R. Alton Gilbert, “ The Effects of Legislating Prompt Corrective Action on the Bank Insurance Fund” Daniel L. Thornton, Targeting M2: The Issue of Monetary C ontrol” Steven Russell, “ Understanding the Term Struc ture of Interest Rates: The Expectations Theory” Mark D. Flood, “ The Great Deposit Insurance Debate” M A R C H /A PR IL David C. Wheelock, “ Monetary Policy in the Great Depression: What the Fed Did, and W hy” Manfred J. M. Neumann, “ Seigniorage in the United States: How Much Does the U.S. Govern ment Make from Money Production?” James B. Bullard, “ The FOMC in 1991: An Elu sive Recovery” K. Alec Chrystal and Cletus C. Coughlin, “ How the 1992 Legislation Will Affect European Finan cial Services” M A Y /JU N E Alison Butler, “ Environmental Protection and Free Trade: Are They Mutually Exclusive?” Michael J. Dueker, “ The Response of Market In terest Rates to Discount Rate Changes” S E PTEM B ER /O C TO B ER Mark D. Flood, “ Two Faces of Financial Inno vation” Kevin L. Kliesen and John A. Tatom, “ The Re cent Credit Crunch: The Neglected Dim ensions” John W. Keating, “ Structural Approaches to Vec tor Autoregressions” Anna J. Schwartz, “ The Misuse of the Fed’s Dis count W indow” Cletus C. Coughlin, “ Foreign-Owned Companies in the United States: Malign or Benign?” N O VE M B E R /D E C E M B E R Michael T. Belongia, “ Foreign Exchange Inter vention by the United States: A Review and As sessment of 1985-89” David C. Wheelock, “ Seasonal Accommodation and the Financial Crises of the Great Depres sion: Did the Fed ‘Furnish an Elastic Currency?’ ” Michelle A. Clark, “ Are Small Rural Banks Credit-Constrained? A Look at the Seasonal Bor rowing Privilege in the Eighth Federal Reserve District” Alison Butler, “ Is the United States Losing Its Dominance in High-Technology Industries?” James Bullard, “ Sam uelson’s Model of Money with n-Period Lifetim es” Daniel L. Thornton and Piyu Yue, “ An Extended Series of Divisia Monetary Aggregates NOVEMBER/DECEMBER 1992 Federal R eserve Ban k o f St. L ouis Post Office Box 442 St. Louis, Missouri 63166 The R e v ie w is published six times p er year b y the Research and Public Information Department o f the Federal R eserve Rank o f St. Louis. Single-copy subscriptions are available to the public f r e e o f charge. Mail requests f o r subscriptions, back issues, o r address changes to: Research and Public Information Department, Federal R eserve Rank o f St. Louis, P.O. Box 442, St. Louis, Missouri 63166. The views expressed are those o f the individual authors and do not necessarily reflect official positions o f the Federal R eserve Rank o f St. Louis o r the Federal R eserve System. Articles herein may b e reprinted provided the source is credited. Please provide the Bank’s Research and Public Information Department with a cop y o f reprinted material.