The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
mn« m i a 1^ ^ 3 ym ■1 The Economic Consequences of Wage-Price Guidelines ........................... i S T .l 2 El LITTLE R O C K Digitized for V o l . 6 0 , N O . 1 2 FRASER Are Prelim inary Week-to-Week Fluctuations in M l B iased? ................... .. 13 A Comparison of Yields On Future Contracts and Implied Forward Rates .. 21 The Economic Consequences of Wage-Price Guidelines MICHAEL E. TREBING OLUNTARY wage and price guidelines have now been adopted as a major element in the Govern ment’s anti-inflation program. The pricing behavior of firms and wage demands of labor are considered by a large portion of the population to be incompatible with the social objective of reducing inflation. Re straint in wage and price movements is believed to be necessary. Monetary and fiscal restraint alone appar ently have been judged as either not being able to accomplish this objective or as carrying too high a cost, in terms of lost output and employment.1 Although the guideline approach is popular with the public, remarkably little discussion has been di rected toward its implications.2 In particular, the probable efficiency and distributional consequences of the program have received little public attention. THE DECELERATION PLAN W a ge and Price Arithm etic The Administration has set explicit numerical stand ards for wage and price increases.3 The basic guide lines specify that annual increases in wage and fringe benefits be held below 7 percent and that price in creases be limited to 0.5 percent less than their an nual rate of increase during 1976-77.4 An alternative test for firms is to apply a “profit-test.” If a firm can not m eet the price standard, it is requested to limit its pre-tax profit margin on sales to the average of the best two of the past three years. In addition, total profit increases must be below a 6.5 percent ceiling, unless accounted for by volume increases. The program requires that deceleration of prices be achieved in each market, purportedly with individual JThis conclusion is clearly stated in White Paper: The Presi dent’s Anti-Inflation Program (accompanied the President’s announcement of the guideline program on October 24, 1978) pp. 1-4. 2A recent public opinion poll demonstrates the popularity of the adopted guideline policy. In a November 1978 Harris Poll 63 percent of the respondents supported the program. See Louis Harris, “Americans Support Anti-Inflation Plan,” St. Louis Globe-Democrat, October 20, 1978. 3In this article the word “standards” is used interchangeably with guidelines and guideposts. For details of the program see U.S. Council on Wage and Price Stability, Fact Book: Wage and Price Standards issued October 31, 1978. 4Ibid. pp. 20-40. The pay standard applies not to individual workers but to average pay increases for “groups” of workers. Page 2 firms sharing equally in the burden of lowering in flation.5 The target for inflation is 6 to 6.5 percent over the first year of the program. In order to reach these objectives, the program’s aim is to have prices rise at the same rate as unit labor costs, with average wage increases of 7 percent minus 1.75 percentage points for projected productivity growth yielding an inflation rate of 5.25 percent. The Administration al lows an additional 0.5 percent for “legislatively man dated payroll costs” and arrives at a rate of 5.75 per cent.6 The Administration states: “The wage/price standards are designed to serve as guides for the be havior of decision-making agents who have d iscre tionary p o w er ov er th e p rices an d th e w ag es that they receive.”7 [emphasis added] W hile the guidelines are “voluntary,” the Adminis tration has emphasized its intention to compel firms to comply by manipulating both Federal procurement policy and the Government’s broad regulatory author ity. The program also encourages that the force of public exhortation be directed at those large firms which exhibit “excessive” price increases. The Administration has requested that Congress pass a “real wage insurance” program. Under this scheme, workers who m eet the pay standard will re ceive a tax rebate if the rate of inflation exceeds 7 percent. The purpose of the rebate is to reduce workers’ fear of cooperation by insuring that they will not have their purchasing power reduced if the rate of inflation is not held to less than 7 percent.8 The price standards are directed at individual firms and apply to an “overall average price” and not to specific products. 5See White Paper, p. 7. 6Fact Book, pp. 15-16. Even with widespread compliance, the Administration concedes that prices will probably rise within the range of 6 to 6.5 percent. This would represent, however, an improvement over the first six months of 1978 when prices rose at a 10 percent annual rate. 7Ibid. p. 16. The paragraph continues: “Thus, standards are not directly relevant to pricing behavior in those markets in which prices are determined by the impersonal workings of supply and demand.” The program exempts raw materials and auction type markets which include (1 ) prices of agricultural and industrial raw materials, (2 ) interest rates, and ( 3 ) prices which historically have moved in tandem with an organized open exchange market. 8Note that a 7 percent pay increase and a 7 percent inflation rate gives zero increase in real income before taxes — even if productivity rises 1.75 percent. Given the progression in the income tax structure, real income (after taxes) declines. See Nancy Jianakopolos, “Paying More Taxes and Affording It Less,” this Review (July 1975), pp. 9-13. F E D E R A L R E S E R V E B A N K O F ST. L O U IS Incom es Policy and Inflation Voluntary wage and price standards can be classi fied as an “incomes policy”. This generic term, loosely defined, includes all of those actions taken by a gov ernment to affect the level of money incomes or prices by actively participating in wage and price decisionmaking. Although more popular in European countries, a wide range of incomes policies have been tried in the United States in recent years.9 Included have been relatively weak attempts to persuade or “jawbone” specific firms and workers to hold down wage or price increases in the spirit of social responsibility.10 Such a program was adopted during the Kennedy Administra tion and carried over into the early years of the John son Administration. At the other extreme, incomes pol icies have included former President Nixon’s rigid program of mandatory criteria for wage and price behavior throughout the entire economy. Guidelines represent an attempt to achieve a compromise between the two extremes. By strengthening the persuasive element used under the jawboning method while at tempting to avoid the harsh consequences of strict wage and price controls, guidelines represent a politi cally tempting route. WHAT “CAUSES” INFLATION? T h e “Cost-Push” View The acceptance of voluntary wage and price standards as an alternative prescription for reducing the general rate of inflation stems from the idea that inflation is generated by “cost-push” factors. This view describes how rising wages, the largest com ponent of business costs, continually force prices upward. The resulting inflation is known to the pub lic as a wage-price spiral. A similar version of this view concludes that inflation is the consequence of increases in the market power of firms and labor over the prices they charge. According to this analysis, prices and wages are “administered” by large firms and trade unions without regard to competitive mar ket forces. 9For historical surveys of incomes policy in the United States and abroad see U.S. Congress, Congressional Budget Office, Incomes Policies in the United States: Historical Review and Some Issues, May 1977; Lloyd Ulman and Robert J. Flana gan, Wage Restraint: A Study of Incomes Policies in Western Europe (Berkeley, California: University of California Press, 1971); Walter Galenson, ed., Incomes Policy: What Can We Learn From Europe? (Ithaca, New York: New York State School of Industrial and Labor Relations, Cornell University, 1973); and Craufurd D. Goodwin, ed., Exhortation and Con trols: The Search for Wage-Price Policy, 1945-1971 (Wash ington, D.C.: The Brookings Institution, 1975). 10These efforts are sometimes referred to as “moral suasion.” DECEMBER 1978 The cost-push view has great popular appeal since it depicts the inflation process as a struggle for in come shares between capitalists and workers. How ever, economic theory reveals that many implications of this view of inflation are illogical or, at best, ques tionable.11 It is argued that monopoly power exists in the market place and that firms have the ability to push prices above competitive levels and raise the average price level. But this analysis ignores the ques tion of why the monopolies had been charging less than the high monopoly price. The theory of monopoly pricing predicts that firms which have protection from the entry of competitors into their markets are able to receive prices above those of competitive markets. Once the monopoly price has been achieved, however, further increases are limited to the opportunities provided by the mar ket. If monopoly power is now causing prices to rise, either monopoly power is increasing or monopolists had been behaving irrationally and have just discov ered their market advantage. There is little evidence to support either alternative.12 Undeniably, many economic groups exhibit enough market power to influence the level of certain prices and wages. These monopoly prices are higher than they would be if the specific market were competi tive. But, except for a slight rise due to the resource misallocation, the overall level of prices and wages will remain substantially unchanged.13 For example, if wages in a particular industry are pushed up above competitive levels less employment will result. Labor will then be released for use in other sectors where a downward pressure on wages will result until a new equilibrium is reached. More importantly, however, this analysis is u n ab le to explain p ersistent in creases in prices, month after month, year after year. A M onetary View An alternative theory is that inflation is a monetary phenomenon. This view holds that changes in money 1'See George J. Stigler, The Organization of Industry (Home wood, Illinois: Richard D. Irwin, Inc., 1968), pp. 244-45 and Milton Friedman, “What Price Guideposts?”, Guidelines: Informal Controls and the Market Place, ed., George P. Shultz and Robert Z. Aliber (Chicago: University of Chicago Press, 1966), pp. 17-39. l2For a survey of the evidence regarding the relationship be tween market concentration and price changes see Steven Lustgarten, Industrial Concentration and Inflation (Wash ington, D.C.: American Enterprise Institute for Public Policy Research, 1975). 13One purpose of a union monopoly, for example, might be to gain real wage benefits for its rank and file. To accomplish this objective the union has several alternatives available to it. It may try to reduce the supply of labor through restric tive licensing practices or by not allowing non-union workers Page 3 F E D E R A L R E S E R V E B A N K O F ST. L O U IS growth exert a strong influence on total spending in the economy. W hen people find that they are holding cash balances which are greater than desired, they spend the excess money on real and financial assets and bid up their prices. The monetary view does not deny the existence of a wage-price spiral, but interprets the cost-push analysis as a confusion of the cause and effect relationships of the inflation process. According to the monetary view, the observed patterns of wage and price adjustments are normal responses to excessive money growth. For inflation to persist, the higher prices, no m atter w h ere th ey originate, must be validated by increases in the money supply. W ith money growth held constant, price increases can be maintained only through re duced production and employment. For such a sit uation to persist, businesses would have to willingly accept lower profits and labor would voluntarily re main unemployed and refuse to accept employment at lower wages. The empirical evidence does not support such irrational behavior.14 Only when mone tary authorities actively ensure that the spiral is fully augmented through increases in the money supply, will inflation result. WILL VOLUNTARISM SUCCEED? The underlying requirement for a successful guide line policy is that firms and wage-earners restrain themselves from acting economically as individuals. In a market economy the motive of individual selfinterest is crucial. Consumer preferences are revealed through the market by nonrestricted opportunities and/or purchases of goods and services at their mar ket price. These prices reflect not only the costs of production, but also the nature of demand for the good in question. The free movement of prices and the consequent incentives and disincentives that are created assure that resources in the economy will move toward satisfying these individual demands. DECEMBER 1978 with the social motives and wealth is transferred to those who stand apart from the program. The Kennedy Administration guideposts of 1962-67 represent a prime example of these conflicts. This voluntary program established upper wage limits which were equal to overall productivity gains for the economy.18 The problem which resulted in the demise of the productivity rule was that the guideline prin ciples did not take into account the fundamental pressure created by accelerated money growth. When the productivity guidelines were initially adopted, the trend of productivity growth was sub stantially above the rise in consumer prices. Wages based on the productivity rule thus provided for growth in real wages (money wages adjusted for price changes). As money growth accelerated and inflationary pressures intensified in 1965 and 1966, the wage standard became viewed as unfair. Labor, discovering that the purchasing power of their wages was falling, found the argument for holding down wages unacceptable. As key settlements began to ex ceed the guideposts, the program was abandoned. The degree of compliance with price guidelines will be associated with the severity of the penalties against those who choose to ignore them. If the con trols are truly voluntary and involve no costs for violation, there is little chance that they will succeed since gains from noncompliance can be realized with out costs. If economic sanctions are used against vio lators, however, each firm will weigh the expected costs and benefits involved. If the expected costs of noncompliance are less than the benefits, the firm will choose to ignore the guidelines.16 Avoidance can also take the form of “black-market” transactions above the controlled price or product quality changes. An appeal for individual restraint conflicts with a very basic economic observation about human be havior — consumers naturally strive to maximize their individual well-being. Econom ic self-interest is the major motivating factor behind economic activity. Guidelines, on the other hand, represent rules that substitute “social responsibility” for self-interest. The conflict between the two views is glaring. E co nomic incentives argue against in dividu al compliance Though the burden of the program is intended to be equally shared, this will not be the case. Gov ernment penalties through procurement policy will not affect private decisions in a uniform fashion. Some firms are dependent upon either government purchases of their output or are directly influenced by government policies. Other firms, however, may be outside the range of government sanction. Holding down the price of particular goods by penalties ben efits the purchasers of these goods and the sellers of unaffected competitive goods. The losers are the sellers of the controlled goods, those prospective buyers of the controlled goods who can no longer to obtain jobs. Secondly, the union might seek a higher wage through collective bargaining and thus accept the un employment forthcoming at this higher wage. 14See Denis S. Kamosky, “The Effect of Market Expectations on Employment, Wages and Prices,” Working Paper # 1 7 , Fed eral Reserve Bank of St. Louis. 15The Council of Economic Advisers, Economic Report of the President (Washington, D.C.: United States Government Printing Office, 1962), pp. 185-90. 