The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
FEDERAL RESERVE BAN K OF ST. LOUIS SEPTEMBER 1970 Economic Slowdown and Stabilization Policy ST LO UlS Selecting a Monetary Indicator — Evidence from the United States and Other Developed Countries...................... 8 EIC5HTH C . A. LITTLE R O C K V o l. 5 2 , N o . 9 Economic Slowdown and Stabilization Policy HE PACE OF real economic activity remains sluggish, as the response to last year’s monetary and fiscal restraint continues. Despite the slowdown in production and employment, the pace of inflation has not yet moderated significantly. Major questions regarding the economy include: (1) How long will real economic activity continue so sluggish and, once recovery begins, how strong will it be? (2) What are the prospects for an easing in infla tionary pressures? The course of prices and output in coming quarters will be influenced greatly by mone tary and fiscal actions which have already occurred, as well as by forthcoming actions. D em and and Production Ratio S c a le Trillions of D o lla rs Q u arte rly Totals at Annual Rates S e a s o n a lly A djusted Ratio S c a le Trillion s o f D o lla rs 1.1 1.0 .9 .8 .7 .6 L l G N P in current dollars. [2 G N P in 1958 dollars. Pe rce n ta g e s a re annual rates of change for p eriods in d icated . Latest d a ta plotted: 2nd quarter Page 2 Source: U.S. Department of Commerce Pattern of Econom ic Slowdown Growth of total spending for goods and services began slowing in late 1969, in response to antiinflationary monetary and fiscal actions. Such spending rose at a 4 per cent annual rate from third quarter 1969 to second quarter 1970, following an 8 per cent aver age rate of advance in the previ ous five quarters. Continued rapid increases in prices, reflecting ex cessive total demand in the pre vious five years, along with the reduction in growth of total spend ing, resulted in a decline in real product. Total real output (GNP adjusted for changes in prices), which rose at a 3 per cent average annual rate in the five quarters ending third quarter 1969, de clined at a 1 per cent rate from third quarter 1969 to second quar- FEDERAL RESERVE BANK OF ST. LOUIS ter 1970. Industrial production, which is usually more sensitive than total output to changes in total spending, has decreased at a 3 per cent average annual rate since mid-1969. This decline in production follows a rise of 4.9 per cent from mid-1968 to mid1969 and a 6.5 per cent gain in the previous year. SEPTEMBER, 1970 Prices Ratio S c a le Ratio Scale 1 9 5 7 -5 9 = 1 0 0 1 9 5 7 -5 9 = 1 0 0 Moderated growth of total spending and output first had little effect on employment growth which did not slow appreciably until early this year. Firms apparendy were reluctant to reduce their work forces until the reduced growth of total spending proved more than temporary. By historical standards, the labor market was very tight early in 1969, with almost 65 per cent of the population of labor force age employed, com pared with an average of 62 per Source: U.S. D epartm ent of Labo r Perce n ta g e s a re annual rates of ch ange for p erio d s ind icated . cent in the Fifties and early Latest d ata plotted: Consumer-July; Sixties. Because labor had been W h o le sale Industrial-August unusually scarce, the competitive the same as in the previous year. In contrast, consumer costs of hiring and training personnel probably con prices rose at a 3.8 per cent average rate in 1967 and tributed to employer reluctance to lay off workers in late 1969 and early 1970. Reflecting this lag in ad 1968 and at a 1.2 per cent rate in the 1961-64 period. justment of the work force to changes in the growth Wholesale industrial prices have risen at about a 4 per of total spending, total employment did not slow cent rate since last fall, about the same as in the significantly until March of this year, but has since previous year. declined at a 2 per cent rate. Nevertheless, 64 per Continued rapid inflation in the face of moderated cent of the population of working force age has growth of spending, output and employment is con been employed this summer, a greater proportion than sistent with past experience in the U.S. economy at any time in the Fifties and Sixties prior to 1967. (Table I). Slowing of growth in spending, output and employment has not yet resulted in a clear reduction in the rate of increase of prices. Inflation apparently is no longer accelerat ing, but a decline in the rate of ad vance of prices has not yet been firmly established. The broadest measure of prices, the implicit price deflator for GNP, increased at a 5.4 per cent annual rate in the first half of 1970, about the same as in the last half of 1969. Consumer prices have increased at about a 6 per cent rate since December, about Table 1 Demand, Production and Prices Annual Rates of Change Total Demand Period Real Product Period Prices 11/54 to 1/57 1/57 to 11/58 0.3 - 2 .5 1/58 to IV/59 11/58 to 1/60 8.2 6.4 IV/59 to IV/60 1.9 1/60 to 1/61 0.1 — 1.6 IV/60 to IV/61 1.1 IV/64 to 1/66 10.3 8.2 IV/65 to IV/66 3.5 1/66 to 11/67 6.0 2.8 IV/66 to 11/67 2.4 11/67 to 111/69 8.5 3.8 11/67 to 11/70 4.7 111/69 to 4.1 — 1.1 11/70 7.3% 4.5 % IV/55 to 1/58 3.7% 1.6 Page 3 FEDERAL RESERVE BANK OF ST. LOUIS For example, in 1957-58 the rate of price in crease did not begin to recede until about a year after output growth slowed, and in 1960-61 the lag was about three quarters. The main difference be tween the most recent period and other periods of slowdown is the greater momentum that inflation was allowed to achieve in the 1964-69 period than in previ ous periods. The length of time required to achieve relative price stability, such as during the 1961-64 period, is likely to be considerable. Not only has in flation been allowed to achieve rates higher than in other inflationary periods in the postwar U.S. economy, but the degree of monetary and fiscal restraint has been less than in other such periods. Recent Monetary and Fiscal Actions The course of spending, output and prices in recent quarters appears to have responded normally to mon etary and fiscal actions. Economic developments in the second half of 1970 will depend in large measure on policy actions already taken in the first half of the year. Both monetary and fiscal actions have been more expansive in 1970 than in 1969, though such stimulative policy actions have persisted for only a relatively short period. Fiscal actions —Federal budget actions in the first half of 1970 were dominated by several factors affect ing disposable income: a 6 per cent pay increase for Federal employees, retroactive to the beginning of the year; an increase in the Social Security benefit schedule, also retroactive; and a reduction in the tax surcharge on January 1. Reflecting these actions, the high-employment budget moved from a surplus of about $11 billion in the year ending in first quarter Page 4 SEPTEMBER, 1970 1970 to a $3.6 billion rate in the second quarter of 1970. This reduced rate of surplus was in marked contrast to the $10 billion average annual rate of deficit in the high-employment budget in 1967 and 1968. The national income accounts budget moved from a $9 billion surplus in 1969 to a $14 billion annual rate of deficit in the second quarter of this year. This budget, in contrast with the high-employment budget, is influenced by variations in growth of total spending in the economy, as well as changes in Federal ex penditures and tax rates. Consequently, the national income accounts budget is misleading as a measure of fiscal actions. Reduced growth of total spending, and thus of income and profits, contributed to a drop in Federal tax receipts at a $5 billion annual rate from fourth quarter 1969 to second quarter 1970. Monetary actions —Monetary actions became rela tively expansionary in early 1970. The money stock rose at a 5 per cent annual rate from the three months ending February 1970 to the three months ending in August, after changing little in the second half of 1969. Acceleration of money stock growth reflected more rapid expansion of Federal Reserve credit and bank reserves since February. Federal Reserve credit grew at a 7 per cent annual rate from February to August, compared with a 3 per cent increase in the previous year. M o n e y Stock ,-,- R atio S c a le B illio n s o f D ollars 230 is of Dollars 1962 R atio S c a le B illio n s o f D o lla rs 230 Monthly Averages of Dally Figures Seasonally Adjusted Weekly Averages of Daily Figures Seasonally Adjusted Billions of Dollars 1964 1965 1966 1967 il rates of change for periods indicated. Latest data plotted: August 1968 » t 1969 1970 FEDERAL RESERVE BANK OF ST. LOUIS SEPTEMBER, 1970 Money Market Rates B a n k C redit* Ratio Seal* Ratio Scale A ll C o m m e rcial B anks Ratio S ca le Billions of D ollars Ratio S ca le Billions o f Dollars 500 Monthly Averoges ol Daily Figures 500 r A/V v , Regula ion Q Maximu mRated r -------i. 7 J / 'A / < / r** j v -057. 300 + 89%^ 7 1 V '‘6.4 1 *297. 200 126%^-'- / 80 1962 i I I 1963 I s s a -S Q 6 Jan.'62 ... 