16The same type of “cost-benefit analysis” will occur when labor contemplates compliance with wage guidelines and the rebate scheme which supplement them. Page 4 DECEMBER F E D E R A L R E S E R V E B A N K O F ST. L O U IS 1978 Toble II Table I Productivity C h a n g e s for Selected Industries, 1 9 7 1 -7 6 (Annual Rates of Change) W a g e s, Prices and Em ploym ent by Industry (1 9 7 1 - 7 7 ) (Annual Rates of Change) Output per Employee-hour1 Hosiery Synthetic Fibers Wet Corn Milling Aluminum Rolling & Drawing Pharmaceutical Preparations Paper, Paperboard, & Pulp Mills Petroleum Refining Steel Concrete Products Primary Aluminum Hydraulic Cement Copper Rolling & Drawing Coal Mining 11.1% 7.4 7.2* 5.5 4.3 2.5 2.1 1.8 0.8* — 1.4 — 1.5 — 1.7 — 4.3 *1971-75 xThe output per employee-hour figure is computed by dividing an output index by an index of aggregate employee-hours for produc tion workers. Sou rce: U .S. Department of Labor, P roductivity In d ex es F o r S elected In dustries, 1977 edition, p. 5. obtain them, and the purchasers of the competitive goods. The losses exceed the gains because of the misallocation of resources. The proposed “real wage insurance” plan which would supplement the guidelines, if enacted, serves to shift the burden of compliance among economic groups. If certain workers are guaranteed a constant real wage, the forthcoming rebates could reduce real incomes to the rest of the nation provided there is an increase in the federal deficit. To the extent that these larger deficits are “monetized”, inflationary pressures will be supplemented, thereby reducing the wealth of all holders of money and monetary instruments.17 The real-wage insurance program is said to be capable of breaking inflationary psychology and be able to bring about more rapidly the achievement of price stability. Lower expectations of inflation in the future, according to this view, would translate into lower demands for wage increases and eventually lower prices. This viewpoint, however, confuses how inflation expectations develop. Expectations p e r se do not cause inflation. Curbing expectations will require controlling of the underlying force which causes them to prevail. If price controls only delay the upward thrust of prices caused by expansive money growth, expectations of future inflation will not be reduced. If inflation expectations are not reduced by slower monetary growth, the longer-run objective of reaching price stability will be abandoned. F or example, since 17For an analysis of the administrative problem with the “realwage insurance” program, see Gardner Ackley, “Okun’s New Tax-Based Incomes Policy Proposal,” Economic Outlook USA (Winter 1978), pp. 8-9. Prices1 Manufacturing Construction Services Trade Finance Communications Transportation Utilities Mining Agriculture 6 .0 % 8.2 7.3 6.5 4.9 3.4 5.9 8.9 18.2 9.5 Employment2 Wages' 1.0% 1.1 3.8 3.1 2.8 0.9 0.8 1.1 5.1 2.7 7 .7 % 5.7 7.7 6.4 6.6 9.9 7.7 7.8 9.6 6.5 1Changes in implicit GNP price deflator by industry. 2Full-time and part-tim e employees. 3Labor compensation per full-time equivalent employee. Source: Su rvey o f C urrent B usiness. market pressures will eventually push prices upward, reaching the objectives of wage stabilization in the future will be made more difficult. Following a period of controls, a stable wage structure is far more likely to allow a resumption of moderate rates of wage in crease than a structure in which distortions, perpet uated by public policies, require rapid readjustment at the bargaining table. Experience with this type of “wage-price explosion” is well documented from European experience with control programs.18 MARKET DISTORTIONS AND CONTROLS Most economists agree that, for the sake of effi ciency, relative wages and prices should remain flex ible. Relations among the wages of workers of differ ent skills and of workers in different localities, industries or even firms should be allowed to vary according to changes in demand and supply. For example, firms which are growing have an incentive to hire scarce resources (labor and capital) away from other firms. Consequently, if upper limits are imposed upon the payments that can be offered to attract scarce inputs, the firms will not be able to meet the demand for their output. Relative prices, therefore, should be allowed to move in order to allo cate resources into their most productive uses. The dynamic character of the U.S. economy is evi dent from Tables I and II. The tables display that changes in employment and prices have varied across industries. Some industries have experienced rapid productivity growth; others have not. Employment growth has varied from industry to industry and gen erally reflects underlying demand conditions. The application of a single price and wage standard to all situations ignores this ongoing adjustment process 18Ulman and Flanagan, Wage Restraint, p. 223. Page 5 F E D E R A L R E S E R V E B A N K O F ST. L O U I S and confuses movements in relative prices (shifts of resources between economic groups) and the general level of prices. To minimize efficiency losses, it would be necessary to keep a watchful eye on individual wage and price relationships and make exceptions based on individual market situations. A tradeoff is therefore faced by the policymaker. The more strin gent the guideline (less exceptions) the greater the efficiency losses. On the other hand, “weak” guide lines are not likely to gain the acceptance of the popu lace who judge the probable success of the guidelines by the strictness of the program. The program is therefore unlikely to reduce inflation expectations. Implicit in the decision of who should be covered by guidelines are important judgments regarding the distribution of income among industries and be tween the factors of production. In other words, con trol programs politicize questions of income distribu tion. In a market economy, relative prices are signals that allocate society’s resources into their most pro ductive uses. Reflecting changing market conditions, these relative prices are always in motion and are independent of political criteria for distributing in come between economic groups. Any guideline based on a simple percentage price increase for all individ ual firms, however, is implicitly centered on an ac ceptance of wage and price relationships (at the time of policy implementation) as stable ones and assumes that the relationships will remain fixed throughout the period of the guidelines. D irect government controls, therefore, offer little inducement for the efficient development and use of resources, and contain no automatic mechanism for resource adjustments and the alleviation of shortages or excesses in production. Rather than being an aid to growth and vitality, they lead to retardation of economic resiliency and replace market forces by political ones. The U.S. experience with control programs dem onstrates these market misallocations. Price controls during World W ar II resulted in the substitution of low-quality goods for higher quality goods and black markets were commonplace as individuals developed lack of respect for the law. In later years, subsidies to producers became an increasing part of the con trol program as fixed prices were insufficient to provide the necessary incentives for production. Recent vol untary programs were also unable to avoid selective scarcities. For example, the Kennedy guideposts were blamed for shortfalls in supply of aluminum and sul fur and potential users were forced into using costly substitutes. Similarly, under the Nixon Administration controls, shortages developed for zinc, lead, steel, fer Page 6 DECEMBER 1978 Table III The 1 9 6 2 -6 6 G uid epo sts Period Annual Percentage Changes M l1 1962 1963 1964 1965 1966 W PI2 CPI3 2 .2 % 2.9 4.0 4.3 4.7 0 .3 % -0 .3 0.2 2.0 3.3 1 .1 % 1.2 1.3 1.7 2.9 1Demand deposits plus currency and coin held by the nonbank public. 2Wholesale Price Index. 3Consumer Price Index. Source: Departm ent of Commerce. tilizer, petrochemical products and a long list of other products. THE ROLE OF MONETARY ACTIONS The program is further complicated by the timing of monetary action. In order to validate decelerating inflationary pressures, it will be necessary to supple ment the program by tighter monetary action — re ducing growth in the money supply. But a problem exists in the timing of monetary actions and the con trol policy. Relations observed in the past indicate that previous changes in the money supply have ef fects on current variables — the pattern of aggregate spending is determined by past monetary actions.19 A perfectly timed effort by monetary and price con trol authorities will be difficult to achieve. The apparent failure of the Kennedy and Nixon control programs to reduce inflation can be inter preted in a monetary framework. A monetary explana tion for th e failure of th e 1 9 6 2 guideposts is evident in Table III. W hen the guidelines were adopted, consumer prices were rising at a moderate 0.7 percent rate. (The average change in the W holesale Price In dex between 1958 and 1964 was near zero.) Through out the life of the guideposts (1962-66), however, money growth increased steadily each year. The money stock grew 1 percent per year from 1959 to 1961, but increased steadily each year of the pro gram.20 Correspondingly, prices and wages moved upward reacting to the more rapid growth of spending. W hen the program was abandoned in 1966, consumer prices were rising at the rate of 3 percent. During the Nixon price control period (1971-74), money growth data reveals that the controls camou 19One study which provides a more detailed statement of the theory and evidence supporting these conclusions is Leonall C. Andersen and Keith M. Carlson, “A Monetarist Model for Economic Stabilization,” this Review (April 1970), pp. 7-25. 20Empirical support for the money-price relationship for the period 1955 to 1971 is presented by Denis S. Karnosky’s, “The Link Between Money and Prices — 1971-76,” this Review (June 1976), pp. 17-23. F E D E R A L R E S E R V E B A N K O F ST. L O U IS flage a period of overly ex pansive monetary policy during the 1970’s. Although non-monetary factors (the oil embargo and the shock of controls) temporarily in fluenced the money-price relationship, the growth of the price level by 1975 par alleled that rate predicted by the trend growth of money.21 1978 Trends and Fluctuations of Money, Prices, Output, and Unemployment Movements in the rate of inflation have been closely associated with movements in the trend growth of the money stock. The accom panying chart shows that both U.S. control periods since World W ar II have been marked by money growth above its long-run trend. Correspondingly, in both cases rates of change in prices eventually moved upward reflecting this long-run trend. W ithout curtailing aggregate spending, individual demands will be simply shifted among controlled and uncontrolled goods.22 Holding prices below their mar ket clearing levels will increase the quantity de manded of controlled goods. If total spending is not reduced through monetary or fiscal actions, those who are unlucky and do not receive the goods that would have been supplied without controls, will shift their spending to other products which represent, in most cases, close substitutes. The increased demand in un controlled markets will put upward pressure on these prices. CONCLUSION “Voluntary” wage and price guidelines have recently been adopted as an accompanying policy alongside the more traditional economic stabilization tools of monetary and fiscal policy. By establishing rules for pricing behavior, the Administration hopes to dampen a wage-price spiral that appears to be self-sustaining. 21Ibid. 22The word “controlled” refers equally to those “voluntary” re sponses that are reactions to government sanctions. DECEMBER According to the monetary view of inflation, the logical foundation of the control program confuses the results and causes of inflation. According to this view, inflation results when money growth persistently ex ceeds growth in the amount of money demanded. The observed wage and price adjustment (the so-called wage-price spiral) are but parts of the general re sponse in the economy to excessive money growth. Inflation expectations, which are generated by exces sive money growth, will be reduced only when the growth rate of money is slowed. Any short-term benefits received from strict com pliance with the guidelines will be costly. The uncon strained market system provides an efficient signaling system for moving resources between alternative uses. Any control framework will probably conflict with these price signals and will cause distortions which reduce the resiliency of the market system to changing market conditions. The emergence of black-markets and disguised price increases through reduced product quality are two examples of devices that have arisen in response to previous programs and may arise in the current program to circumvent the controls. Past incomes policies in the U.S. have been unable to reduce inflationary pressure because monetary ac tions remained expansive. If monetary actions remain expansive throughout the current program, acceler ated inflation appears inevitable. The fundamental forces of supply and demand cannot be repealed through any type of control program. Page 7 Is Inflation All Due to Money? ALBERT E. BURGER IN F L A T IO N is an all-pervasive problem which af fects everyone’s decisions. Individuals must consider the outlook for prices when planning budgets or wage demands, when deciding whether to buy a house or in what form to hold savings, as well as a multitude of other economic decisions. Also, business is increas ingly concerned about the outlook for inflation, espe cially as it relates to planning and capital investment.1 It is not surprising, therefore, that persistent inflation has led to increased public demands that something be done to correct the problem. The current Administration has responded to these demands by announcing an anti-inflation program which includes, among other aspects, a promise to intervene in individual price and wage decisions in an attempt to reduce inflationary pressures. But such an approach, at best, has only a very limited chance for success because it fails to distinguish between two key characteristics of the inflation process. First, there are increases (or decreases) in prices which result from nonmonetary factors that cover a gamut of influences such as the effects of weather on agriculture and actions of foreign oil producers. The basic character istic of all these nonmonetary factors is that they have a transitory influence on inflation. They have their impact on the level of prices in selected periods, but their influence is either reversed in following periods or ceases to be a cause of period-after-period changes in prices in the same direction. It is the sec ond aspect of inflation, the trend or persistent yearafter-year increase of prices, that is really “public enemy number one.” This is the aspect of inflation to which corrective economic policy must be directed. Otherwise, all other economic programs to stop infla tion will end in frustration. Contrasting Explanations of Inflation The rate of change of prices can show considerable short-term fluctuation. For example, the implicit price deflator for gross national product rose at a 5 percent 1John A. Tatom and James E. Turley, "Inflation and Taxes: Disincentives for Capital Formation,” this Review (January 1978), pp. 2-8. Page 8 rate in the third and fourth quarters of 1977, accele rated to about a 7 percent rate in the first quarter of 1978, rose further to an 11 percent rate in the second quarter, only to recede back to a 7 percent rate in the third quarter of 1978. In addition to this variation in the general price index, there are also frequent fluctuations in the prices of individual items included in the general price in dexes. Since pronounced swings in the prices of spe cific goods or services sometimes coincide with fluc tuations in the general index of prices, specific items are frequently cited as the cause of the current infla tion. Also, because the magnitude and timing of price changes vary from item to item, the blame for in flation is often transferred, from period-to-period, from one item to another. Consequently, a number of explanations of the inflation process have been offered, involving at various times the behavior of such diverse items as steel prices, exchange rate move ments, union wage demands, agricultural conditions, changes in minimum wages and even the behavior of the periodically elusive anchovy. Such an analysis provides an ever-changing array of inflation villains. The blame for inflation is shifted from Arabs to coffee producers to beef producers to steel producers to specific union leaders to large banks and so on. Concentration on such short-term oscillations in the various elements of price indexes clouds the issue of the fundamental force behind the persistent increase in the general level of prices. The problem of inflation is much more than an unfortunate sequence of in creases in the prices of particular items. Focusing attention on movements in the price of particular items or each wiggle in the general price indexes gives only a description of where and when the general infla tionary pressures fall in the economy. The important issue is why prices, on average, continue to rise over an extended period of time. An explan ation of the fundamental source of a con tinued pressure on prices requires a broader, longerrun perspective that incorporates monetary develop ments. W hen the money stock grows too rapidly relative to the rate of increase of goods and services, DECEMBER F E D E R A L R E S E R V E B A N K O F ST. L O U IS individuals find themselves holding more money than they demand, given existing income, prices, and yields (including interest rates) on other assets. In the proc ess by which they attempt to pull their holdings of money in line with the quantity demanded, inflation results. To put the matter more simply, when “too much money is chasing too few goods” there will be persisten t increases in prices. Consequently, analysis of persistent increases in the general level of prices requires consideration of the growth of the money stock. Such a monetary view contends that although prices can periodically rise or fall sharply due to nonmonetary factors, inflation continues only if these nonmonetary factors recur in succeeding periods, or if there is a continued excessive expansion of money. 1978 Table I M onetary G ro w th as an Indicator of Inflation (1) Period (2) Growth Rate of Money Growth Rate of Prices Period { i >- (2 : 2.3 % 1953 1.5% 0.8 2.7 1954 1.4 1.3 1949-54 3.2 1955 2.2 1.0 1950-55 3.3 1956 3.1 0.2 1951-56 2.7 1957 3.4 - 0 .7 1952-57 1.8 1958 1.6 0.2 1953-58 1.5 1959 2.2 - 0 .7 1954-59 2.0 1960 1.7 0.3 1955-60 1.3 1961 0.9 0.4 1956-61 The emphasis which is placed on the role of mone tary actions in the fight against inflation depends very much on which of these two aspects of price changes is the center of attention. Concentration on movements in individual prices or short-term movements in the general price indexes typically leads to assignment of a limited role to monetary actions, a focus of at tention on nonmonetary factors, and the recommen dation of some form of direct controls on the prices of specific items. In contrast, consideration of why prices continue to increase period after period, pin points the rate of monetary expansion as the prime factor in the fight against inflation. 1947-52 1948-53 1.5 1962 1.8 -0 .3 1957-62 1.8 1963 1.5 0.3 1958-63 2.2 1964 1.6 0.6 1959-64 2.2 1965 2.2 0.0 1960-65 3.1 1966 3.3 -0 .2 1961-66 3.6 1967 2.9 0.7 1962-67 4.0 1968 4.5 -0 .5 1963-68 4.8 1969 5.0 -0 .2 1964-69 5.2 1970 5.4 -0 .2 1965-70 5.1 1971 5.1 0.0 1966-71 5.5 1972 4.1 1.4 1967-72 6.1 1973 5.8 0.3 To illustrate the difference between inflation caused by monetary factors and short-term movements in price indexes caused by nonmonetary factors, consider the following simple monetary guide to inflation:2 The rate of change of prices over the next year is equal to the average rate of growth of the money stock over the previous five years. The results of using this shorthand representation of the driving force behind the inflation process and its long-run character are presented in Table I.8 The information in this table shows that, over the period 1953-71, past or trend growth rates of money were a reasonably good guide to the year-to-year behavior of prices. During this nineteen-year period, the 2For other examples of the use of monetary guides to inflation, see Richard T. Selden, “Inflation: Are We Winning the Fight?,” Morgan Guaranty Survey (October 1977), pp. 7-13, and Allan Meltzer, “It Takes Long-Range Planning to Lick Inflation,” Fortune (December 1977), pp. 96-106. 3Annual data are calculated as to the average of the four quarters of data in a given year. For example, the growth rate of prices from 1976 to 1977 on an annual basis is com puted by comparing the average of the four quarters in 1977 to the average of the four quarters in 1976. 6.2 1974 9.7 -3 .5 6.1 1975 9.6 -3 .5 1970-75 M oney and Inflation 1968-73 1969-74 6.2 1976 5.2 1.0 1971-76 6.0 19 77 5.9 0.1 average difference between actual yearly inflation and that indicated by the past rate of monetary ex pansion was only 0.2 percentage point, and in twothirds of the years the error was 0.5 percentage point or less. On a quarter-to-quarter basis, the rate of change of prices oscillated around the trend rate of inflation. However, the rate of change of prices re turned consistently to that dictated by the rate of monetary expansion. Also during this period, changes in the five-year trend growth of money accurately indicated changes in the year-to-year rate of inflation. As the trend growth of money slowed in the period 1958-63, in flation was reduced. Over the next eight years, the trend growth of money accelerated steadily from less than a 2 percent rate to a 5 percent rate, and inflation rose from less than 2 percent to 5 percent per year. In contrast to the 1953-71 period, the last six years present some examples of abnormally large differences Page 9 F E D E R A L R E S E R V E B A N K O F ST. L O U IS between changes in the price index and the inflation indicated by past growth rates of money. In particu lar, 1972 and 1974-75 stand out as glaring exceptions to the previous performance of the monetary guide to inflation. To understand the behavior of inflation since 1971, and how this experience fits into the general monetary explanation of inflation, it is crucial that one clearly understand the effect of nonmonetary factors on the behavior of prices. Specifically, it is very important to realize that, although the level of prices can change, sometimes even for a prolonged period, the rate of change of prices cannot continue to substantially deviate from the rate of monetary expansion.4 W hat special nonmonetary factors in 1972 and 1974-75 operated to cause such large deviations of actual changes in prices from those indicated by past growth rates of money?5 First, 1972 was a year of price controls. By law, reported prices were not allowed to fully reflect market pressures, especially those pushing prices upward. Under such circum stances, the reported change in prices would be expected to be considerably less than inflation in dicated by a monetary guide. From the perspective of a monetary interpretation of inflation, the gap in 1972 between price changes consistent with past money growth (about 5.5 percent) and those re ported during wage and price controls (about 4 percent) indicates (1 ) an upsurge of prices when price controls were removed, and (2) an incentive for transactions to take place at prices above posted prices. Other major differences between reported changes in prices and those indicated by past monetary expan sion occur in the more recent period of 1974-75. Over this period, the level of prices was sharply and un expectedly raised by the now well-known pricing actions of the major oil-producing nations and the nonmonetary effects of weather on agriculture. The actions of the Organization of Petroleum Exporting Countries (O P E C ) resulted in a substantial, unex pected rise in the price of energy. Since energy is a basic input into most production processes, these 4Even here, however, monetary factors still play a role, although indirectly. Autonomous events can have an effect on the de mand for money, which, if not matched by a one-time change in the money supply, result in a one-time increase in the level of prices. In such instances, prices rise not because of an excessive increase in the money supply but because of exces sive money balances created by a decrease in money demand. 5For a more complete technical discussion of the effects of special developments in 1972-74, see Denis S. Kamosky, “The Link Between Money and Prices— 1971-76,” this Review (June 1976), pp. 17-23. Digitized for Page 10 FRASER DECEMBER 1978 Table II G ro w th Rates o f Selected Com ponents o f Consum er Prices Period 11/71 _ IV /72 Food Prices 3 .9 % Energy Prices 3 .7 % All Items Monetary less Food Rate of and Energy Inflation 3.1 % 6 .0 % IV /72 — 1/74 19.6 22.3 4.6 6.0 1/74 — 111/75 8.8 13.9 9.7 6.0 111/75 — 11/78 6.2 6.9 6.6 6.0 O PEC actions had a widespread, and unexpected up ward effect on costs of production. There was a de crease in the effective productive capacity of the economy. W ith aggregate demand affected to only a minor extent and real output reduced, the lev el of prices rose sharply.6 Consequently, the rate of change of prices, computed over the period when these sharp upward adjustments in the level of prices took place, would be expected to be substantially, but tempo rarily, higher than that indicated by past monetary expansion. As their effect was absorbed in the econ omy, however, the rate o f ch an g e of prices fell back to that dictated by the trend rate of money growth. Although in 1976-77 inflation returned to the rate dictated by monetary expansion, the lev el of prices remained about 4 percent higher, reflecting the effect of the O PEC actions. Table II shows the movement of prices of selected groups of items during the period from mid-1971 to mid-1978. As show n in the table, price increases of all items other than food and energy were held to about a 3-4 percent rate while general price controls were in effect (August 1971 through April 1974). After controls were removed from most items, prices accelerated to about a 10 percent rate, as shown in the period 1/74 — III/75. Table II also clearly shows that the sharp surge in prices from late 1973 into late 1975 was initially led by the sharp rise in agri cultural and energy prices7 and then was reinforced by the adjustment of prices of all other items re sulting from the removal of price controls in early 1974. None of these components of consumer prices 6For a technical discussion of the effect of the rise in energy prices, see Kamosky, “The Link Between Money and Prices — 1971-76,” pp. 17-23; Robert H. Rasche and John A. Tatom, “The Effects of the New Energy Regime and Economic Ca pacity, Production, and Prices,” this Review (May 1977), pp. 2-12; and “Energy Resources and Potential GNP,” this Re view (June 1977), pp. 10-24. 7Price controls on agricultural products were removed in Sep tember 1973. The initial OPEC rise in oil prices came in late 1973. F E D E R A L R E S E R V E B A N K O F ST. L O U IS (food, energy, all other) continued the sustained double digit rate of increase. Since 1975 the average rate of increase of all these prices has fallen back into line with the sustainable rate indicated by the past rates of monetary expansion. Implications for M onetary Actions The above discussion has important implications for assessing the effects of past, current, and pros pective monetary actions in the battle against in flation. The experience of the last six years makes it clear that it can be just as misleading to ascribe each and every reported increase in prices entirely to monetary factors as it is to ignore the effect of money on inflation. Consequently, failure to separate the monetary (trend) and nonmonetary (transitory) aspects of inflation can lead to confused demands on policymakers. To illustrate the importance of this distinction, consider economic developments over the last six years. During the three-year period ended in the second quarter of 1971, the persistent rate of infla tion was very much in line with the rate indicated by a monetary guide to inflation. Over the next six quarters, however, prices rose at about a 4 per cent rate. Could this fall in inflation be attributed to monetary actions? The answer is no, the fall in rep o rted inflation was strictly due to nonmonetary factors, that is, price controls that went into effect in August 1971. From early 1973 through early 1975 prices rose very rapidly. From a 4 percent rate, inflation ac celerated to about an 8 percent rate in the year ended first quarter 1974. Then, over the next four quarters inflation took another sharp leap upward, averaging 11.6 percent. If one attributes all of these increases in prices during this period to the cumu lative effect of past monetary actions, then it appears that the Federal Reserve had let things get seriously out of hand. On the other hand, if the short-run in fluences of nonmonetary developments on prices are taken into consideration, quite a different conclusion emerges. Careful analysis of the effects on prices of weather, O PEC actions, and the removal of price controls would indicate a sharp rise in the level of prices beginning in late 1973 that was not the result of past monetary actions. The basic rate of inflation, the one determined by the cumulative effect of past monetary actions, remained at about 6 percent. Early in 1975, inflation dropped sharply, and aver aged 6.5 percent over the remainder of the year. DECEMBER 1978 Then inflation eased further to a 4.4 percent rate over the first three quarters of 1976. W as this sub stantial slowing in inflation the result of monetary policy actions? Again the answer is no. The slowing in the rate of change of prices from the double-digit pace of 1974 reflected only that the O PEC actions of late 1973 were not repeated in the following years, the general adjustment of other prices to the removal of price controls had been completed, and favorable agricultural conditions resulted in a sharp drop in the rate of increase of food prices. Did the basic infla tion slow to a sustained 4.5 percent rate by late 1976? Again the answer is no. From late 1976 to the end of 1977 inflation returned to a 6 percent rate, the same as that indicated by the trend growth of money. W hat was the effect of monetary actions, as mea sured by the growth of the money stock, on inflation over the six-year period 1972-77? In particular, what was the effect of allowing M l to grow at about an 8 percent rate from late 1971 to early 1973, then cut ting M l growth to 6 percent for a year, further slashing it to 4 percent for a year, and then progres sively reaccelerating M l growth, first to 5 percent for six quarters, and then to almost 8 percent over the two-year period ended in the third quarter of 1978? Did these gyrations in money growth substan tially change the basic rate of inflation? Using the past growth pattern of M l as a guide to inflation, then again the answer is no. Money had grown at a 6 percent rate over the five years (20 quarters) ended in the fourth quarter of 1971, remained at 6 percent in the 20 quarters ended in early 1975 and by the end of the fourth quarter of 1977 the twentyquarter growth rate of M l was still essentially 6 percent. Conclusions In analyzing the inflationary process, one must be careful to avoid shortsightedness. In particular, shortrun gyrations in prices must be distinguished from persistent changes in prices. Monetary policy cannot prevent the quarter-to-quarter fluctuations in the price level that naturally result from the dynamics of eco nomic activity. However, concentrating on only these short-run fluctuations in the level of prices can result in falsely blaming nonmonetary factors for a persistent rise in prices. The analysis of inflation then tends to bounce, month-to-month, quarter-to-quarter from one item or sector of the economy to another. Such an approach diverts attention from the role of monetary actions, results in failure to permanently reduce in flation and ultimately means that inflation will return Page 11 F E D E R A L R E S E R V E B A N K O F ST. L O U IS to plague the economy. The monetary actions of the government must be given a key position in any program to permanently reduce persistent inflation. Over the first half of 1978, prices rose at about a 9 percent rate. Should inflation be expected to continue at this rate? The monetary guide presented in this paper, indicates a persistent inflation of about 6.2 percent for the period 111/78 — 111/79. Some econo mists would contend that individuals now adapt their expectations of inflation more rapidly than previously, hence, a five-year trend rate of growth for M l is too long. If the period for calculating the trend rate of money is shortened to four years, the inflation indi cated for 111/78 — 111/79 rises to 6.4 percent. Short ening the period further to three years, raises the basic inflation rate to 6.8 percent for the next year. Consequently, the lasting rate of inflation indicated by past monetary developments falls in a fairly nar row range of 6.2-7 percent, nowhere near a 9 per cent rate. However, just because a rough monetary guide to inflation, such as the one presented in this article, does not indicate that past monetary actions have yet cumulated into a 9 percent persistent inflation should not be taken as a cause for rejoicing. A per sistent inflation of 6.5-7 percent is still at least three Page 12 DECEMBER 1978 times as fast as any lasting inflation the U.S. economy experienced from the end of World W ar II through 1965. Furthermore, historical evidence indicates that the development of such a persistent inflation is a rather sluggish process that does not adjust immedi ately to accelerations or decelerations of the growth of money.8 Currently, the trend rate of money growth is being held down by the 5 percent growth rate that pre vailed from the third quarter of 1973 to the third quarter of 1976. In sharp contrast, over the last two years (111/76 — 111/78), the average rate of monetary expansion accelerated to 8 percent. As the effect of the 5 percent growth wears off, if money continues to grow at an 8 percent or faster rate, inflation will rise sharply to a persistent, year-after-year, 8-9 percent rate. R For example, growth of money (M l) accelerated to