1962 1963 1964 1965 -957. -— ^ 90 3-Mor Ih Treasury Bills 13°j8 - +297. 70 J +1927. +89% +2197. Inve tments 100 '— + 69%^- Loans V sTO 4 284.1 *1237. 200 *4\ ; / \ / n 4U 9 Total *167. 4- to 6- lonth Prin e Comme rcial Pape ♦9 87. 27. 400 1 1966 100 1 1967t ... " tl t ....... 1968 1969 ° £ 3 -1 t i_ 1970 90 80 70 ‘ Data are estimated by the Federal Reserve Bank o( St. Louis. Percentages are annual rates oi change (or periods indicated. Latest data plotted: August estimated 1964 1965 1966 1967 1968 1969 1970 [l_Rate on deposits in am ounts of $100,000 or more maturing in 90-179 days. Latest da ta plotted: August More rapid expansion of Federal Reserve credit and the money stock since early 1970 has been accom panied by a sharp acceleration in the growth of time deposits, reflecting primarily a decline in market in terest rates and changes in Regulation Q. This regula tion has been changed twice this year (January 21 and June 24), each time in the direction of increasing the maximum rates which banks are permitted to offer on time deposits. As a result, banks became more competitive in bidding for funds, and time deposits at commercial banks have increased sharply since Janu ary. Total time deposits increased at an 11.5 per cent annual rate from January to June. Since June, these deposits have increased yet more rapidly, at a 37 per cent annual rate, reflecting the suspension of ceilings on rates paid for large certificates of deposit maturing in 30 to 89 days. Over half of the increase in time deposits since June has been in large certificates of deposit. Since January, these CD’s outstanding at commercial banks have increased by $9 billion, off setting a large part of the $13 billion decline from December 1968 to January 1970. The large inflow of time deposits has been accom panied by an increase in bank credit at a 10 per cent annual rate since February. Over the previous year, the level of bank credit remained essentially un changed. The recent upsurge of bank credit reflects partly an increase in Federal Reserve credit, and to that extent represents an increase in total credit in the economy. But the bulk of the increase in bank credit reflects rather a reintermediation of funds previously flowing through nonbank credit channels. From late 1968 to early 1970, banks were not able to compete effectively for funds with open market interest rates which were generally above the Regulation Q ceilings. With the increase in ceilings in 1970 and the general decline in short-term market interest rates, banks have been able to bid funds away from these other markets. The yield on Treasury bills fell from a 7.87 per cent average in January to 6.41 per cent in August. Long term corporate rates, however, rose slighdy on bal ance; the rate on corporate Aaa bonds rose from 7.91 per cent in January to 8.13 per cent in August. Projections of Total Spending, Prices and Output Monetary and fiscal actions of recent months will continue to affect total spending in the second half of 1970, but a more complete assessment of forthcoming developments requires assumptions about the future course of monetary and fiscal actions. The pattern of fiscal actions for the year ending June 30, 1971 can be assessed by looking at Federal budget plans for that period. Current estimates in dicate Federal spending in fiscal 1971 will be about 5 per cent higher than last year. The large increase in expenditures in the second quarter of calendar 1970 was associated with the Government pay increase and rise in Social Security benefits, and is not considered indicative of a trend. To provide projections of total spending, prices, and output beyond mid-1971, Fed eral expenditures were assumed to grow at a 6 per cent annual rate after mid-1971. Page 5 FEDERAL RESERVE BANK OF ST. LOUIS SEPTEMBER, 1970 Table II SIMULATION OF THE EFFECTS OF MONETARY AND FISCAL ACTIONS* Prrii 1970 Onto r\( Chnnnt* in Money Stock 1 II 1971 III 1 IV Actual II 1972 ill IV 1 II Projections 3 Per Cent Rate of Change in: Nominal GNP Real G N P G N P Price Deflator 3.3% — 2.9 6.6 4.9% 0.6 4.3 5.5% 0.6 4.8 4.3% — 0.3 4.6 3.1% — 1.2 4.3 2.9% — 1.1 4.0 6.5% 2.7 3.8 5.3 % 1.8 3.5 4 .9 % 1.8 3.1 4.6% 1.8 2.7 Unemployment Rate 4.2 4.8 5.2 5.5 5.8 6.2 6.6 6.7 6.9 7.1 Corporate Aaa Rate 7.9 8.1 7.7 7.7 7.7 7.7 7.6 7.5 7.3 7.0 Commercial Paper Rate 8.6 8.2 7.5 7.2 6.7 6.2 5.8 5.3 4.8 4.3 3.3 — 2.9 6.6 4.9 0.6 4.3 6.0 1.1 4.9 5.6 0.9 4.7 5.0 0.5 4.5 5.2 0.9 4.2 8.8 4.6 4.1 7.5 3.5 3.9 7.2 3.4 3.7 6.8 3.3 3.5 6.3 5 Per Cent Rate of Change in: Nominal G N P Real GNP G N P Price Deflator Unemployment Rate 4.2 4.8 5.2 5.4 5.7 6.0 6.2 6.2 6.2 Corporate Aaa Rate 7.9 8.1 7.6 7.7 7.7 7.7 7.6 7.5 7.4 7.2 Commercial Paper Rate 8.6 8.2 7.2 7.0 6.7 6.3 6.1 5.8 5.4 5.1 •The simulation is based oni equations estimated through 11/1970. High-employment expenditures are estimated through 11/1971, based on the Federal budget as revised in May 1970. Thereafter, expenditures are projected at a 6 per cent rate. For purposes of projecting total spending, two alter native courses of monetary action are considered. A 5 per cent rate of growth in money for the two-year period starting in second quarter 1970 would be con sidered as a moderately stimulative course of mone tary action. A 3 per cent rate of monetary expansion for the next two years would be a moderately restric tive monetary policy. An optimal rate of monetary expansion must be selected on the basis of a choice among attainable combinations of price and output increases in coming quarters. With the effects of past excessive total de mand still working through the economy, a period of relatively slow advances in output is probably neces sary to complete the adjustment from the accelerating inflation experienced since 1965. If the most likely length of this correction process is not taken into ac count in the determination of stabilization actions, what might appear to be a strong recovery, resulting from stimulative monetary and fiscal actions, would prove only temporary. The adjustment process in volved in unwinding from the 1965-69 inflation can be expected to be long and painful, but it may be even more painful if not consistently pursued. Estimates made by this Bank indicate that if Fed eral expenditures grew at a 5 to 6 per cent annual rate and the money stock were increased at a 5 per cent rate from the quarter ending in June, the growth rate of total spending would most likely be about 5 to 6 per cent in the year ending in mid-1971, and 7 to 8 per cent in the following year (Table II ) 1. Page 6 Real product would grow moderately in the coming year, then increase at about a 3.5 per cent rate in the year ending in mid-1972. Such a policy would most likely result in the rate of increase in prices being down to about 3.5 per cent by mid-1972. A slower rate of growth of money, for example 3 per cent, would probably result in a more moderate 4 per cent advance of total spending in the coming year, about the same as in the first half of 1970. From mid-1971 through mid-1972, total spending would in crease about 5 per cent. Real product, under such a policy, would most likely decline slightiy through mid1971, then increase slowly in the following year. Such total spending growth would probably reduce the rate of inflation to below 3 per cent by mid-1972. Reliability of Projections: A Postscript Economic forecasting based on econometric models continues to be imprecise, and information concern ing the past forecasting performance of a given model should be carefully considered before such a model is used as a guide in policy formulation. To assist the reader in forming his own judgment about the re liability of projections based on research conducted at this Bank, some comparisons are presented to cast light on the model’s forecasting reliability. To examine the reliability of projections, the equa tions of the model were estimated using the experi 'See “A Monetarist Model for Economic Stabilization,” this Review (April 1970), pp. 7-25. FEDERAL RESERVE BANK OF ST. LOUIS SEPTEMBER, 1970 put change corresponded closely to total spending change, and SIMULATION OF THE EFFECTS OF MONETARY AND FISCAL ACTIONS* the model succeeded in produc 1969 1970 ing that similarity of movement. IV II III 1 II 1 For the six-quarter period, total Rate of Change in Nominal G N P output increased at a 0.7 per cent 8.4% 3.9% 3.3% 4.9% Actual 7.5% 7.3% average annual rate; the model 6.4 Model Estimate 6.2 5.7 4.9 4.0 7.6 estimated a 1.4 per cent average Rate of Change in Real GN P 2.2 - 0 .9 - 2 .9 0.6 Actual 2.6 2.7 rate. 0.5 — 0.2 2.2 3.2 1.7 1.3 Model Estimate Forecasting the course of prices, Rate of Change in G NP Price Deflator 5.8 4.9 4.7 6.6 4.3 which has been a prime concern Actual 4.6 4.4 4.1 4.4 4.4 4.2 Model Estimate 4.3 in the U.S. economy in recent Unemployment Rate years, did not satisfy the criterion 3.4 4.2 3.5 3.6 3.6 4.8 Actual of avoiding over- or under-estima 4.1 4.4 3.9 4.7 Model Estimate 3.6 3.7 tion. The model did succeed in Corporate Aaa Rate 6.9 7.5 7.9 8.1 6.7 7.1 Actual that it did not project any