about a 7 percent annual rate in 1968, after rising at an average rate of about 4 percent over the previous five years. Inflation did not rise to 7 percent in 1968, instead it was 4.5 percent, about in line with the 4 percent average growth of Ml over the previous five years. However, as the money stock continued to grow rapidly by past standards — at a 6 percent rate in 1969 — the five year average growth of M l rose to 5.2 per cent by the end of 1969 and the rate of inflation moved up to 5.4 percent in 1970. This increase in inflation took place even though the growth of Ml subsequently decreased to about a 4 percent rate in 1970. Are the Preliminary Week-to-Week Fluctuations in Ml Biased? COURTENAY C. STONE and JEFFREY B. C. OLSON F h E preliminary seasonally adjusted estimate for weekly M l — the money stock consisting of currency in the hands of the public and net private demand deposits — released each Thursday after noon by the Federal Reserve has become one of the most eagerly awaited, widely publicized, and closely watched of all economic statistics. Changes in stock prices, movements in interest rates, vari ations in the volume of trading on finan cial markets — even fluctuations in the foreign-exchange value of the U.S. dol lar — are frequently cited as conse quences of the public’s reactions to the week-to-week changes reported for the money stock. The impact attributable to the publication of these weekly money numbers has been described, with only slight hyperbole, by one economist as follows: Table I M eans and S tan d ard D eviations for Prelim inary S e a so n a lly A djusted Short-Run M l G ro w th R ates: 1 9 7 1 -7 7 One-Week Growth Rates Period Standard Deviation Mean Two-Month Growth Rates Standard Deviation Mean Standard Deviation 5 .5 % 1971 6 .4 % 2 5 .2 % 6 .1 % 6 .5 % 6 .3 % 1972 7.5 24.1 8.0 4.9 7.5 3.2 1973 8.7 35.7 5.6 5.1 5.4 4.1 1974 5.2 29.8 4.9 4.5 5.1 3.0 1975 4.6 25.8 4.7 7.7 4.7 5.4 1976 5.0 26.9 5.6 5.2 5.2 3.3 1977 7.2 29.9 7.2 6.6 7.3 3.8 Sources: The preliminary Ml data series used for computing the one-week growth rates were obtained from the initial estimates published in the Federal Reserve Statis tical Release H.6. The one- and two-month growth rates are from Alfred Broaddus and Timothy Q. Cook, “Some Factors Affecting Short-Run Growth Rates of the Money Supply,” Federal Reserve Bank of Richmond E con om ic R evieiv (November/December 1977), p. 4. E ach Thursday has become a Day of Judgement of anticipatory trembling over the latest Fed report on money supplies. E ach set of weekly statistics is combed as heralding a new wave of the business cycle, a new round of inflation, a new course of stock prices, and a new state of the economy ahead. Civili zation itself appears to hang in the balance.1 The attention devoted to these numbers recently motivated the Chairman of the Board of Governors of the Federal Reserve System to wish that “we could get away from the habit in this country of looking at those [money supply] figures every Thursday and assuming that the world is going up or down based on a weekly figure.”2 The growing popularity of this “habit” is puzzling to many economists for a variety of reasons. First, and perhaps most important, week-to-week fluctuations in M l are irrelevant for assessing the impact of money 'Sidney Weintraub, “Wall Street’s Mindless Affair with Tight Money,” Challenge (January/February 1978), p. 35. 2G. William Miller, “Hearings,” Second Meeting on the Con duct of Monetary Policy, U.S. Congress, Senate, Committee on Banking, Housing and Urban Affairs, 95th Cong., 2nd sess., April 25, 1978, p. 153. Mean One-Month Growth Rates growth on employment, output and prices. Only the longer-run variations in M l growth — over periods of several quarters or more — are generally consid ered to have significant effects on aggregate economic behavior. One-week growth in the money stock per se simply does not matter unless it can be used as a guide to the longer-term money stock movements. Second, as the period decreases over which the money stock growth rates are calculated, the greater is the influence of purely random events on the indi vidual growth rates — and the greater is the likeli hood of obtaining misleading results when using these growth rates to estimate the longer-run M l fluctua tions. An illustration of this problem appears in Table I which shows the means and standard deviations for annualized short-run growth rates of preliminary sea sonally adjusted M l for the 1971-77 period. Compar ison of the standard deviations, year by year, across the alternative short-run M l growth rates indicates that the one-week growth rates are more volatile than the one-month growth rates, which, in turn, display greater variation than the two-month growth rates. This greater variation around the mean growth rate Page 13 F E D E R A L R E S E R V E B A N K O F ST. L O U IS demonstrates how the impact of random events, which tends to “wash out” over longer periods, can mislead those who want to use the short-run growth rates to estimate the longer-term growth in M l. For ex ample, the preliminary rate of money growth for 1977 was about 7.2 percent regardless of which short-run money growth estimates are used. Yet, one-third of the week-to-week M l growth rates during that year were either less than -2 2 .7 percent or greater than 37.1 percent. This wider variation in the one-week growth rates makes it difficult to decipher the under lying longer-run trend growth in M l using the weekly money data. DECEMBER 1978 Table II M eans and Stan d ard D eviations for Fin ally Revised S e a so n a lly Adjusted O n e -W e e k M l G row th R ates: 1971- 77 One-Week Growth Rates The purpose of this article is to describe the nature of the bias that exists in using the preliminary money stock fluctuations to estimate the actual money stock movement on a week-to-week basis. As such, it in vestigates the extent to which the preliminary money stock estimates released each Thursday provide re liable information about the actual w eek ly growth in M l. This article demonstrates that the most widely cited of the money estimates, those for preliminary seasonally adjusted M l, are generally unreliable guides to the actual weekly growth in the money stock. Therefore, whatever explains the popular mystique as sociated with the Thursday release of the weekly money estimate, it does not appear to be due to its Page 14 Mean Standard Deviation 1971 6 .2 % 7 .4 % 1972 9.0 6.1 1973 5.6 9.1 1974 4.2 6.7 1975 4.5 11.5 1976 Finally, preliminary estimates of the money stock are subject to substantial revisions over an extended period of time after their initial public release. Com parison of the means and standard deviations for the finally revised one-week growth rates of seasonally adjusted M l for 1971-77, shown in Table II, with the equivalent statistics for the preliminary one-week growth rates in Table I provides an initial indication of the impact of the money stock revision process. The average one-week M l growth rates were revised upward for three of the seven years; four of the seven mean M l growth rates declined as a consequence of these revisions. Moreover, the volatility displayed by the one-week growth rates was substantially reduced as a result of the revisions. Because of the sizable ef fect of the revision process on the initially published growth rates for seasonally adjusted M l, the prelim inary one-week growth rates for M l may provide un reliable estimates of the actual movement in the money stock ev en on a w e e k -b y -w e e k basis. If the preliminary weekly money growth rates are biased, using them to estimate the longer-run growth in the money stock is even more troublesome. Period 6.1 11.0 1977 7.5 9.3 Sou rce: The finally revised M l series incorporates all revisions up to, and including, those appearing in the September 21, 1978 Federal Reserve Statistical Release H.6. usefulness in providing accurate information about the actual week-to-week growth in seasonally adjusted money. Revising the Prelim inary M oney Stock Estimates Although many economic data series remain vir tually unchanged once they are collected and pub lished, the money stock series are not among these. Exhibit I reproduces the first page of the Federal Beserve Statistical Belease H.6 — the initial public source of the preliminary weekly money stock esti m ates —■for T h u rsd ay , N o v em b er 2, 1978 to show one example of h o w the revision process affects the weekly M l numbers. There are several points to consider in Exhibit I. First, although the H.6 release is dated November 2, the most recent weekly money stock figures shown are those for the week ending on Wednesday, October 25; the weekly money stock is reported with a lag of eight days. Second, the H.6 release contains estimates for five different definitions of the money stock, M l through M5.3 Because M l is the most commonly cited money stock in the reports linking weekly money fluc tuations to financial market activity, only M l will be discussed in this article. Third, although financial ana lysts concentrate primarily on the behavior of the sea sonally adjusted money stock, the H.6 release includes estimates for both seasonally adjusted (SA) and not seasonally adjusted (NSA) weekly M l. Both are stud3Beginning with the November 16, 1978 H.6 release, an addi tional money stock measure, M 1+ , is now being published. F E D E R A L R E S E R V E B A N K O F ST. L O U I S DECEMBER 1978 Exhibit I A REPR O D U CTIO N O F THE FEDERAL RESERVE STATISTICAL RELEASE H .6 MONEY STOCK MEASURES In Billions of Dollars Mi m2 M i Plus Time De posits at Com mercial Banks Currency Other Plus Than Demand Large Deposits' CD's2 Date M3 M4 Ms For Immediate Release Nov. 2, 1978 Ml M2 M3 M4 M5 M2 Plus Deposits at Nonbank Thrift Institu tions3 M2 Plus Large Nego tiable CD's 4 M3 Plus Large Nego tiable CD's5 m T p I us M2 Plus Deposits at Nonbank Thrift Institu tions3 M2 Plus Large Nego tiable CDs* M 3 Plus Large Nego tiable CD's5 Time De posits at Com mercial Banks Currency Other Plus Than Large Demand Deposits' CD's2 Seasonally Adjusted Not Seasonally Adjusted 1977 — SEPT. 333.0 795.1 1344.9 858.9 1408.7 331.1 791.3 1339.7 856.7 1405.1 OCT. 335.9 801.4 1357.9 867.8 1424.3 335.2 798.7 1353.0 867.1 1421.3 NOV. 336.2 805.4 1367.1 876.3 1438.0 338.4 802.8 1360.1 874.4 1431.7 DEC. 338.5 809.5 1376.1 883.5 1450.1 348.2 814.9 1377.5 890.9 1453.4 1465.4 341.7 815.9 1386.6 892.2 1462.9 347.5 820.6 1389.0 897.0 FEB. 341.8 819.1 1393.1 898.5 1472.5 335.9 813.9 1386.0 890.8 1462.9 MAR. 342.9 822.6 1400.3 904.7 1482.3 338.2 821.1 1400.2 901.4 1480.5 APR. 348.5 830.3 141 1.4 913.7 1494.9 350.9 836.6 1421.2 917.9 1502.6 MAY 350.6 835.2 1419.9 922.2 1506.9 345.3 833.6 1420.3 918.2 1505.0 JUNE 352.8 840.6 1429.8 927.3 1516.5 351.7 842.0 1435.2 928.3 1521.5 JULY 354.2 846.2 1440.9 933.6 1528.3 356.0 848.7 1447.9 936.0 1535.2 AUG. 356.7 853.5 1455.1 939.8 1541.4 354.2 850.8 1452.9 938.8 1541.0 SEPT. 360.9 r 862.4 r 1472.0 r 950.5 r 1560.1 358.8 r 858.4 r 1466.4 r 948.7 r 1556.7 1978 — JAN . WEEK ENDING: 1978 — AUG. 30 SEPT. 6 355.5 854.3 940.9 350.0 848.0 936.8 361.4 861.3 948.8 360.3 859.4 948.7 13 360.5 861.7 950.3 362.4 861.9 952.1 20 361.1 862.6 951.8 360.3 858.8 949.5 27 361.8 864.1 95 1.7 353.0 852.6 943.3 953.1 4 360.2 864.5 951.4 360.5 862.5 11 r 364.3 869.4 955.8 r 364.9 868.6 958.4 18 P r 364.3 r 869.3 r 956.3 r 364.8 r 868.0 r 957.7 25 P OCT. 358.9 865.7 954.3 356.1 860.2 951.0 inclu d es (1) demand deposits at all commercial banks other than those due to domestic commercial banks and the U.S. Government, less cash items in the process of collection and F .R . F lo a t; (2) foreign demand balances a t F.R , B a n k s; and (3 ) currency outside the treasury, F.R . Banks and vaults of all commercial banks. 2Includes, in addition to currency and demand deposits, savings deposits, time deposits open account, and time certificates of deposits other than negotiable time certificates of deposit issued in denominations o f $100,000 or more by large weekly reporting commercial banks. 3Includes M2 , plus the average of the beginning and end of month deposits of mutual savings bank, savings and loan shares, and credit union shares. 4Includes M2, plus negotiable time certificates of deposit issued in denominations of $100,000 or more. includ es M3, plus negotiable time certificates of deposit issued in denominations of $100,000 or more. P — Prelim inary ; R — Revised ied in this article. Fourth, the two most recent weeks’ numbers are clearly designated as preliminary (as indicated by the “P” following their dates) to show that they are still being checked for processing errors. Finally, the previous two weeks’ M l numbers have been revised to correct an error detected since the previous H.6 release was published. Although processing errors in the estimation of M l occur irregularly, there are two standard revisions Page 15 F E D E R A L R E S E R V E B A N K O F ST. L O U I S that regularly affect the initially reported M l numbers — benchmark revisions and changes in the seasonal adjustment factors.4 Benchmark revisions in the money stock occur because, unlike the member bank data on vault cash and demand deposits which are available to the Federal Reserve each week, data for the majority of nonmember banks are reported to the Federal R e serve infrequently and then only for a one-week pe riod.5 Because weekly data for the periods between the nonmember banks’ reporting dates must be esti mated to obtain the preliminary weekly money figures, the money stock numbers are “subsequently revised as more information becomes available, in order to ‘benchmark’ the estimated weekly data to the few weeks of actual nonmember bank data.”6 As a conse quence of the correction of processing errors and in corporation of the benchmark changes, the prelimi nary not seasonally adjusted M l estimates are revised into final estimates of the NSA money stock over a period of months after their initial publication. These “final” NSA M l estimates are subject to yet further revision over a period of years whenever previously undetected processing errors are discovered or defini tional changes occur.7 The seasonally adjusted money stock is obtained by separately adjusting the currency and demand deposit components of NSA M l to take account of sea sonal patterns in money holdings.8 Therefore, in addi tion to being subject to benchmark revisions and correction of processing errors (which change the underlying NSA money stock components), the pre liminary SA money stock is also subject to revision if the initial seasonal factors used to obtain the season ally adjusted M l series are found subsequently to be inaccurate. The process of “firming up” the seasonal factors takes at least four years after the initial SA money stock numbers are publicly released. 4In this article, the term “processing errors” is used to indicate all revisions except benchmark revisions and changes due to reestimation of seasonal factors. 5Currently, FDIC-insured nonmember bank data are reported four times each year while data for noninsured nonmember banks are reported twice each year. For detailed explanations of the benchmark revision process, see Darwin Beck and Jo seph Sedransk, “Revision of the Money Stock Measures and Member Bank Reserves and Deposits,” Federal Reserve Bulle tin (February 1974), pp. 81-89, and Richard W. Lang, “Benchmark Revisions of the Money Stock and Ranges of Money Stock Growth,” this Review (June 1978), pp. 11-19. ®Lang, “Benchmark Revisions,” p. 11. 7Recently, for example, the money stock was revised back to mid-1975 to correct a bias discovered in the cash items adjustment. See the September 21, 1978 Federal Reserve Statistical Release H.6. 