decline 7.2 6.9 Model Estimate 6.6 6.8 7.1 7.1 in the rate of price advance before Commercial Paper Rate 1970, though it did not succeed in 6.7 7.5 8.5 8.6 8.6 8.2 Actual projecting the degree to which in 5.8 6.2 6.7 6.9 6.5 5.9 Model Estimate flation accelerated in 1969 and *The simulation is based on equations estimated through IV/1968. Model estimates indicate what the model would have projected when the simulation starts in 1/1969, using actual subsequent early 1970. changes in money and high-employment expenditures. The upward movement of un employment as a percentage of the labor force in 1969 ence of the period from 1953 through 1968. Observa and 1970 was projected by the model, though the in tions since the beginning of 1969 can then be used crease was over-estimated in 1969. The long-term to gauge the performance of the model in simulating interest rate was tracked very closely by the model in the effects of monetary and fiscal actions in 1969 and 1970. For 1969 and 1970, observed monetary and 1969, but the projections fell below the actual rates in fiscal developments were treated as if they were the first half of 1970. Since only two quarters have tran known in advance to the forecaster; thus forecasts spired yielding these discrepancies between actual and of these policy variables were not considered as a predicted, it is probably too early to form a judgment as factor contributing to error in the projections. to whether the criterion of marked and sustained overThe results of this simulation experiment are or underestimation is violated. The short-term interest shown in Table III. Observed values for the varia rate was consistently underestimated by the model, bles are shown along with model estimates. Econo though the pattern of increasing through 1969 and then metric model-building has not yet progressed to the declining in 1970 was projected by the model. point where the models are able to capture accurately Conclusions quarter-to-quarter movements in the variables. What A continuing period of uncomfortable adjustment is important in assessing forecasting reliability is the to reduced inflation may be in prospect for the U.S. degree to which marked and sustained over- or under economy. Given Federal budget plans for fiscal 1971, estimation is avoided. stimulative monetary actions could bolster output and Examination of the forecasts of total spending for employment, but any significant gains in the battle 1969 and the first half of 1970 indicates that the against inflation would be sacrificed. On the other model succeeded in projecting the pattern of total hand, more restrictive actions aimed at reducing the spending growth. The model underestimated the in inflation rate rapidly would probably give rise to fur crease of total spending in the second and third ther cutbacks in output and employment. quarters of 1969, and overestimated the change in Projections based on research at this Bank appear succeeding quarters. For the six quarters as a whole, to be most reliable with respect to growth of total total spending increased at a 5.9 per cent average spending. Projections of the other variables of the annual rate, while the model projected a 5.8 per cent model, though useful and informative, should be ac rate of increase. cepted tentatively. Additional observations will pro Model forecasts of real output change also tracked vide further basis for a judgment about the reliability actual movements quite closely on average. Real out of model projections for these variables. Table III Page 7 Selecting a Monetary Indicator —Evidence from the United States and Other Developed Countries by MICHAEL W. KERAN X HERE IS a long tradition of comradeship between central bankers of different countries. This is due not only to the similarity of professional backgrounds of central bankers but, perhaps more importantly, to the similarity of monetary tools and technical prob lems in the conduct of monetary policy. Most central banks have a common set of mone tary tools:1 (1) the discount rate, or the price at which it loans reserves to the banking system; (2) open market operations and “window guid ance,” or the quantity of direct reserves it provides; and (3) reserve requirements, or the amount of reserves that the banking system is required to hold as idle balances and therefore cannot use for loans and investments. Which of these tools will be dominant depends upon the institutional and financial conditions of each country. In the United States, with its well-developed short-term financial markets, the primary monetary tool is Federal Reserve open market operations. In Germany, where the short-term money market is not eAn earlier version of this paper was presented on June 10, 1970, in Seoul, Korea, on the occasion of the twentieth anniversary of the Bank of Korea. The author gives special thanks to Professors Karl Brunner and Allan Meltzer, and to his colleagues Leonall Andersen, Christopher Babb and Keith Carlson, for helpful comments. 1These monetary tools represent indirect controls of the central bank on the banking system, because they constrain only the total balance sheet of the banking system, and the banks are free to adjust the individual components of their portfolio. The central bank may also have monetary tools which di rectly affect specific sectors of the banking system’s balance sheet. Interest rate ceilings on time deposits constrain a seg ment of the banking system’s liabilities; quantitative limits on the amount of business loans restrict a component of a bank’s assets. The discussion with respect to the indicator question applies to both direct and indirect central bank tools. Digitized for Page FRASER 8 well developed, and where large reserve injections come from balance-of-payments surpluses, the pri mary monetary tool is changes in reserve require ments (Mindestreservepolitik). In Japan, where com mercial banks are in large and continuous debt to the central bank, the primary monetary tool is rationing central bank credit through the discount window (Madoguchi Shido). In Korea, the primary tools are reserve requirements and rationing at the discount window ( Chan-gu Kyu-jai). Once monetary policy is determined and the mone tary tools activated, the next question central bankers face is “are net monetary influences on the economy moving in line with policy?” In a world of uncer tainty, this question can only be answered in the con text of a properly specified indicator of monetary in fluence on the economy. This article will (1) briefly discuss the need for an indicator and the method of testing alternative indi cators; (2) develop the criteria of a good indicator; (3) present statistical evidence regarding which in dicator has given the most consistently correct infor mation for various periods of American history and for recent experience of other developed countries; and (4) consider the general factors which would make one indicator superior to another. The N eed for An Indicator An indicator is defined here to be some readily observable economic time series which can be used to “scale” monetary or fiscal influences on economic activity. If the indicator shows an increase, we want to be able to say with some confidence that monetary or fiscal influences are easier or tighter, depending on what sign the indicator is postulated to have. FEDERAL RESERVE BANK OF ST. LOUIS In a world of perfect knowledge of the financial and economic interrelationships in an economy, one would not need an indicator of monetary influence. Given a particular monetary policy goal, such as to restrict total demand, the policymaker could directly link the manipulation of his monetary tool (open market operations, reserve requirements, or the dis count rate) to a desired change in total demand. This would be possible because, with perfect knowledge of the relationships in the economy, the policymaker would know the exact linkage between his manipula tion of the monetary tool and its consequence with respect to total demand. Unfortunately, we do not have perfect knowledge about the links between monetary tools and financial markets or between financial markets and real mar kets. We know relatively little about the transmission mechanism between central bank actions and the final effect on the economy. This uncertainty is not only due to a lack of statistical data, since it exists in all countries irrespective of whether they have strong or weak statistical gathering services. An example may help illustrate the problem of un certainty in the implementation of monetary policy. Suppose the United States wishes to follow a restric tive monetary policy. To do this the Federal Reserve may raise the discount rate, raise reserve requirements, or sell Government securities on the open market. However, any of these movements in the monetary tools may not by themselves lead to tight mone tary influences on the economy. A rise in the discount rate may not raise the relative price of central bank credit if, because of an increase in the demand for credit, money market interest rates rise by as much as, or more than, the rise in the discount rate. An increase in reserve requirements designed to impound reserves may be offset by an increase in Federal Reserve float, because of a rise in bank transactions. The reserves lost by the banking system through Federal Reserve selling of Government securities may be neutralized by a gold inflow. Some of these neutralizing influences can be ac counted for and offset by the central bank. However, given the current state of knowledge about economic relationships, many other factors which could neu tralize Federal Reserve actions are not known. The central bank needs a summary indicator of net mone tary influences on the economy as a check against whether the manipulation of its monetary tools is achieving the previously established goals. SEPTEMBER, 1970 By observing the movement of the indicator, the central bank should be able to determine whether monetary influences are expansionary, contractionary, or neutral. If the indicator shows monetary influences are expansionary, and policy calls for contraction, then the monetary tools can be manipulated in a more contractionary way. If the monetary indicator is moving in the same direction as that called for by policy, then the monetary tools need not be mani pulated as vigorously as in the previous case. Method of Testing An Indicator The indicator problem can be considered either in the context of a large structural model or in the con text of a single equation, reduced form approach. The single equation approach will be used here.2 The single equation approach to the indicator issue has a number of virtues. First, it includes most of the monetary and fiscal variables that are components of the economic theories developed in most textbooks, and which are used in the estimations of most structural econometric models. Generally, it is these monetary and fiscal variables which are, within the framework of these large models, the dominant factors influenc ing economic activity. Thus, if the monetary and fiscal variables are properly specified, the single equa tion approach will include the generally recognized major factors in economic stabilization. Second, there is a considerable degree of uncertainty, given our lack of knowledge about the economic world, as to the major channels by which these monetary and fiscal variables influence the economy. In conse quence, it is a useful research strategy to consider these issues by employing the single equation ap proach where the transmission mechanism is not 2In the case of a large structural model, a theory is stated about the interaction of decision-making units in the econ omy. Such a theory would, naturally, include information about how monetary and fiscal policy tools affect economic activity. The monetary or fiscal indicator would be implicit in the hypothesized structure of the economy and, by standard theoretical analysis, could be made explicit. Different indi cators could be derived analytically from alternative theories about the structure of the economy. If we are not certain which of the hypothesized economic structures is “true,” then even if we have the optimal indicator for each structural theory, we do not necessarily have the “true” indicator of monetary or fiscal influences. For an example of analytically deriving monetary indicators from a number of structural econometric models, see Richard Zecher, “An Evaluation of Four Econometric Models of the Financial Sector,” Disserta tion Series No. 1, Federal Reserve Bank of Cleveland Eco nomic Papers (January 1970). Page 9 FEDERAL RESERVE BANK OF ST. LOUIS specified.3 Third, this approach is consistent with a wide range of theories (hypotheses) about the struc tural interrelations in the economy. The key to the single equation approach is the proper specification of the monetary and fiscal varia bles. On the fiscal side, there is general consensus that some measure of changes in government spending and tax rates transmits important fiscal influences. On the monetary side, there is a controversy as to the ap propriate measure of monetary influences. Some eco nomic theorists and model-builders use various market interest rates as a measure of monetary influences; others use various monetary aggregates. To help re solve which class of measures provides the better indi cator of monetary influences, a statistical test is employed. For this test, a representative of each class of indicators is selected; a long-term interest rate and the narrowly defined money stock. The test would allow us to assert one of three propositions: (1) the money stock is superior to long-term interest rates as an indi cator; (2) long-term interest rates are superior to the money stock as an indicator; or (3) neither the money stock nor interest rates are clearly superior as an indicator. Criteria for Selecting An Indicator There are no generally accepted criteria of a good indicator wifh the single equation approach. Three criteria are suggested here which are plausible, but not necessarily exhaustive: (1) to be useful as a guide to central bank policy implementation, an in dicator should be responsive to the monetary tools of the central bank; (2) in order to interpret move ments in the indicator as expansionary or contrac tionary, it should have a theoretically unambiguous association (or sign) with total demand; (3) to be of practical use to central bankers, it should have a high degree of statistical association (with the theoretically expected sign) with total demand. If the indicator changes in value today, we want to be able to predict with some degree of confidence what will happen to total demand in the future. How do the money stock and interest rates com pare with the criteria of a good indicator? With respect to the first criterion, the central bank’s ability to sub3This, of course, would only give a first approximation meas urement of impact, which could later be refined when we have greater confidence in the structural models. Indeed, the results of the single equation estimates could help guide structural model-builders in the most fruitful direction. Digitized for Page FRASER 10 SEPTEMBER, 1970 stantially affect interest rates or the money stock is widely accepted among economists. This is based on the general proposition that because a central bank has, in effect, unlimited financial resources, it can determine the value of any financial variable, including interest rates or the money stock ( but not both simul taneously ). There has been a relatively limited amount of empirical work directed to the question of re sponsiveness of monetary indicators to central bank tools, but what has been done supports this general proposition.4 It is also not hard to find theoretical justification for the role of both interest rates and the money stock as an important element in the transmission of cen tral bank actions to the rest of the economy. Both the Keynesian Income-Expenditure Theory and the Modern Quantity Theory of Money place money and interest rates in strategic roles.5 These two theories differ substantially with respect to how money and interest rates operate on the economy, but do not differ on the proposition that both variables are im portant. In the Keynesian theory, the money stock is positively associated and interest rates are negatively associated with economic activity. In the Quantity theory, the money stock is also positively associated with economic activity; however, the interest rate link to economic activity is ambiguous, because the link between money and interest rates is negative in the short run but it could be positive in the long run. Both interest rates and the money stock pass the first two tests of a good indicator, which leaves the third criterion for differentiating between money and interest rates. Which of these two variables has been observed to have the closest statistical association (with the expected sign) with economic activity? 