8For extended treatment of the seasonal adjustment of the money stock, see Thomas A. Lawler, “Seasonal Adjustment of the Money Stock: Problems and Policy Implications,” Federal Page 16 DECEMBER 1978 The money stock revision process represents a con tinuously ongoing attempt to produce more accurate money stock data. Consequently, the finally revised money stock numbers are not necessarily “final”. They are always subject to the possibility of additional future revision. However, if the revision process pro duces more reliable money stock data by correcting all known sources of error, the most recently revised money stock figures can be thought of as the best current estimates of the actual or “true” money stock. In the following discussion, the actual, or underlying, money stock is defined as the finally revised money stock incorporating all revisions up to, and including, those published in the September 21, 1978 Federal Reserve Statistical Release H.6, which contains the most recent benchmark revisions. M easuring the Reliability of the Preliminary W eekly Fluctuations in M l Because the weekly money stock estimates undergo a series of revisions after their initial release, questions concerning the reliability of the preliminary fluctua tions in weekly M l naturally arise. How closely do the preliminary weekly changes in M l, as derived from the H.6 releases, conform to the actual money stock changes after incorporating all corrections and revi sions? Do the growth rates computed from the ini tially reported M l numbers provide reliable estimates of the actual weekly growth in the money stock? The evidence from the 1970s suggests that the pre liminary money stock estimates are significantly af fected by the revisions that occur after they first ap pear in the H.6 releases. During the 1971-77 period, over 99 percent of the preliminary weekly money numbers were altered by subsequent revisions. The impact of these revisions on the week-to-week fluctu ations in M l can be determined by comparing the preliminary weekly change (or rate of growth) with the final change (or rate of growth) in M l after all revisions have been incorporated. Table I II presents summary statistics for this comparison using the NSA weekly money stock series for the 1971-77 period. Table IV presents similar results for the SA money stock. The absolute value, rather than the arithmetic value, of the difference between the final and the pre liminary changes (A M ) or annual rates of growth (%AM ) is used to focus on the magnitude of the dis crepancy between the initially reported weekly changes or growth rates in the money stock and their finally revised values. Reserve Bank of Richmond Economic Review (November/ December 1977), pp. 19-27. F E D E R A L R E S E R V E B A N K O F ST. L O U I S DECEMBER 1978 Table IV Table III M eans a n d Stan d ard D eviations for A bsolute Differences Between Fin a lly Revised and Initially Reported W e e k ly S e a so n a lly Adjusted M l Fluctuations: 1 9 7 1 -7 7 M eans a n d S tan d ard D eviations for A bsolute Differences Between Fin ally Revised and Initially Reported W e e k ly Not S e a so n a lly A djusted M l Fluctuations: 1 9 7 1 -7 7 NSA M l Changes (billions of dollars) Final A M I minus Preliminary A M I SA M l Changes (billions of dollars) Final A m i minus Preliminary A M I SA M l Growth Rates (annual percentage rates) Final % A M 1 minus Preliminary % A M 1 Period Mean Absolute Difference Standard Deviation Mean Absolute Difference NSA M l Growth Rales (annual percentage rates) Final % A M 1 minus Preliminary % A M 1 Mean Absolute Difference Standard Deviation 1971 $ .38 $ .32 8 .6 % 7 .4 % 1971 $ .74 $ .67 17.1% 15.7% 1972 .36 .30 7.5 6.3 1972 .83 .66 18.1 14.3 1973 .35 1.25 6.9 24.9 1973 1.16 1.16 23.3 23.7 1974 .40 .40 7.5 7.5 1974 1.17 .81 21.8 15.1 1975 .34 .28 6.2 5.0 1975 .95 .64 16.9 11.5 Period Mean Absolute Difference Standard Deviation Standard Deviation 1976 .35 .32 5.9 5.3 1976 1.03 .79 17.6 13.5 1977 .26 .26 4.1 4.2 1977 1.28 .83 20.6 13.2 1971-77 .35 .55 6.7 10.9 1971-77 1.02 .82 19.4 15.7 Comparison of the results shown in Tables III and IV yields two general conclusions about the effects of the revision process on the initially published weekto-week fluctuations in M l. First, the mean absolute differences are sufficiently large enough, given their standard errors, to be significantly different from zero.9 Therefore, the revisions in the money stock series have had a significant impact on the initially reported weekly movements in M l. Second, the money stock revisions have had a more substantial impact on the SA money stock fluctuations than on the NSA money stock movements. The mean absolute difference between the final and the prelim inary weekly changes or rates of growth in SA M l ranges from approximately two to five times the equivalent difference in NSA M l, depending upon the year of comparison. For the 1971-77 period as a whole, the mean abso lute difference between the final and preliminary week-to-week changes in SA M l was $1.02 billion; between the final and preliminary weekly growth rates, the mean absolute difference was 19.35 per cent. Thus, during this period, the final weekly change in SA M l differed in absolute value from its prelimi nary estimate by slightly more than $1 billion, on average, each week. Similarly, the final weekly growth 9Standard errors are obtained by dividing the standard devia tions by the square root of the number of weeks in the year. The mean absolute differences are all significantly greater than zero at the 5 percent level. in SA M l varied, on average, about 19 percent each week from the preliminary growth rate. Over the same period, these differences for the NSA money stock fluctuations were roughly one-third as large. The mean absolute difference for week-to-week changes in NSA M l was $.35 billion; for weekly growth rates, it was 6.67 percent. Another assessment of the reliability of the prelimi nary changes reported for M l can be obtained from the estimation of the following equation: (1 ) AMlt = a 0 + aiA M lPt where A M lt designates the actual change in M l from week t-1 to week t based on the most recently revised M l data and A M lP t designates the preliminary weekly change derived from the weekly M l numbers initially reported for week t-1 and week t. If the pre liminary week-to-week changes in M l (A M 1P) pro vide unbiased estimates of the underlying changes in the money stock (A M I), we would expect the esti mates to show that a 0 = 0 and a , = 1, or, alterna tively, that A M lt = A M lP t. A similar test for growth rates can be obtained by estimation of the equation: (2 ) %AMlt = | o + Pi%AMlPt 3 where % A M lt designates the actual annualized per centage growth rate in M l from week t-1 to week t based on the most recently revised M l data and % A M lP t designates the preliminary annualized per centage growth rate derived from the weekly M l Page 17 F E D E R A L R E S E R V E B A N K O F ST. L O U I S Finally, if the preliminary changes in M l are to be useful in estimating the actual changes in the money stock, the closer these two variables are related, the better. The R2 statistic calculated from the estimated relationship represents one measure of the closeness between the preliminary and the actual fluctuations in M l.11 Each R 2 shows, approximately, the propor tion of the total variation in the actual M l fluctuations that is associated with the fluctuations in the prelimi nary M l series. To the extent that the initially pub lished M l fluctuations closely parallel the actual move ment in the money stock after all computational errors have been corrected and the necessary revisions have been incorporated, the R 2 would be expected to have a value close to one. If the preliminary M l fluctuations do not closely anticipate the actual changes in M l after all necessary adjustments have taken place, the R 2 will have a value closer to zero. Thus, the closer the value of the R 2 is to one for the estimated relationship, the closer these variables are correlated. Assessing the Reliability of the Preliminary Not Seasonally Adjusted M l Fluctuations Table V shows the results obtained from estimating the above relationships between preliminary and final weekly NSA M l fluctuations over the period 1971-77. W hat do these tell us about the reliability of the pre liminary changes reported for weekly NSA M l? First, the preliminary weekly changes and rates of growth in the initially reported NSA M l appear to provide reasonably reliable estimates of the actual weekly changes occurring in the NSA money stock — despite the existence of various processing errors and bench mark revisions. The estimated coefficients for NSA M l over the entire 1971-77 period, displayed in the next to the last row in Table V, show that, if the one-week change in NSA M l was initially reported as $5 billion, for example, after all processing errors are corrected and benchmark revisions have been made, the actual change would be estimated to be $5.04 billion. Sim ilarly, if the one-week growth in NSA M l was initially 1978 Table V Tests for Bias in the Prelim inary O n e -W e e k Not S e a so n a lly A djusted M l Fluctuations Equation 1: Equation 2: AMlt = %AMlt = ao + aiAM lPt Period OCo Po CCi** R2 Po + P i% AMlPt Pi** 301 1 Id numbers initially reported for the respective weeks.10 Again, if the preliminary growth rates yield unbiased estimates of the underlying growth rates, we would expect the estimates to show that po = 0 and = 1, or, alternatively, that % A M lt = % A M lP t. DECEMBER 1971 .00 1.01 .95 .02 .98 .95 1972 .08 1.03 .97 1.58 1.00 .97 1973 - .1 1 .94 .79 -2 .0 8 .93 .78 1974 - .07 1.04 .97 -1 .3 6 1.04 .97 1975 - .01 1.03 .98 - .19 1.03 .98 1976 .04 1.02 .99 .60 1.02 .99 1977 .03 1.00 .99 .50 .99 .99 1971-77 - .01 1.01 .96 - .18 1.00 .96 Unbiased Values 0.00 1.00 1.00 0.00 1.00 1.00 **A11 qcj and 3 i coefficients are significantly different from zero at the 5 percent level. #Denotes oc0 or j30 coefficient significantly different from zero at the 5 percent level. * Denotes or B l coefficient significantly different from one at the 5 percent level. reported as 10 percent, for example, the estimate for the actual rate of growth in NSA M l is 9.82 percent.12 The reason that the preliminary changes and growth rates in the NSA M l so closely match the actual changes and growth rates is that the estimated co efficients do not differ significantly from those values necessary to assure that the initially reported fluctua tions in M l are unbiased (repeated in the bottom row of Table V ). All of the a 0 and | 0 estimates are numer 3 ically close to zero and none is significantly different from zero statistically. Similarly, all of the a , and (3! estimates are numerically close to one and none is significantly different from one. Overall, the results indicate that the week-to-week changes between the revised NSA M l numbers remain essentially the same as those initially calculated from the preliminary NSA money stock numbers. The R2 statistics for the NSA weekly money stock relationships over the 1971-77 period indicate that the initial changes and growth rates reported for NSA M l closely track the actual movements in NSA M l despite the existence of processing errors and bench mark revisions. Roughly 96 percent of the variation in the actual week-to-week changes in NSA M l are anticipated by the movement in the preliminary 10 %AMlt = 5200 (AlnM lt) and %AMlPt = 5200 (AlnMlP,). 11The R2 statistic is the coefficient of determination adjusted for degrees of freedom. Digitized for Page 18 FRASER 12 AMI = -.01 + 1 .0 1 (5 ) = 5.04; %AM1 = -.1 8 + 1.00(10) = 9.82. F E D E R A L R E S E R V E B A N K O F ST. L O U IS changes reported for NSA M l for the 1971-77 period as a whole. Similarly, the fluctuations in the initially reported annualized growth rates for NSA M l account for 96 percent of the actual movement in the rate of growth of NSA M l over the entire period. Year-byyear analysis confirms the closeness of the relationship between the initial and the final NSA M l fluctuations. These results indicate that the preliminary changes and growth rates reported in NSA M l provide reason ably accurate estimates of the actual changes occur ring in the money stock. DECEMBER 1978 Table VI Tests for Bias in the Prelim inary O n e -W e e k S e a so n a lly A djusted M l Fluctuations Equation 1: Period Equation 2: AMit = 0(0 + O i AM I Pt C % Amu = Po + P i% A M !P t ao a i* * R2 po P i'* R2 1971 .24 # .12 * .13 5.48 # .11 * .14 1972 .40 # .08 * .07 8.42 # .07 * .06 On the Reliability of the Preliminary Seasonally Adjusted M l Fluctuations The results from estimating the relationships for changes and rates of growth between preliminary and final seasonally adjusted M l, as shown in Table VI, indicate that the initially published SA money fluctuations do not provide accurate estimates of the actual movements occurring in the money stock after all revisions have been made. Using the results for the entire 1971-77 period, presented in the next to last row of Table VI, if the preliminary change in SA M l was $5 billion, for example, the estimate of the actual change that will be reported, after all processing error corrections, benchmark revisions and seasonal factor changes have been incorporated, is only $1.12 billion. Similarly, if the initially reported growth rate in SA M l was 10 percent, for example, the estimate for the actual rate of growth in weekly SA M l is only 6.73 percent. W hat accounts for the wide disparity between the preliminary changes and the final changes in the SA money stock? First, compare the estimated coefficients for the SA money stock relationships in Table VI with the values necessary to assure their reliability as shown in the bottom row of Table VI. Not only are the various estimates of a 0 and | 0 numerically greater 3 than zero, they are all statistically significantly differ ent from zero. This means that, even if the preliminary M l change was reported as zero, the estimate of actual change that will be reported after all re visions have been made is not zero, but rather ranges from $.18 to $.40 billion, depending upon the year of comparison, with an estimate of $.27 billion for the period as a whole. Similarly, if the weekly growth rate was initially announced as zero percent (that is, the preliminary SA money stock was unchanged from the previous week), the estimate is that M l had actually grown by more than 5 percent for that week, using the results for the 1971-77 period. Second, none of the ax and (3! estimates is close to one numerically and all 1973 .24 # .10 • .13 4.71 # .1 0 * .13 1974 .19 # .12 * .29 3.53 # .12 * .29 1975 .18 # .28 * .39 3 .1 7 # .28 * .38 1976 .29 # .24 * .34 4.89 # .24 * .34 1977 .38 # .21 * .44 5.99 # .21 * .44 1971-77 .2 7 # .17 * .28 5.13 # .16 * .25 Unbiased Values 0.00 1.00 1.00 0.00 1.00 1.00 **AI1 (Xi and $ 1 coefficients are significantly different from zero at the 5 percent level. # Denotes or J30 coefficient significantly different from zero at the 5 percent level. * Denotes or |3i coefficient significantly different from one at the 5 percent level. are significantly less than one statistically. Thus, it is clear that the final changes (or growth rates) in weekly SA M l are only slightly related to the prelim inary changes (or growth rates). The relatively poor correspondence between the preliminary and the final fluctuations in SA M l is also shown by the value of the R 2 statistics for the relation ships which range from .06 to .44, depending upon the year of comparison. Only about 28 percent of the actual fluctuations in the changes in M l, and only 25 percent of the actual movement in M l growth rates, are associated with the movements in the respective preliminary SA money stock estimates over the entire period. Put somewhat differently, more than 70 per cent of the actual variations in weekly SA M l changes and growth rates are not directly related to the varia tions in the preliminary M l estimates for the 1971-77 period as a whole. Since the seasonal adjustment process requires at least four years before the seasonal factors are con sidered final, only the earlier years, 1971-73, can be considered “fully” revised for seasonal purposes. Thus, it can be argued that the more recent of these R 2 statistics are misleadingly high — that the finally re vised changes and growth rates in SA M l are even less closely related to the preliminary SA M l move ments than these R 2 estimates indicate. Note that the Page 19 F E D E R A L R E S E R V E B A N K O F ST. L O U IS R 2 statistics for the earlier years are the lowest in Table V I. The SA M l estimates for the later years, 1974-77, are still undergoing seasonal revisions and will continue to do so for several more years. There fore, the R 2 value shown in Table VI for each of these later years are likely to be reduced when additional revisions occur. Consequently, the values of R2 shown for the later years, and for the 1971-77 period taken as a whole, probably overstate the closeness of the relationship between the initial movements and the finally revised SA M l fluctuations. Why are the preliminary weekly SA M l fluctuations unreliable while the preliminary NSA M l changes accu rately forecast the actual week-to-week changes in the NSA money stock? One approach to answering this question is to assess the importance of the dif ferent factors which cause the revisions in the pre liminary money stock numbers. The preliminary SA M l numbers are affected by the same processing errors and benchmark revisions that affect the pre liminary NSA M l numbers. In addition, they are affected by revisions of the seasonal factors. Since the preliminary NSA M l fluctuations do not generally appear to be unreliable, the problem with the initial SA M l numbers apparently is created by the revisions produced in reestimation of seasonal factors. The differential impact of processing errors and benchmark revisions, compared to those errors result ing from reestimation of seasonal factors, can be determined by analyzing the error associated with using the preliminary rate of growth in weekly SA M l as an estimate of the actual rate of growth in the SA money stock. Defining the “estimation error” to be the difference between the actual and the pre liminary weekly rates of growth, the estimation error for SA M l can be shown to equal the sum of the estimation errors for NSA M l and the seasonal ad justment factors. Analysis of these errors for the 1971-77 period shows that the estimation error asso ciated with the preliminary weekly growth rates in the seasonal factors accounts for approximately 70 per cent of the estimation error in SA M l growth rates.13 The preliminary weekly SA M l fluctuations provide 13If ESAM1, ENSAM1, and ESF represent the estimation errors for SA M l, NSA M l, and the seasonal factors, re spectively, it can be shown that ESAM1 = ENSAM1 + ESF. Consequently, VAR(ESAM l) = VAR(ENSAMl) + VAR (E S F ) + 2 COV(ENSAMl, E S F ). For the 1971-77 period, the latter equation had the following values: 623.7 — 164.4 + 556.6 + 2 (-4 8 .7 ). Page 20 DECEMBER 1978 generally unreliable guides to the movement in the actual money stock for any given week because the reestimation of seasonal factors introduces consider ably more erratic revisions than do the correction of reporting errors and benchmark revisions. Conclusion The week-to-week fluctuations in the preliminary seasonally adjusted M l, as reported each Thursday by the Federal Reserve, provide biased and generally unreliable information about the underlying weekly growth in the seasonally adjusted money stock. Earlier studies have commented on this problem for the pre liminary monthly and quarterly seasonally adjusted money stock estimates.14 Moreover, the Federal R e serve is sufficiently troubled by the lack of correspond ence between the preliminary and actual money growth rates that it has recently established a com mittee to study the seasonal adjustment process. Economists, by and large, have tended to ignore this issue because these extremely short-run varia tions in money are irrelevant for assessing the impact of money growth on employment, output and prices. It is only the longer-run fluctuations in money growth — over a period of several quarters or more — that generally are considered to influence these economic variables. The unreliability of the preliminary weekly growth rates in the seasonally adjusted money stock only poses a problem if financial market traders and mone tary policy authorities believe these rates accurately portray the underlying longer-term growth in money. Whatever explains the current fascination with the pre liminary week-to-week fluctuations in the seasonally adjusted money numbers, it clearly does not result from their usefulness in detecting the actual week-toweek growth in the seasonally adjusted money stock. 