4See A. Burger, An Explanation of the Money Supply Process, Wadsworth Publishing Company (forthcoming); Keran and Babb, “An Explanation of Federal Reserve Behavior (193368),” this Review (July 1969); Allan Meltzer, Controlling Money, this Review (May 1969); John Wood, “A Model of Federal Reserve Behavior,” Staff Economic Study No. 17, Board of Governors of the Federal Reserve System; and Zecher, An Evaluation of Four Econometric Models of the Financial Sector; G. Kaufman, “Indicators of Monetary Policy,” National Banking Review, June 1967. “Until recently, most econometric models along Keynesian lines have ignored the explicit role of money. However, more recent work, specifically the MIT-FRB model, has included monetary aggregates. Keynesian economic theory is compatible with eimer a monetary or interest rate measure of central bank actions. The quantity theory of money also treats interest rates as the strategic price variable which transmits monetary influ ences to the rest of the economy. See Milton Friedman “The Quantity Theory of Money —A Restatement,” in Studies in the Quantity Theory of Money, (Chicago: University of Chicago Press, 1956), pp. 3-21. FEDERAL RESERVE BANK OF ST. LOUIS Statistical Tests of Alternative Indicators A number of recent studies in this Review have measured the relative impact of monetary and fiscal influences on economic activity in the United States and in other developed countries.6 The single equa tion tests were of the following general form: where AY — do “I- cci AM 0C 2 AF -t- e Y is a measure of economic activity ( total demand) M is a measure of monetary influence F is a measure of fiscal influence A is quarterly change The symbol a , stands for the coefficient relating monetary influences to economic activity. The symbol a 2 is the coefficient relating fiscal influences to eco nomic activity. The symbol a 0 represents the coeffi cient for the trend value of all other influences on economic activity. The symbol e represents the error term or nontrend values of all other influences on economic activity. These earlier studies found this single equation ap proach useful, as a first approximation, in measuring monetary and fiscal influences on total demand. This same single equation approach is used here to test alternative monetary indicators. One difference from earlier studies is that alternative monetary indi cators must be estimated in separate equations, be cause they are conceptually measuring different as pects of the same phenomenon.7 All variables are measured as quarterly differences or changes from one quarter to the next. The data are drawn from fifty years of American history (1919/11 to 1969/IV), divided into a total and five sub-periods, and the postwar periods of five other developed countries: Canada, Germany, Japan, South Africa, and the United Kingdom.8 All equations were estimated using the Almon distributed lag technique (see Appendix for discussion). 6The rationale for this approach to empirical estimation has been discussed before and will not be repeated here. The in terested reader is referred to Andersen and Jordan, “Mone tary and Fiscal Actions: A Test of Their Relative Importance in Economic Stabilization,” this Review (November 1968); DeLeeuw and Kalchbrenner, “Comment,” this Review (April 1969), and Keran, “Monetary and Fiscal Influences on Eco nomic Activity —The Historical Evidence,” this Review ( No vember 1969). 7Fiscal indicators are included in the statistical estimations, but are not considered explicitly in the text which is con cerned only with monetary indicators. If the fiscal variables had not been included, the estimated coefficients of the monetary variables could have been biased or their statisti cal significance over- or under-stated. 8Detailed description of data and sources is given in the Appendix. SEPTEMBER, 1970 For each country, and for each period of American history, three tests were performed, and the results are summarized in Tables I, II, and III of the Appendix. The first test consisted of regressing changes in eco nomic activity against changes in Government spend ing, the Government tax rate,9 and the money stock. Government expenditures and tax rates are the indi cators of fiscal influence, and the money stock is the indicator of monetary influence. The second test was identical to the first test, except that changes in long term interest rates were substituted for the money stock as the indicator of monetary influence. In the third test, the level of long-term interest rates was used for the monetary indicator.10 Several interesting observations could be made on the basis of these statistical results. However, just one question will be considered —whether the money stock or interest rates is a more reliable indicator of monetary influence. According to the discussion in the previous section, the monetary variable which is most consistent in predicting future movements in economic activity is a superior indicator. Predictable association of one of a number of independent varia bles with respect to the dependent variables is measured by the “t” statistic. A “t” statistic of 1.96 or larger for a coefficient is considered statistically sig nificant within the conventional 95 per cent confi dence intervals. A “t” statistic of less than 1.96 is not considered statistically significant. The higher the “t” statistic, the greater confidence one has that the es timated coefficient is drawn from the same “universe” as the “true” coefficient. To facilitate comparisons of the “t” values for the monetary “sum” coefficients in the Appendix, they have been grouped into Table I. Of the eleven test periods —six from the United States and five from other countries —only in two 9Total tax receipts are a function of both the level of income and the average tax rate established by the Government. Only the tax rate can be considered a policy variable, because changes in tax receipts due to changes in GNP are not di rectly controllable by Government action. To take account of this consideration, the tax variable in this study is com puted as an average tax rate on all sources of income as follows: where Tx is total receipts and Y is nominal GNP. The change in the tax rate is scaled by the level of (Y) to convert it into a billions of dollars equivalent. 10There are two exceptions in the use of long-term rates, Japan and South Africa. A short-term rate was used for Japan because the long-term rates are subject to informal interest rate ceilings imposed by the government. A short term rate was used for South Africa because no suitable long-term rate was available. Page 11 FEDERAL RESERVE BANK OF ST. LOUIS SEPTEMBER, 1970 Table 1 “ t” VALUES OF ALTERNATIVE MONETARY INDICATORS Expected Sign Estimated Signs United States 1919-69 1919-29 1929-39 1939-46 1947-52 1953-69 Canada 1953-69 Germany 1960-69 Japan 1955-69 South Africa 1955-69 United King dom 1954-69 Monetary Influences Meas ured by Changes tn Level Of Money Interest Interest Stock Rates Rates Positive ( + ) Negative ( — ) Negative ( — ) + 4.08 + 3.22 + 2.48 + 1.49 + 9.35 + 5.38 + 5.62 + 2.82 + 2.76 + .46 — .22 — 3.13 + 1.76 + 2.60 + 1.01 — 2.95 — 1.87 + -21 + -42 - .1 1 - .76 — 3.12 + .17 + 5.19 + 2.44 + 1.02 - 1 .7 0 + 2.58 — 1.01 + 2.22 + 4.61 — .99 + 3.22 of monetary indicators would not have supported the superiority of a monetary aggregate over an interest rate measure. However, such a result is not likely, because most monetary aggregates move in line with the money stock, and most interest rates move in line with the long-term bond rate. In this type of test, it is unnecessary for the magnitudes of the movements to be similar. periods did changes in interest rates (column 2) have a statistically significant negative value (U.S. 1929-39 and Canada). In the other nine periods, the change in the interest rate coefficient was significantly posi tive in one period (U.S. 1947-52), and statistically insignificant in the eight other periods. Clearly, changes in interest rates do not give a systematic or consistent indication of monetary influences on eco nomic activity and thus are not a reliable indicator. A second test of alternative monetary indicators consists of looking at the average quarter-by-quarter pattern of their association with economic activity, in contrast with their total (sum) association with eco nomic activity (Table I). The charts on the next page present the results of such a test for changes in money and changes in interest rates, where each chart can be thought of as representing the pattern of statistically estimated coefficients relating changes in money (the solid line) and changes in interest rates (the dotted line) to changes in eco nomic activity. Because the money stock and interest rates are measured in different dimensions, the esti mated coefficients have been multiplied by the ratio of the standard deviation of the independent and dependent variables, so that the coefficients can be compared directly. When the estimated coefficients are thus modified, they are referred to as Beta coefficients.11 The statistical test was also performed using levels of interest rates and first differences for the other vari ables (column 3). These results are less satisfactory than using changes in interest rates. Of the eleven test periods, only one had a statistically significant and negative coefficient. That result occurred for the United States in 1939-46. Of the other ten cases considered, four are statistically significant but of the wrong sign (positive), and six are statistically insignificant. The pattern of the Beta coefficients for the money variable (AM) is very similar for all periods and countries represented. The coefficients have a con sistently positive value through most of the time pe riods. If there are any negative coefficients on the money variable, they appear in the longest lag time period, usually in excess of t-4. The only exception to this “standard” pattern is the United Kingdom, where there is one virtually zero value of the AM coeffi cient in the middle of a pattern of positive coefficients. The money stock, on the other hand, had a posi tive relationship with economic activity in all eleven periods and was statistically significant in all but one period, World War II (U.S. 1939-46). In spite of the wide diversity of institutions and economic circum stances represented in the different time periods and different countries, changes in the money stock have almost always led to a predictable change in eco nomic activity in the direction consistent with eco nomic theory. The Beta coefficients for changes in interest rates (AR) also have a degree of consistency. However, it is not the kind of consistency which increases pol icymakers’ confidence in interest rates as an indicator. In all but one case, changes in interest rates show an initial positive association with economic activity which only gradually diminishes and becomes a nega tive association after three to five lagged quarters. The interest rate coefficient has the theoretically ex pected negative association with economic activity consistendy only in the case of the United States from With respect to the propositions considered on page 10, the one which is most consistent with the evi dence just presented is (1) the money stock is superior to long-term interest rates as an indicator of monetary influence. It is possible that a different pair Digitized for Page FRASER 12 11The results for the War and immediate Postwar periods for the United States and South Africa are omitted from the chart, because of space limitations. The pattern of the Beta coefficients for the omitted periods is quite similar to that of the included periods. SEPTEMBER, 1970 FEDERAL RESERVE BANK OF ST. LOUIS Beta Coefficients of Alternative M onetary Indicators First Differences United States .6 .6 .4 .4 .4 .2 .2 .2 0 0 0 .2 -.2 .4 -.4 .6 -.6 t+ t-1 CANADA II '19 — ll'29 .6 .6 .6 .4 .4 .2 .2 -/ 0 0 .2 -.2 .4 -.4 .6 -.6 -.2 AR -.4 I t t-1 t-2 t-3 I____ L -.6 J ____ L t-4 t-5 153 — Iir69_ .6 .4 A AR \ .2 AM \ -.2 -.4 ............ -....... ... .... ..... -___________ 1 t+1 t t-1 t-2 t-3 t-4 t-5 Other Developed Countries 0 X V7 \/ 1 1........ 1. -L.. G erm a n y .6 4 .2 0 .2 4 6 .6 .4 .2 0 -.2 -.4 -.6 Note: Beta coefficients are for changes in the money stock (AM} and interest rates (AR)- These Beta coefficients are calculated as the products of the regression coefficients for the respective variables and the ratio of the standard deviation of the independent variables to the standard deviation of the dependent variable (AY) or (AGNP). Lags were selected on the basis of the minimum standard error of the estimate adjusted for degrees of freedom. Page 13 FEDERAL RESERVE BANK OF ST. LOUIS SEPTEMBER, 1970 1929 to 1939. In the Canadian case, interest rates have a small positive influence in the initial quarter, and consistently negative influences in all subsequent quarters. It is not surprising that these two cases are also the only ones where the sum coefficients were statistically significant and negative, as described in Table I, column 2. In a world of perfect knowledge about the financial and economic structure, both the money stock and interest rates would give identical information about monetary influences on the economy.13 The indicator problem arises because there is ignorance at the empirical level about exact specification of the link ages of monetary and other variables in the economy and the time lags associated with them. The evidence In the other six cases in the chart, the pattern of which was considered above suggests that the money the interest rate coefficients is such as to virtually stock has an overwhelmingly more predictable asso wash out any consistent effect on economic activity. ciation with economic activity than interest rates. The early positive influences are matched by the later Knowledge is one of our scarcest resources, and it negative influences. This is also consistent with Table apparently takes less knowledge to properly evaluate I, column 2, where the value of the “t” statistic in the impact of the money stock than the impact of dicated that these same six cases had statistically in interest rates. Conversely, to see the workings of in significant sum coefficients. terest rates it takes more knowledge of the workings The results presented in Table I and the chart are of the economy than we currently have. There are a highly consistent with each other, and provide a number of possible reasons for this state of affairs: strikingly strong case that monetary influences, meas 1. Difference between theoretical and actual ured by changes in the money stock, have a more measures. The range of interest rates which are theo predictable and uniform pattern of effect on economic retically relevant in indicating monetary influence on activity than monetary influences, measured by economic activity is much broader than that available changes in long-term interest rates. in the published interest rate series. The transmission of monetary impulses to the rest of the economy op What do these results imply for the monetary pol erates through changing prices of a wide range of icymaker? If he desires to minimize his errors in pre assets and liabilities, which is equivalent to changes dicting the effects of his actions on the economy, he in their associated interest rates. The value of finan will use the money stock as an indicator of monetary cial assets reflected in the yield on any one type of influence. This selection is not dependent on his ac bond may be too narrow to represent the wide spec ceptance of a “Quantity Theory” view of the trans trum of assets and liabilities represented in the bal mission mechanism. It is equally consistent with a ance sheets of households and firms which transmit Keynesian view of the transmission mechanism which also postulates a positive association of money with < monetary influences. The measured money stock, on the other hand, is a economic activity. Rather, the selection is based on the much more complete enumeration of the liquidity empirical observation that interest rates have proven position of all households and firms. Only commercial to be a misleading indicator in most periods, while bank demand deposits and currency issued by the the money stock has proven to be an accurate in central bank and Government can perform the role dicator in virtually all periods.12 of a medium of exchange. Even other financial in stitutions must hold their working balances as demand Why is Money Superior to Interest Rates deposits in a commercial bank. Therefore, the ob as an Indicator? served money stock series comes closer to a theoreti The empirical results just discussed should not be cal measure than the observed interest rate. interpreted as denying the central role of interest 2. Difference between real and nominal values. rates in transmitting monetary influences to the rest It is generally asserted that it is changes in real of the economy. The large body of theoretical litera interest rates which affect economic activity, but ture on the paramount role of interest rates is not in only changes in nominal interest rates are actually dispute. Most monetarists acknowledge the role of measured and reported. The difference between interest rates in the transmission mechanism. real and nominal interest rates is the result of 12For another study along similar lines see M. Hamburger the change in prices which is expected to occur “Indicators of Monetary Policy: The Arguments and the between now and the maturity of the financial in Evidence,” The American Economic Review, May 1970, and M. Willms, “An Evaluation of Monetary Indicators in Ger many,” in K. Brunner, ed., Proceedings of the First European Conference on Monetary Theory ana Monetary Policy, forthcoming. Digitized for Page FRASER 14 13See Karl Brunner and Allan Meltzer “The Nature of the Policy Problem” in Targets and Indicators of Monetary Policy, (San Francisco: Chandler Publishing Co., 1969). SEPTEMBER, 1970 FEDERAL RESERVE BANK OF ST. LOUIS strument.14 Measurement of these expected price changes is both conceptually and empirically a diffi cult process, subject to many errors. If nominal in terest rates are rising because of expected inflation in the future, the real interest rate may actually be un changed or falling. Thus, to evaluate monetary ac tions in a period of inflation or deflation by looking at nominal interest rates may be misleading. This problem does not arise with measurements of the money stock, because in its most generally used form it is nominal values of money which influence nominal values of economic activity. 