14See, for example, William Poole and Charles Lieberman, “Improving Monetary Control,” Brookings Papers on Eco nomic Activity, (2 : 1972), pp. 293-335; Report of the Ad visory Committee on Monetary Statistics, “Improving the Monetary Aggregates,” Board of Governors of the Federal Reserve System (Washington, D.C., 1976); Alfred Broaddus and Timothy Q. Cook, “Some Factors Affecting Short-Run Growth Rates of the Money Supply,” Federal Reserve Bank of Richmond Economic Review (November/December 1977), pp. 2-18; Herbert M. Kaufman and Raymond E. Lombra, “Short-Run Variations in the Money Stock,” South ern Economic Journal (April 1977), pp. 1515-27; and Rob ert D. Laurent, “Effects of Seasonal Adjustment on the Money Stock,” Federal Reserve Bank of Chicago Economic Perspective (September/October 1978), pp. 12-17. A Comparison of Yields On Futures Contracts and Implied Forward Rates RICHARD W. LANG and ROBERT H. RASCHE j^ I N C E the introduction of futures trading in 3-month Treasury bills in 1976, yields on these futures contracts have been examined for clues as to market expectations of the future course of interest rates. Al though there are difficulties in isolating these expec tations, the yields on futures contracts do embody information about market expectations of future in terest rates.1 However, similar information is also em bodied in the forward rates of interest that are implicit in the spot market yield curve. Yields on Treasury bill futures contracts (futures rates) are essentially the market counterpart to the implied forward rates embodied in the Treasury yield curve. The correspondence between yields on financial futures contracts and forward rates derived from a yield curve is readily apparent in the work of Sir John Hicks. Hicks interpreted the term structure of interest rates as a futures market for loans in formulating his theory about the relationship of long- and short-term interest rates.2 To the extent that futures rates and forward rates represent the yield on the same type of loan contract, market traders will arbitrage between yields in the futures market and yields in the spot market (from which implied forward rates are de rived) if profitable trading opportunities exist. In this case, it would not be surprising to find yields on Treasury bill futures contracts to be closely related to implied forward rates embodied in the Treasury yield curve. This paper compares yields on 3-month Treas ury bill futures contracts with forward rates derived from spot yields on Treasury securities, for compa rable periods, to examine how closely these interest rates are related. Specifically, this paper tests the hypothesis that futures rates are equal to implied foriFor a discussion of these difficulties, see Albert E. Burger, Richard W. Lang and Robert H. Rasche, “The Treasury Bill Futures Market and Market Expectations of Interest Rates,” •his Review (June 1977), pp. 2-9. 2J. R. Hicks, Value and Capital: An Inquiry into Some Funda mental Principles of Economic Theory, 2nd ed. (Oxford: Clarendon Press, 1946), pp. 144-47. ward rates, and finds that this hypothesis must be rejected. Various explanations as to why the rates are not equal are then examined. Recently, W illiam Poole and others have argued that the yields on 3-month Treasury bill futures con tracts can be expected to be less than the correspond ing implied forward rates, that these futures rates are unbiased market estimates of future Treasury bill spot rates, and that it is not necessary to allow for risk premia when using yields on futures contracts to measure market expectations of future interest rates.3 If these arguments are correct, a great deal of em pirical work in economics that includes variables on interest rate expectations will be greatly simplified. In addition, such conclusions would allow policymakers to easily assess the differences between their own in terest rate forecasts and the market’s expectations of the future course of interest rates. As Poole notes, policymakers face difficult problems when market in terest rate forecasts differ from the policymakers’ fore casts, since they then must decide whether their own estimates of economic activity are incorrect or whether th e m ark et is m isin terp retin g th e policym akers’ p lans.4 Unfortunately, the results reported in this paper do not support these conclusions about the relationship between futures rates and forward rates for futures contracts, except for the ones closest to delivery, which were the ones investigated by Poole. Extrapolation of Poole’s conclusions to other futures contracts is there fore unwarranted, and other explanations for the re lationship between forward and futures rates must be explored. One factor considered here is the possibility of default risk affecting yields on futures contracts. 3William Poole, “Using T-Bill Futures to Gauge Interest-Rate Expectations,” Federal Reserve Bank of San Francisco Eco nomic Review (Spring 1978), pp. 7, 14 and 15; and Kenneth Froewiss and Michael Gorham, “Everyman’s Interest Rate Forecast,” Federal Reserve Bank of San Francisco Weekly Letter (September 8, 1978), p. 1. 4Poole, “Using T-Bill Futures to Gauge Interest-Rate Expecta tions,” pp. 16-17. Page 21 F E D E R A L R E S E R V E B A N K O F ST. L O U IS EXPECTATIONS OF INTEREST RATES AND FORWARD RATES Expectations of future interest rates play an im portant role in many areas of economics; topics in both micro- and macroeconomics deal with interest rate expectations. Since such expectational variables generally are not observable, researchers have only been able to proxy them by using various substitutes, such as by constructing expectational variables on the basis of the past history of each variable. This ap proach is problematical in that when such expecta tional proxies are used in empirical research, a joint test is made: both the hypothesis and the assumed expectations-formation mechanism are tested. One alternative to such joint tests is to survey a specific group (such as financial consultants) as to their expectations of interest rates for various future periods. Such a survey has been reported since 1969 in the Goldsmith and Nagan B on d an d M oney M ar k et L etter.5 However, such surveys are subject to problems that may limit their usefulness. One prob lem is basically statistical, but another deals with the tim ing of the survey. The Goldsmith-Nagan survey is quarterly, which makes its use for shorter periods very difficult. An alternative approach that allows the use of daily data focuses upon ch an g es in interest rate expectations rather than levels, and is based upon changes in the shape of the yield curve from one date to another. A yield curve relates the yi'elds-to-maturity of a group of secu rities to their term s-to -m atu rity , fo r securities with similar characteristics other than maturity. In particular, all of the securities used in constructing a yield curve have similar default risk. For example, yield curves are usually drawn for Treasury securities, or for corporate Aaa bonds, as of a particular date. The yield curve indicates the structure of interest rates on a given date for securities with the same risk of default and different terms-to-maturity. Changes in the shape of the yield curve from one date to another involve changes in implied forward rates. A forward rate is the yield on a loan or invest ment over some period beginning at a specified future time. Such a forward rate can be obtained by an ap propriate combination of buying and selling bonds outstanding. For example, by selling a 1-year bond 5Other interest rate surveys have been collected by various researchers, but are not regularly published. For example, see Edward J. Kane and Burton G. Malkiel, “The Term Structure of Interest Rates: An Analysis of a Survey of Interest-Rate Expectations,” The Review of Economics and Sta tistics (August 1967), pp. 343-55. Page 22 DECEMBER 1978 and buying a 2-year bond, a 1-year investment is effectively made that will begin 1 year hence at a rate of interest established by the difference in the spot market yields for the 1- and 2-year bonds. The forward rate on this loan is defined by: M 4.1M (1+R2)2 (1-j-Ri)* where j F j is the forward rate on a 1-year loan to begin in 1 year, R 2 is the spot rate on 2-year bonds, and R, is the spot rate on 1-year bonds. More generally, for a 1-period investment to begin n -1 periods in the future the forward rate is: /1 , „ (1+n-lFi). - ((1-+f R n^)n - l l R n where n 1 F 1 is the forward rate on a 1-period loan to begin in n -1 periods, R n is the spot rate on n-period bonds, and R n t is the spot rate on (n -l)-p erio d bonds. Thus, the yield curve at any given point in time im plies a set of 1-period forward rates to prevail on forward (or future) transactions. Such forward rates have economic content, however, only if the implied transactions are possible in the market, and can be carried out by market traders.6 In theories of the term structure of interest rates, the forward rates (n iFx) are often decomposed into a 1-period expected rate („ -iE ,) plus a premium (a li quidity premium associated with interest-rate risk or a term premium associated with investors’ preferences for bonds with specific ranges of m aturities).7 n-iFi == n-iEi + Premium For a set of 1-period forward rates on a given date, there is then a set of 1-period expected rates stretch ing out into the future. Under the assumption that the premia are stable over time, changes in the structure of interest rates (measured by changes in the yield curve) reflect changes in interest rate expectations. Thus, by examining the changes in the implied for ward rates contained in the term structure, research ers can obtain an estimate of the ch an g es in interest rate expectations, even though the lev el of expected interest rates is not readily estimable. ,;For a thorough discussion of yield curves, forward rates of interest, and the term structure, see Burton Gordon Malkiel, The Term Structure of Interest Rates: Expectations and Be havior Patterns (Princeton: Princeton University Press, 1966), Chapters I and II. 7Malkiel, p. 26; Franco Modigliani and Richard Sutch, “Debt Management and the Term Structure of Interest Rates,” Jour nal of Political Economy (Supplement: August 1967), pp. 571-73; Charles R. Nelson, The Term Structure of Interest Rates (New York: Basic Books, 1972) pp. 20 and 28-31. F E D E R A L R E S E R V E B A N K O F ST. L O U IS However, such calculations are time consuming and costly — in terms of both data collection and com puter time. One must obtain quotations on securities outstanding (e.g. Treasury issues), fill in missing data points by estimating a yield curve, then calculate forward rates. This is a difficult task if done monthly, and expensive to do weekly or daily. As a result, it is expensive to use yield curve data to assess the effect of new information about economic policies or of the state of the economy on expectations of future in terest rates. FUTURES MARKETS IN FINANCIAL INSTRUMENTS Starting in the fall of 1975, the difficulties of exam ining changes in market expectations of future interest rates on a weekly or daily basis have been alleviated. Trading in futures contracts in financial instruments began to develop in late 1975, and currently there are futures markets in seven financial instruments.8 This paper focuses on the futures market in 3-month Treasury bills. Futures markets in 3-month Treasury bills allow us to observe directly the yields or prices on 3-month bills to be delivered at certain dates in the future. Thus, they are the market counterpart of the implied forward loans or investments which can be con structed from Treasury yield curve data. But instead of requiring large efforts at data collection, estima tion, and calculation, these yields are readily avail able from daily quotations in T h e W all Street Journal and other newspapers. As new information about the economy or economic policy becomes available to market traders, this in formation is incorporated into the market prices and yields of Treasury bill futures contracts. To the extent that such new information changes market expecta tions of interest rates, it is reflected in changes in the Treasury bill futures rates. Yields on futures contracts could also be broken down into expectational and pre mium components, just as in the case of forward rates implicit in the yield curve. Again, the lev el of expected future interest rates may not be readily estimable, but chan ges in market expectations of future interest rates can be observed from changes in yields on fu tures contracts, under the assumption that the premia are stable.9 8Currently there are futures markets in 3-month and 1-year Treasury bills, 15-year Treasury bonds, 3-month commercial paper, and three GNMA instruments. A number of other fu tures markets in other financial instruments have also been proposed. #Burger, Lang, and Rasche, “The Treasuiy Bill Futures Mar ket and Market Expectations of Interest Rates,” pp. 4-5. DECEMBER 1978 YIELDS ON FUTURES CONTRACTS AND IMPLIED FORWARD RATES FROM YIELD CURVES Since implied forward rates calculated from the yield curve are, in theory, rates on forward loans or investments such as those actually made in the Treas ury bill futures market, the question arises as to whether yields on 3-month Treasury bill futures con tracts are equal to 3-month forward rates calculated from the Treasury yield curve. It would be convenient if the two sets of yields were equal, so that we would not have to be concerned with any separate informa tional content of either data set (especially since yields on Treasury bill futures are easier to obtain). A test for the equality of the two yields, as of a quotation date, can be made by comparing yields on Treasury bill futures contracts with yields on implied forward contracts for the same periods. First, we choose a set of quotation dates. Then, we obtain the yields on Treasury bill futures contracts on those dates for each available delivery date. Next, we obtain quo tations on U.S. Treasury securities outstanding on those same quotation dates. From these data we cal culate implied 3-month forward rates that match the 3-month Treasury bill futures contracts. Finally, we calculate the difference (in absolute value) between the two sets of rates to determine whether they are significantly different from each other. T h e Data The selection of quotation dates for yields on Treas ury bill futures contracts and yields on outstanding Treasury securities were obtained by random selec tions of thirty quotation dates from each of three periods of roughly equal length — eight to nine months.10 The first thirty