3. Confusion between supply and demand. Even if one could measure real interest rates, the change in interest rates may be due to a change in the de mand for credit rather than to a change in the supply of credit, engineered by the central bank. In a period of economic expansion, the demand for credit in creases, which pushes interest rates up. In a period of economic decline, there is typically a reduction in the demand for credit, which pushes interest rates down. Such movements in interest rates are not the result of central bank action but of feedback from the rest of the economy. Yet, if interest rates are used as an indicator of monetary influence, it would ap pear as if the central bank has taken countercyclical actions when, in fact, it may have taken no action at all. This problem is not as serious when the money stock is vised as an indicator. Most studies on the determinants of the supply of money lead to the con clusion that central bank operations dominate the money stock and tend to offset demand-induced changes in the money stock.15 In other words, the behavior of the public, acting on the demand side of the market, does not bias the money stock as an in dicator of monetary influence as much as it does in terest rates. 4. Greater stability in the demand for money than in the demand for commodities. If our current state of knowledge allows us to more accurately predict the demand for money than the demand for goods and services, then the money stock will be more closely related to economic activity than interest rates in any statistical analysis.18 This point can be illustrated in a standard Keynesian LM-IS framework, as in Figure I. Figure I The D e m a n d for M o n e y (LM0) a n d C o m m o d itie s (IS-IS') interest rate, r is the nominal interest rate, and pe is the rate of change in expected prices of goods and services over the life of the financial assets. If price expectations are formed very slowly, then the gap between real and nominal interest rates will be small. Until quite recently, this was the gen erally held position among economists. However, Yohe and Kamosky (this Review December 1969) have developed new evidence which indicates that price expectations are formed quite rapidly, thereby creating a substantial gap between real and nominal interest rates even during relatively short periods of inflation and deflation. If the demand for money is well specified, then the locus of points representing the LM curve can be described by a line ( LM0). If the demand for com modities, however, has a large random (stochastic) element, the IS curve can be described only as a band, the dimensions of which are IS —IS'. In this circumstance, the link between any interest rate R0 and income would be represented by the gap Y0 - Yx. On the other hand, the relationship between any given money stock M0 (which is implied by a given LM curve) and income would be represented by the band Y2 - Y3. Because the spread between Y0 and Y! is greater than the spread between Y2 and Y3, the degree of statistical association between changes in R and changes in Y would be less than between changes in M and changes in Y. 1BSee John Wood, “A Model of Federal Reserve Behavior,” Staff Economic Studies No. 17, Board of Governors, 1968. Also, “An Explanation of Federal Reserve Actions,” this Review, July 1969. “Reply to Comments on the St. Louis Position,’ August 1969, and “Comment,” May 1970. 16The rationale for the greater stability for the demand for money than the demand for commodities is presented by William Poole in “Optimal Choice of Monetary^ Policy In struments in a Simple Stochastic Macro-Model.” Quarterly Journal of Economics, May 1970. 14The difference between real and nominal interest rates can be presented as follows: r* = r — pe, where r* is the real Page 15 FEDERAL RESERVE BANK OF ST. LOUIS 5. Government controls. Governments historically have imposed ceilings on interest rates. When such ceilings exist, interest rates cannot be used simul taneously as an indicator of monetary influence on the economy. An indicator that is not allowed to move with changes in market forces can give mis leading and wrong information. This point applies only to the use of legal authority to control interest rates by fiat. The use of standard monetary tools to control interest rates does not, of course, weaken its role as an indicator. Usually when an interest rate is used to measure monetary influence, it is selected from among those which are not under direct government constraints. For example, the corporate Aaa bond rate, which is used in the statistical tests on the United States, has always been free of legal constraints. However, when the government controls one interest rate, like that which banks can pay on time deposits (Regulation Q), credit flows away from banks and into other fi nancial markets in which the rates are uncontrolled. These distortions in credit flows could distort the in terest rate quoted in those markets as an indicator of monetary influence. It is, of course, possible that interest rate controls on time deposits could distort the money stock, es pecially when money is defined to include time de posits. However, the money stock definition used here includes only demand deposits and currency, and therefore the distorting effects of controls are apt to be minimized. Conclusion The main point of this article is that selection of an indicator of central bank actions need not be made only on theoretical grounds. If we are not certain of Digitized for Page FRASER 16 SEPTEMBER, 1970 the theoretical structure of the economy, the selection of the indicator can also be made on empirical grounds. We have observed, in a wide range of his torical and institutional contexts, that the money stock is a reliable and predictable indicator of monetary influence, and that interest rates are not. The reasons for this difference in results stem largely from the fact that it apparendy takes more knowledge about the workings of the economic system to evaluate the im pact of interest rates than to evaluate the impact of the money stock. There are at least five possible fac tors responsible for this: (1) the reported interest rates do not cover all the financial markets which transmit monetary influences to the rest of the eco nomy; (2) the data reported are of nominal interest rates, while it is real interest rates which affect eco nomic activity; (3) it is difficult to distinguish changes in interest rates which are induced by de mand pressures of the public from those caused by central bank actions; (4) uncertainty about the de mand for commodities relative to the demand for money increases the uncertainty of the relation of interest rates to economic activity; (5) Government interest rate ceilings in some markets induce arbitrage flows which distort the movements of interest rates in other markets. We can summarize these factors by saying that they represent the greater degree of knowledge we must have about the economic system to make inter est rates a successful indicator of monetary influence on economic activity. This does not imply that the money stock is not subject to some of the same uncer tainties as those attached to interest rates. Rather, the statistical results suggest that the uncertainties are less with the money stock than with interest rates. This article is available as Reprint No. 59 The statistical appendix begins on the next page. STATISTICAL APPENDIX The following tables summarize the regression results which are the basis for the assertions in the text. The only aspect of these results which is discussed in the text are the “t” values on the alternative monetary variables. There are other implications which can be drawn from these re sults. Specifically, the fiscal variables play a stronger role and have greater statistical significance when an inter est rate, rather than the money stock, is used as the monetary variable. This result is not surprising. Omitting money from the equation allows the Government deficit to be financed by increases in the money stock rather than just through increases in debt sales to the public. Thus, following the analysis of Fand (this Revieiv, January 1970), one would expect a stronger measured fiscal influence when interest rates are the monetary variable, and a weak fiscal influence when the money stock is the mone tary variable. This point and others will be developed in a future article. The Almon lag technique was used to estimate all equations presented below. By constraining the distribu tion of coefficients to fit a polynomial curve of n degree, it is designed to avoid the bias in estimating distrib uted lag coefficients which may arise from multicollinearity in the lag values of the independent variables. The theoretical justification for this pro cedure is that the Almon constrained estimate is superior to the uncon strained estimate, because it will cre ate a distribution of coefficients which more closely approximates the dis tribution derived from a sample of infinite size. In order to minimize the severity of the Almon constraint, the maximum degree of the polynomial was used in each case. The maximum degree is equal to the number of lags Table 1 UNITED STATES Monetary and Fiscal Influences on Economic Activity (Quarterly First Differences — Billions of Dollars) A y t = oco + oci A m + a * A e + 0 C3 A tx tt Period Lags* Constant Term Monetary Influence ao Otl (Sum) Fiscal Influences a2 (Sum) RV D-W a3 (Sum) 11/1919 - 111/1969 t— 8 1.24 (1.18) 4.55 (4.08) ( - .09 .40) - 3 .7 3 ( — 2.61 ) .39 1.84** 11/1919 - 11/1929 t— 2 111/1929 - 