quotation dates were taken from the period March 1, 1976 to November 30, 1976 (Period I ) ; the second thirty quotation dates were taken from the period Decem ber 1, 1976 to July 31, 1977 (Period I I ) ; and the last thirty quotation dates were taken from the period August 1, 1977 to March 31, 1978 (Period I I I ) . Yields on the available futures contracts for each quotation date were based on the settlement prices obtained from the “Daily Informa tion Bulletin” of the International Monetary Market of the Chicago Mercantile Exchange. Yields on out standing U.S. Treasury securities used to construct forward rates for each quotation date were obtained 10The random numbers were obtained from The Rand Corpo ration, A Million Random Digits with 100,000 Normal Devi ates (Glencoe, Illinois: The Free Press, 1955). Page 23 F E D E R A L R E S E R V E B A N K O F ST. L O U I S DECEMBER 1978 Table I Summary Statistics for Absolute Differences: Futures Rates Less Forw ard Rates Categories' 1 2 3 4 5 6 7 8 0.13 30 6.47 0.16 0.12 30 7.30 0.35 0.21 30 9.13 0.58 0.34 29 9.19 0.58 0.45 23 6.18 0.51 0.31 11 5.46 0.09 0.06 30 8.22 0.14 0.11 30 6.97 0.34 0.20 30 9.31 0.37 0.26 30 7.79 0.63 0.35 30 9.86 0.88 0.34 29 13.94 0.97 0.26 12 12.92 1.43 0.44 12 11.26 0.19 0.13 30 8.01 0.26 0.15 30 9.49 0.16 0.34 0.22 30 8.46 0.48 0.29 30 9.07 0.56 0.27 30 11.36 0.55 0.31 30 9.72 1.01 0.36 28 14.85 Period 1 (3 / 1 / 7 6 - 1 1 /3 0 /7 6 ) Mean, X Standard Deviation, S Number of Observations, N t-statistic2 0.11 Period II (1 2 / 1 / 7 6 - 7 / 3 1 / 7 7 ) Mean, X Standard Deviation, S Number of Observations, N t-statijtic2 Period III (8 / 1 / 7 7 - 3 / 3 1 / 7 8 ) Mean, X Standard Deviation, S Number of Observations, N t-statistic2 0.11 30 7.97 C ategory 1 includes futures rates fo r the futures contract closest-to-delivery ; Category 2 includes futures rates for the futures contract next nearest-to-delivery, and so on. 2A11 t-statistics are significantly different from zero at the 1 percent level. from the Federal Reserve Bank of New York’s “Com posite Closing Quotations for U.S. Government Se curities.” All yields were converted from a discount basis to a bond equivalent yield basis. Forward rates were calculated for each quotation date to match up with each available Treasury bill futures contract. Thus, if a futures contract were to be delivered in 30 days, at which time the delivered Treasury bills would have 90 days to maturity, a for ward rate was calculated using the yields on an out standing Treasury bill maturing in 30 days and an outstanding Treasury bill maturing in 120 days. If no bills were outstanding with the exact number of days to maturity, say 120, then the yield was estimated by linearly interpolating from the yields on two se curities with maturities surrounding 120 days — say one with 130 days and one with 115 days. The result ing forward rate is the implied yield on a loan or “security” that begins in 30 days and has 90 days to maturity — the same time frame as the futures contract.11 n Spot rates used to calculate forward rates were the average of the bid and asked yields in the spot market. For futures contracts to be delivered more than one year in the future, yields on Treasury coupon securities were used (since Treas ury bills are not available) to calculate the forward rates. Page 24 Once the forward rates matching the available fu tures contracts were calculated for each quotation date, they were compared to the yields on the futures contracts (futures rates) by taking the absolute differ ence between the two. For each quotation date, these differences were categorized as being associated with the futures contract nearest-to-delivery (Category 1), next nearest-to-delivery (Category 2 ), and so on. All the available contracts for each quotation date were categorized in this way. W hen the market was first formed in 1976, only four contracts were traded, ex tending out one year into the future. As trading in Treasury bill futures has increased, the number of contracts has been extended. By March 1978, the end of the third sample, there were eight contracts traded, (This introduces a slight measurement error in the calcula tion of the forward rates since the formulae given below and in the text assume that the spot rates used are for non-cou pon securities.) The formula used to calculate the forward rates is that given by Richard Roll, The Behavior of Interest Rates: An Application of the Efficient Market Model to U.S. Treasury Bills (New York: Basic Books, Inc., 1970), p. 16: _ nRD - (n -9 1 )R n-9i A comparison of the above formula’s estimates of forward rates with estimates based on the traditional formula given in the text showed only minor differences. Consequently, the above formula was used for computational ease. DECEMBER F E D E R A L R E S E R V E B A N K O F ST. L O U I S extending out two years into the future. Consequently, the number of observations in Categories 4 through 8 are not always equal to 30 for each sample, and the first sample does not have as many categories as the last two samples. Table II Test Values of Com parisons of M ean Absolute Differences Across Sam ples Comparison of Periods Categories Results Summary statistics for the futures rates minus the associated forward rates are given in Table I for each category in each sample period. The mean of the absolute value of the differences between the rates are given for each category, as well as the standard deviation and the number of observations. For each category in each sample, the hypothesis that the fu tures rate is equal to the associated forward rate was tested by determining whether the mean absolute dif ference in each category is significantly different from zero. The t-value for each test is also given in Table I. The two futures contracts nearest to delivery (C ate gories 1 and 2 ) tended to have the smallest mean absolute differences between the futures and forward rates, while the contracts furthest from delivery tended to have the largest mean absolute differences. All of the mean absolute differences were significantly dif ferent from zero at the 1 percent level. Thus, al though the mean absolute differences between the futures and forward rates for the two futures con tracts closest to delivery (Categories 1 and 2 ) were generally less than 20 basis points, the hypothesis that the rates are equal is rejected in each sample. Samples were taken from three different time peri ods in order to determine whether the differences between the futures and forward rates have narrowed over time. Such an observation would suggest that in its first year of trading the futures market might have been poorly developed, or “thin,” in terms of the num ber of traders in the market and the availability of in formation about the market. W e could then expect that as the volume of trading in this market increased and information about possible arbitrage opportunities between futures and spot markets was more effectively utilized, the differences between the futures and for ward rates would decrease between the first and sec ond samples, and would decrease further between the second and third samples. Neither casual observation of the data in Table I nor statistical tests for significant differences across sample periods support the hypothesis that the differ ences between futures and forward rates have consist ently narrowed over time. Table II presents the results of statistical tests to determine whether the mean ab 1978 1 and II II and III 1 and III 1.93s 1 -1 .7 5 * 3 .8 3 1- 2 2 -0 .6 7 3.5 3 1, 2 3 -0 .1 9 — .3 2 1- 2 4 4 — 2.662 -0 .4 8 -3 .2 1 * 5 0.44 — 1.81 3 - 0 .9 3 1 6 3 .2 8 2 — .9 9 2 3 0.47 7 — — .4 7 2 4 — 8 _ — 2.912 -- 2.852 - 4 .3 9 1- 2 in d ica tes th at critical value of the significance test was determined using Cochran’s approximation to the Behrens-Fisher problem, see footnote 12 of text. 2Significantly different from zero at the 1 percent level. 3Significantly different from zero a t the 5 percent level. solute difference in each category of a sample was significantly different from the mean absolute differ ence in the same category in the other two samples.12 The results shown in Table II do not present a consist ent pattern over time. For example, a comparison of the mean absolute differences between the first and second samples for Categories 1, 2, and 3 indicates that the means are not significantly different from each other at the 5 percent level. Thus, the slight declines in the mean absolute differences for the first three categories nearest to delivery between the first and second samples do not represent statistically significant differences in the re lationship of the futures and forward rates. On the other hand, the increases in the mean absolute dif ferences between the second and third samples for the first two categories are statistically significant, as is the decrease for the third category. On the basis of this evidence, we cannot conclude that the differences between the futures and forward rates have been narrowing consistently over time as the futures market for Treasury bills has become more developed. Other explanations for the statistically sig12A t-test for the difference between two means generally re quires the assumption that the variances of the two samples are equal. When this assumption cannot be made, one is faced with what has been called a “Behrens-Fisher prob lem.” An approximation to the t-test due to Cochran that provides a solution is given in George W. Snedecor and William G. Cochran, Statistical Methods, 6th ed. (Ames, Iowa: Iowa State University Press, 1967), pp. 114-16. This method was used in calculating the t-values and their sig nificance in Table II for the cases where an F-test of the equality of the variances of the samples being compared rejected the hypothesis of equality. Page 25 F E D E R A L R E S E R V E B A N K O F ST. L O U IS DECEMBER 1978 Table III Summary Statistics for Arithmetic D ifferences: Futures Rates Less Forw ard Rates ___________________________________________________Categories1 Period I (3 / 1 / 7 6 - 1 1 /3 0 /7 6 ) -0.12 Mean, X 0.01 0.35 0.48 0.49 0.12 0.21 0.22 0.48 0.55 Standard Deviation, S 30 30 29 23 0.45 0.40 Number of Observations, N 30 t-statistic2 -5 .4 8 0 .2 6 1 8.71 5.39 4.27 — 0.08 0.04 0.32 0.37 0.62 0.88 0.97 1.43 0.08 0.17 0.23 0.27 0.38 0.34 0.26 0.44 11 3.73 Period II (1 2 / 1 / 7 6 - 7 / 3 1 / 7 7 ) Mean, X Standard Deviation, S Number of Observations, N •-statistic* 30 -5 .4 8 30 1.29’ 30 30 30 7.62 7.51 8.94 29 12 12 13.94 12.92 11.26 Period III (8 / 1 / 7 7 - 3 / 3 1 / 7 8 ) -0 .1 9 Mean, X - 0 .2 3 0.08 0.33 0.47 0.56 0.52 1.01 0.13 0.19 0.18 0.24 0.31 0 .2 7 0 .3 7 0.36 Standard Deviation, S Number of Observations, N 30 t-statistic2 - 8.01 30 30 -6 .6 3 2.43 30 7.53 30 8.30 30 11.36 30 7.70 28 14.85 'Category 1 includes futures rates for the futures contract closest-to-delivery; Category 2 includes futures rates for the futures contract next nearest-to-delivery, and so on. 2AU t-statistics are significantly different from zero at the 1 percent level, except for those with footnote 3 references. 3Not significantly different from zero a t the 5 percent level. nificant spreads between the futures and forward rates must be explored. EXPLANATIONS OF THE DIFFERENTIAL Given that there are significant differences between futures and forward rates that have not declined over time, the question arises as to whether or not these differences are systematic. If the differences are syste matic, can we identify some factor or factors that would cause such systematic differences? A further issue is to re-examine the argument that market traders will arbitrage away differences between fu tures and forward rates. This argument was based on the assumption that a futures contract is essen tially identical to an implied forward contract. If a futures contract is substantially different from an im plied forward contract, then market traders will not necessarily drive futures rates to equality with for ward rates. However, even if a futures contract is essentially identical to an implied forward contract, the existence of transactions costs in trading spot and Digitized for Page 26 FRASER future Treasury bills may provide few profitable arbi trage opportunities to traders. In this case, trading in spot and futures markets will not necessarily result in equalizing futures and forward rates. To examine whether there are systematic differences between futures and forward rates, the mean arith metic difference for each category in each sample period is given in Table III. The arithmetic differ ences are systematically negative in all periods for Category 1, zero or negative for Category 2, and systematically positive in all periods for Categories 3 through 8. W ith the exception of Category 2 in Periods I and II, all of the arithmetic differences are significantly different from zero. Thus, futures rates for contracts closest to delivery are generally lower than their associated forward rates, while futures rates for later-dated contracts are generally higher than their associated forward rates. Explanations of the spread between futures and forward rates must be able to account for both the spread itself and its change in sign as the delivery date is extended into the future. DECEMBER F E D E R A L R E S E R V E B A N K O F ST. L O U IS Transactions Costs: Poole’s A pproach In a recent article, W illiam Poole hypothesizes that futures and forward rates should not be equal because of the effect of transactions costs on these yields.13 He argues that transactions costs are basically zero for futures contracts but positive for trades in the spot market. Other factors affecting futures and forward rates (such as term or liquidity premia and interest rate expectations) are assumed to be about the same, while the effect of transactions costs would tend to increase forward rates. Consequently, Poole concludes that futures rates should be lower than forward rates.14 Poole obtains empirical support for his hypothesis by examining the futures contract closest to delivery (our Category 1). He finds that the mean (arithm etic) difference between futures and forward rates is in deed negative, indicating that futures rates are lower than forward rates. The mean difference also tends to be significantly different from zero; a result consistent with that reported in our Table III. In Poole’s subsequent discussion of the policy impli cations of the Treasury bill futures market, he assumes that his findings apply to all futures maturities (i.e., all categories in Table I I I ) , not just to the contract closest to delivery.15 This assumption is not supported by our data. The results shown in Table III indicate that Poole’s hypothesis holds only for Categories 1 and 2 (the two contracts closest to delivery). For the other contracts that are delivered further in the future, the futures rates are higher than the forward rates — contrary to Poole’s hypothesis. Poole seems to argue that futures rates are close to being equivalent to the market’s expectations of future interest rates. Quotes on the nearest maturity in the bill futures market can, therefore, be interpreted for all practical purposes as the market’s unbiased estimates of the future spot rates on 13-week bills.16 If the findings in the previous section apply to all future maturities, then the differences between the futures rates and the realized spot rates over the last two years reflect genuine expectational errors rather than term premiums attached to the futures rates.17 13Poole, “Using T-Bill Futures to Gauge Interest-Rate Expec tations,” pp. 7-19. 14Ibid., p. 14. 15Ibid., p. 15. 16Ibid. iHbid. 1978 The evidence presented here indicates that it is mis leading to extrapolate from the evidence on the futures contract closest-to-delivery to the later-dated con tracts. Futures rates on the later-dated contracts are generally 50 to 100 basis points higher than their as sociated forward rates, which suggests the existence of some substantial differences between the factors affecting the futures and forward rates. A D igression — A rbitrage Opportunities The relatively large and statistically significant dif ferences in Table III between the futures and forward rates for the later-dated futures contracts raises the issue of whether substantial arbitrage opportunities exist for these contracts. Poole investigated this issue for the contract closest to delivery and found that few arbitrage opportunities exist. Poole defined upper and lower critical points for profitable arbitrage for the futures rate given the spot yields, taking into account transactions costs. Values of the futures rates that lie between these upper and lower critical points indicate that profitable arbitrage opportunities do not exist. Poole calculated upper and lower arbitrage points using daily data between Jan uary 6, 1976 and June 23, 1977 for the contract closest to delivery. He found that profitable arbitrage oppor tunities rarely existed, and were small in magnitude when they did exist. By converting Poole’s formulae for the upper and lower arbitrage points to a bond equivalent yield basis (from his discount yield basis), we applied his ap proach to our three samples of data. In doing so, the formulae are not exact since the transactions costs associated with arbitraging the futures contracts fur ther from delivery are larger than for the contracts closest to delivery. This is because maturities for se curities in the spot market do not exactly match up with the maturities associated with the futures con tract. In addition, for futures contracts to be delivered more than one year out, yields on Treasury coupon securities were used to calculate forward rates (see footnote 11). Consequently, transactions costs associ ated with arbitraging the later-dated contracts would be higher than the ones used in Poole’s formulae. This means that our adoption of Poole’s formulae under states the upper arbitrage point, and overstates the lower arbitrage point. The spread between the upper and lower points is therefore understated, so that there may appear to be arbitrage opportunities which would not in fact be profitable if we took all the transactions costs into account. Page 27 F E D E R A L R E S E R V E B A N K O F ST. L O U IS Nevertheless, the application of Poole’s formulae will at least indicate the extent of arbitrage opportunities using a con servative estimate of the transactions costs involved. For each category of con tract in each sample period, Table IV shows the number of futures rates that are above the upper arbitrage point, b e low the lower arbitrage point, or within the upper and lower points. Table IV also shows the number of futures rates that are within or “close” to (defined as within .10 of) the upper or lower arbi trage points. DECEMBER 1978 Table IV Futures Rates Relative to A rb itra g e Points ____________________________ Category1 _________ 1 2 3 4 5 6 Number of Futures Rates: Below Lower Point Above Upper Point Within Points W ithin or "C lo se"2 10 0 20 28 6 1 23 27 0 19 11 18 1 18 10 11 0 12 11 13 0 5 6 8 Number of Observations 30 30 30 29 23 11 7 8 Period 1 (3 / 1 / 7 6 - 1 1 /3 0 /7 6 ) Period 1 1 (1 2 / 1 / 7 6 - 7 / 3 1 / 7 7 ) Number of Futures Rates: Below lower Point Above Upper Point Within Points Within or "C lo se"2 Aggregating over all three sample 5 2 1 0 0 0 0 0 periods, the results for Categories 1 and 0 5 20 14 18 26 12 12 2 tend to support Poole’s findings. Over 25 23 16 12 9 3 0 0 75 percent of the futures rates in C ate 29 14 26 19 15 6 1 0 gories 1 and 2 are within, or “close” to, Number of Observations 30 30 30 30 30 29 12 12 the upper and lower arbitrage points, Period III taking all three periods as a whole. How (8 / 1 / 7 7 - 3 / 3 1 / 7 8 ) ever, the percentage for Period III alone Number of Futures Rates: is considerably lower than for Periods I Below Lower Point 21 4 17 0 0 0 0 0 and II. Furthermore, the percentage Above Upper Point 0 0 5 8 12 15 17 27 tends to decline as the delivery date Within Points 9 13 21 22 18 15 13 1 extends further into the future. For cate Within or "C lo se"2 17 17 28 23 22 18 14 2 gories 6, 7, and 8 over all three sample Number of Observations 30 30 30 30 30 30 30 28 periods, the number of futures rates 1 includes futures rates for Category 2 within, or “close” to, the upper and lower ■Category futures rates for the futures the futures contract closest-to-delivery ;so on. includes contract next nearest-to-delivery, and arbitrage points are only 45, 36, and 5 in clu d es futures rates that are within the upper and lower arbitrage points as well as those th at are "close” in that they are within .10 of the upper or lower points. percent, respectively. Of course, the cal culation of the arbitrage points for these calculations of the upper arbitrage points for the laterlater-dated contracts are most likely to be subject to dated contracts are substantially underestimated, it error since they are based on yields on Treasury cou appears that systematic arbitrage opportunities fre pon securities rather than Treasury bills, and since the quently existed for the later-dated futures contracts spot maturities of the securities used do not match during our sample periods.18 up exactly with the later-dated futures contracts. Nevertheless, there are still some puzzling features If profitable arbitrage opportunities exist but are about the results. not acted upon by market traders, then we should not expect futures and forward rates to be as closely re First, when the futures rate falls outside the upper lated as we had earlier suggested, and we certainly and lower arbitrage points for the two contracts closest should not expect them to be equalized. Such a situ to delivery (Categories 1 and 2 ), it is almost always ation could explain the results obtained earlier, that b elo w the lower arbitrage point. Futures rates for futures rates and forward rates are not equal. How later-dated contracts, on the other hand, are almost ever, such a situation implies that there is a market always a b o v e the upper arbitrage point when they fall outside the upper and lower bounds. Second, when the futures rate is above the upper arbitrage point for the later-dated contracts, the difference between the futures rate and the upper bound ranges from less than 10 basis points to over 100 basis points (one full percentage point), and generally averages over 30 basis points in each category. Thus, unless the Digitized for Page 28 FRASER 18That frequent arbitrage opportunities have existed in the futures market has also been argued in two other papers. See Donald J. Puglisi, “Is the Futures Market for Treasury Bills Efficient?” The Journal of Portfolio Management (W in ter 1978), pp. 64-67; and Anthony J. Vignola and Charles J. Dale, “Is the Futures Market for Treasury Bills Efficient: A Comment,” The Journal of Portfolio Management (Winter 1979), forthcoming. F E D E R A L R E S E R V E B A N K O F ST. L O U IS inefficiency or failure present. Such inefficiency or failure could be due to lack of information about trading opportunities or to institutional constraints on trading. Since information about trading in futures markets is likely to improve over time, and since insti tutional constraints encourage innovations that reduce their effectiveness, such a situation of market ineffi ciency or failure will probably be reduced over time. If transactions costs are substantially larger than those used here, it may be that profitable arbitrage opportunities rarely existed despite the large spreads between the futures and forward rates for the laterdated contracts. In this case, we again should not expect futures and forward rates to be as closely related as was earlier suggested. Given transactions costs, futures rates and forward rates may not be equalized. However, Poole’s argument based on transactions costs led him to conclude that futures rates should be lower than forward rates. Even though transactions costs might explain why futures rates and forward rates are not equalized, it is still puzzling that futures rates are substantially higher than their associated forward rates for the later-dated contracts, contrary to Poole’s argument. This suggests that factors other than transactions costs may affect futures, rates differ ently than forward rates, and we now turn to a con sideration of these other factors. D efault Risk Poole implicitly assumes that factors other than transactions costs have the same effects on both fu tures and forward rates. Thus, since transactions costs are expected to increase forward rates, and since transactions costs are close to zero for futures con tracts, Poole concludes that futures rates will be less than their associated forward rates.19 That this con clusion is not supported by evidence for the laterdated futures contracts suggests that there are other factors embodied in futures and forward rates that have effects in th e op p osite direction to the transactions-cost effect discussed by Poole. Furthermore, this effect is stronger for the later-dated futures con tracts than for those close to delivery. Both futures rates and forward rates can be broken down into expectational and premium components. Since one-period expectations of future interest rates should be the same in both rates, we must consider the premium components of these rates. The premium 19Poole, “Using T-Bill Futures to Gauge Interest-Rate Expec tations,” p. 14. DECEMBER 1978 associated with a forward or futures rate is generally considered to be a liquidity premium associated with interest-rate risk, or a term premium associated with investors’ maturity preferences. For a liquidity premium embodied in a futures rate to be different from the liquidity premium embodied in a comparable forward rate implies that the interestrate risk associated with the futures contract is differ ent than that associated with the comparable implied forward contract. For futures rates to be higher than forward rates for the later-dated contracts as a result of differences in liquidity premia, a given rise in interest rates would have to generate a larger risk of capital loss in the futures contract than in the implied forward contract. It is not obvious why this would be the case. For a term premium embodied in a futures rate to be different from the term premium embodied in a comparable forward rate implies that investors’ ma turity preferences vary both across maturities and across financial instruments. It is again not obvious why this would be the case. One factor that has been ignored in the discussion of futures contracts is default risk. Treasury bills traded in the spot market are considered to be default free. Hence, implied forward rates would not embody premia related to default risk. However, a futures contract is not guaranteed by the U.S. Government, but is rather guaranteed by the exchange on which it is traded. Although the futures contract involves de livery of Treasury bills that are default free, the con tract itself is not default free. Consequently, the futures rate may contain a risk premium associated with default risk. This default risk factor would be more important for the futures contracts that are further from deliv ery, those for which Poole’s hypothesis fails to be supported in our samples. The furthest-dated futures contracts involve delivery of Treasury bills which have not yet been issued; they do not exist. The possibility exists, although it may be small, that there would not be a sufficient amount of 3-month Treas ury bills available to meet the deliveries required by the number of open futures contracts held for deliv ery. Although the Chicago Mercantile Exchange guarantees that a settlement would be made, at least a monetary settlement, the item promised for delivery (3-month Treasury bills) may not be delivered.20 -°Defaults on futures contracts for commodities are rare, but result in quite an uproar when they do occur. A recent ex ample was the May 1976 default on the delivery of Maine Page 29 F E D E R A L R E S E R V E B A N K O F ST. L O U IS This risk of default, or risk of non-delivery of the Treasury bills, would tend to make yields on the laterdated futures contracts higher than the yields on the two contracts closest to delivery (where Treasury bills that can be used for delivery have been issued), other things constant. The results shown in Table III are consistent with this hypothesis. However, whether or not the size of the spreads between the futures and forward rates for the later-dated contracts can be accounted for solely by default risk is an open question. SUMMARY AND CONCLUSIONS Since yields on futures contracts are the market counterpart of implied forward rates of interest de rived from the yield curve, the hypothesis that fu tures rates and forward rates are identical was tested using data from the Treasury bill futures market and the spot market for Treasury securities. The results indicate that futures rates are significantly different from the associated forward rates. Furthermore, the differences between the two rates have not narrowed consistently over time. Thus, it is difficult to attrib ute the significant differences between the two rates as being due to the initial “thinness” in the devel opment of the Treasury bill futures market. DECEMBER 1978 results on the contract closest-to-delivery should not be extrapolated to other futures contracts.21 An explanation which is consistent with the empiri cal results is that there is a default risk premium that affects and is embodied in the futures rates (since the futures contracts themselves are not obligations of the U.S. Government) but that does not affect the for ward rates. The default risk would be greater for the later-dated contracts, which involve delivery of Treas ury bills not yet issued, than for the contract closest to delivery, which Poole investigated. Although fur ther testing and examination is required to fully explore the implications of the evidence given here, the consideration of the default risk of futures con tracts should be a useful starting point. The results of this study imply that we cannot interpret yields on later-dated 3-month Treasury bill futures contracts as the market’s unbiased expecta tions of future spot rates on 3-month Treasury bills. Futures rates do not necessarily reflect the expected lev el of future interest rates. However, these results do not conflict with the proposition that chan g es in market expectations of future interest rates can be inferred from chan g es in futures rates. Poole’s argument that the two rates should not be equal, but that the futures rate should be below the forward rate, was also examined. Poole’s results were based on the effect of transactions costs on forward and futures rates, and were supported by evidence using the futures contract closest to delivery. Results from our samples for later-dated futures contracts do not support Poole’s hypothesis. Instead, we find that the futures rates are consistently above the for ward rates for the later-dated contracts. Thus, Poole’s The examination of interest rate expectations em bodied in futures rates is therefore more complicated than Poole’s results suggest. Furthermore, if default risk is a significant factor affecting futures rates, then estimates of term or liquidity premia in forward rates will not be comparable to the premia embodied in futures rates. This would make the estimation of the levels of expected future interest rates even more difficult. Consequently, policymakers who want to compare their own interest rate forecasts to the mar ket’s expectations should use caution in employing futures rates to measure market expectations. potato futures. Recently a bill was introduced in Congress that would ban all futures trading in potatoes. 21Poole, “Using T-Bill Futures to Gauge Interest-Rate Expec tations,” p. 15. Page 30 F E D E R A L R E S E R V E B A N K O F ST. L O U I S DECEMBER 1978 REVIEW INDEX - 1978 Issue Jan. Title Inflation and T axes: Disincentives fo r Capital Form ation Issue July Pitfalls to the Current Expansion D oes the F ederal Reserve Invest M em ber Bank Reserves? C oordinated International Econom ic Expansion: A re Convoys or Locom otives the Answer? Aug. Oil Im ports and the F all o f the D ollar An Explanation o f M ovements in the Labor Force Participation Rate, 1957-76 M oney-Income R elationships and the Exchange Rate R egim e Sept. Inflation, U nemployment, and M oney: Com par ing the Evidence from Tw o Sim ple M odels D oes the Stage o f the Business Cycle A ffect the Inflation Rate? Effects o f State Reserve Requirements Oct. Taxation o f Capital Gains: Principle Versus Practice Budget Deficits and T rade D eficits: Is there a Link? Econom ic G row th and U nemployment: A R e appraisal o f the Conventional View Nov. M ovements in the Foreign Exchange Value o f the D ollar During the Current U.S. Expansion International Reserves and the R ole o f Special Drawing Rights O utlook fo r F o o d and Agriculture Feb. T h e M echanics o f Intervention in Exchange M arkets A Tax-Based Incom es Policy (T IP ): W hat’s It A ll A bout? D oes th e St. Louis Equation N otv B elieve in Fiscal Policy? Mar. T h e F ederal Open M arket Com m ittee in 1977 B ank Reserve Requirem ents and T heir E n force ment: A Com parison A cross States Apr. T h e Recent U.S. Trade D eficit — N o Cause for Panic H ave M ultibank H olding Com panies A ffected Com m ercial Bank Perform ance? O perations o f the F ederal Reserve Bank o f St. Louis — 1977 May A Perspective on the Econom y: T h ree Y ears o f Expansion Commentary on Monetary Economics: A n Inter view with Karl Brunner C om paring Per Capita Output Internationally: Has th e United States Been Overtaken? Rising Farm land Prices and Falling Farm Earn ings: Is Agriculture in T rouble? June Im ports and Jo b s — T h e O bserved and the U nobserved Benchm ark Revisions o f the M oney Stock and Ranges o f M oney Stock G row th Title Dec. T h e Econom ic Consequences o f W age-Price Guidelines Is Inflation A ll D ue to M oney? Are Preliminary W eek-to-W eek Fluctuations in M l Biased? A Com parison o f Y ields On Future Contracts and Im plied Forw ard Rates Page 31