11/1939 t— 4 Ml/1939 - IV/1946 t— 8 1/1947 - IV /1952 t— 8 1/1953 - 111/1969 t— 4 — .41 ! — -66) - 1 .1 7 (- 1 .3 4 ) - 5 .1 0 ( - -95) — 3.57 ( - 1 .0 5 ) 1.24 (-63) 5.61 (3.22) 4.10 (2.48) 4.40 (1.49) 27.74 (9.35) 9.25 (5.38) .06 (.11) -2.91 ( - -97) — 6.70 (- 2 .8 7 ) - 4 .6 0 (- 6 .6 6 ) - .87 1 ( - 1 .0 5 ) - 7 .5 8 ( — 3.36) 1.1 1 (.49) 10.97 (1.64) 9.43 (3.33) - .18 ( - -15) .52 1.90 .35 1.47 .85 2.64 .85 3.25** .49 1.61 A y t = - cxo + (Xi A r + a 2 A e + 0C3 A tx tt Period Lags* Constant Term Monetary Influence ao ai (Sum) Fiscal Influences a2 (Sum ) R2 / D-W as (Sum) 11/1919 ■111/1969 t— 5 3.86 (3.46) 18.03 (46) .97 (1.95) — .54 ( — .45) .26 1.86** 11/1919 - 11/1929 t— 4 111/1929 - 11/1939 t— 3 111/1939 - IV/1946 t— 6 1/1947 - IV/1952 t— 7 1/1953 - 111/1969 t—4 .34 (•54) - 2 .4 4 (- 2 .0 2 ) 12.94 (3.19) 6.27 (3.91) 6.60 (2.65) — 1.78 ( - -22) - 3 4 .9 0 (- 3 .1 3 ) 249.58 (1.76) 141.67 (2.60) 18.32 (1.01) 1.19 (1.24) 3.40 (1-53) 2.70 (3.70) 2.95 (2.13) .68 (.38) - 9 .2 5 ( - 2 .3 2 ) — 2.95 ( - 1 .4 0 ) -10.25 ( - 4 .1 8 ) - 6 .8 2 ( - 2 .6 9 ) 1.1 1 (.47) .66 2.19 .34 1.83** .76 1.77** .61 3.03** .44 1.95** Note: Regression coefficients are the top figures; their “t ” statistics appear below each f ft coefficient, enclosed by parentheses. R- is the per cent of variation in the dependent variable which is explained by variations in the independent variables. D-W is the Durbin-Watson statistic. A Y is measured as a proxy for economic activity because quarterly GNP data are not available before 1947. This proxy is equal to the Industrial Production Index times the Consumer Price Index, times GNP in the base year of these indexes. The ta x variable is computed as follows: A tx = A * ** Y Lags selected on the basis of minimum standard error adjusted for degrees of freedom. A transform ation of this equation was made to eliminate possible bias in reported “t ” values. For details, see text of this Appendix. Page 17 FEDERAL RESERVE BANK OF ST. LOUIS SEPTEMBER, 1970 term £t equals [p • £t-i + Hi ] with (it normally distributed. In the cases where the D-W statistic is significantly greater or less than 2, autocorrelation in the error term is indicated, and the “t” values of the coefficients will be biased. In such cases, unbiased estimates of the “t” values can be obtained by transforming the original equation, plus one of the independent variables up to five lags. If there are n lags, t + 1 and t — n — 1 are both constrained to equal zero.1 The regressions were also run without con straining the beginning and ending values to zero, and the results are virtually identical. The Durbin-Watson (D -W ) statistic is constructed to equal 2 (1 —p), where it is hypothesized that the error oc0 + 2 0Ci Xt,i, i=l Yt iThis follows the convention established by Shirley Almon, “The Distributed Lag Between Capital Appropriations and Expenditures,” Econometrica, January 1965. into (Yt - pYt-i) = oco (1 — p) + 2 oci(Xt,i - pXt-i,i), i=l Table II OTHER DEVELOPED COUNTRIES Monetary and Fiscal Influences on Economic Activity (Quarterly First Differences — Billions of National Currency) Agnp = ao + oci Am + a 2 Ae + 0C 3 Atxtt Country and Time Period Lags* Constant Term Monetary Influence Fiscal Influences ao CXI (Sum) a2 (Sum) R2/ D-' a3 (Sum) Canada 11/1953 - 11/1969 t— 6 .27 (1.77) 6.11 (5.62) — 2.62 (- 1 .6 0 ) 1.43 (.96) Germany 111/1960 - 11/1969 t— 6 — 7.13 (- .8 5 ) 10.58 (2.82) 2.60 (.7 2) -2 .4 0 (.00) .66 2.01 * Japan 11/1955 - 11/1969 t— 3 — .04 ( - .35) 2.81 (2.76) 1.73 (2.61) 3.23 (.46) .74 1.91 * South Africa1 11/1955 - 1/1969 t— 6 .05 (2.03) 3.02 (2.58) .91 (0.76) - 2 .4 0 ( - 2 .4 4 ) .46 1.86 United Kingdom1 111/1954 - 11/1969 t— 6 .15 (2.11) 3.40 (4.61) — .89 ( — 1-38) .35 (•65) .52 2.17 .46 2.00 Agnp = ao + a i Ar + cc2 Ae + 0C3 Atxtt Country and T im e P e r io d Lags* R2/ Constant Monetary F is c a l Te rm In f lu e n c e In f lu e n c e s ao ai (Sum) a2 (Sum) D-W as (Sum ) Canada 11/1953 - 11/1969 t- 9 .46 (2.49) — 6.45 ( — 2.95) 7.83 (3.90) 4.08 (1.77) .35 2.15 Germany 111/1960 - 11/1969 t— 6 11.92 (2.90) - 2 5 .5 5 (- 1 .8 7 ) — 3.14 ( - .79) - 9 .3 5 ( - -85) .37 2.43 Japan 11/1957 - 11/1969 t— 5 — .12 (41) .09 (.21) 5.07 (4.79) 2.60 (.47) .85 2.53 South Africa1 11/1955 - 1/1969 t— 5 .06 (1.93) - .15 (- 1 .0 1 ) 2.69 (2.64) -2.41 (- 1 .9 4 ) .48 2.17 United Kingdom1 111/1954 - 11/1969 t— 7 .37 (4.22) (- .66 (.92) .07 (.10) .27 1.90 .63 .99) N o te : Regression coefficients are the top figures; their “t ” statistics appear below each tf * ** 1 coefficient, enclosed by parentheses. R 2 is the per cent of variation in the dependent variable which is explained by variations in the independent variables. D-W is the Durbin-Watson statistic. Tx The tax variable is computed as follows: A tx = A Lags selected on the basis of minimum standard error adjusted for degrees of freedom. A transform ation of this equation was made to eliminate possible bias in reported “t ” values. For details, see tex t of this Appendix. A proxy for economic activity is used for the United Kingdom because GNP does not give statistically significant results. The proxy is equal to the Industrial Production Index, times the Consumer Price Index, times GNP in the base year of these indexes. Gross Domestic Product (GDP) was used for South Africa. Page 18 and using the “t” values obtained from the latter form. Those cases where the “t” values of the transformed equa tion were used are starred (* *) next to their respective D-W statistics. In these cases the results reported from the transformed equation are the “t” and D-W values. The values of the coefficients and R2 are from the origi nal equation.2 Data Sources Canada — Gross National Product: Canadian Statistical R eview , Domin ion Bureau of Statistics; Money Stock: International Financial Statistics, IM F; Interest Rates ( Long-term Govern ment Bond Yield): International F i nancial Statistics, IM F; Government Receipts and Expenditures: Canadian Statistical R eview , Dominion Bureau of Statistics. Germany — Gross National Product: International Financial Statistics, IMF; Money Stock: International Financial Statistics; IM F; Interest Rates (Mort gage Bond Yield): International F i nancial Statistics, IM F; Government Receipts and Expenditures (Federal finance on a cash basis): M onthly R e port o f th e D eutsche Bundesbank. Japan — Gross National Product: Annual R eport on N ational In com e Statistics, Economic Planning Agency of Japan and N ihon K eizai Shimbun; Money Stock: E con om ic Statistics Monthly, Bank of Japan; Interest Rates (Bank Lending R ate); Main E con om ic Indicators, OECD; Government Re ceipts (Tax and Stamp Revenue) and Government Expenditures (Treasury Cash Payments): Basic D ata fo r E c o nom ic Analysis 1964, 1968, 1969, and E conom ic Statistics Monthly, Statistics Department of the Bank of Japan. 2See Arthur S. Goldberger, Econometric Theory, (New York: John Wiley and Sons, 1964) pp. 236-8 for further discussion. FEDERAL RESERVE BANK OF ST. LOUIS SEPTEMBER, 1970 Table III UNITED STATES Monetary and Fiscal Influences on Economic Activity (Quarterly First Differences — Billions of Dollars)1 A Y t ; = oto + oci R + a.-> A e + 0 C3 A tx tt Lags* Period Constant Term Monetary Influence cx0 Oil (Sum) Fiscal Influences a-z (Sum ) 11/1919 - 111/1969 t— 5 2.37 (.53) .38 (.42) .96 (1.81) 11/1919 - 11/1929 t— 4 111/1929 - M/1939 t— 4 111/1939 - IV / 1946 t— 6 1/1947 - IV/1952 t— 8 1/1953 - 111/1969 t— 6 1.13 (.15) 3.02 (•40) 128.62 (3.41) - 3 0 4 .3 8 ( - .19) — 28.66 ( - 4 .1 3 ) — .16 ( - - ID - 1 .3 7 ( - .76) - 4 3 .6 5 (- 3 .1 2 ) 115.10 (.17) 9.74 (5.19) 1.17 ( 82) 4.41 (1.44) 3.86 (6.06) 2.78 (.77) — 1.74 (- 1 7 5 ) R2/ D-W 0t3 (Sum) — .54 ( - -50) — 10.04 (- 2 .6 3 ) — 3.55 (- 1 .3 5 ) - 1 2 .0 7 (- 5 .9 9 ) — 9.25 (- 1 .5 3 ) — 1.02 ( - -70) South Africa — Gross Domestic Product: South A frica R eview , Gov ernment of South Africa; Money Stock: International Financial Statistics, IMF; Interest Rates (Treasury Bill Rate) International Financial Statistics, IM F; Government Receipts and Expendi tures: International Financial Statis tics, IMF. .26 1.86** .65 2.16 .31 1.84* * .84 2.21 .59 2.78** .63 1.97 OTHER DEVELOPED COUNTRIES United Kingdom — Industrial Pro duction Index and Consumer Price Index, 1963 = 100: Main E con om ic Indicators, OECD; Money Stock: In ternational Financial Statistics, IMF; Interest Rates (Long-term Govern ment Bond Yield): International F i nancial Statistics, IM F; Government Receipts and Expenditures: Interna tional Financial Statistics, IMF. (Quarterly First Differences — -Billions of National Currency)1 A g n p = oto + ocj R + Country and Time Period Lags* A e + 0(3 A tx tt Constant Term Monetary Influence ao ai (Sum) Fiscal Influences 0(2 (Sum) R2/ D-W CX3 (Sum) Canada 11/1953 - 11/1969 t- 2 — .69 (- 1 .3 4 ) .28 (2.44) 1.20 (1.14) .92 (1.28) .30 1.68 Germany 111/1960 - 11/1969 t— 7 — 4.16 ( - .24) 2.66 (1.02) — 4.03 ( - .99) — 2.83 ( - .24) .34 2.50 Japan 11/1957 - 11/1969 t— 5 2.96 (1.60) — .37 (- 1 - 7 0 ) 4.20 (3.53) 3.62 (-61) South Africa 11/1955 - 1/1969 t— 6 — .09 (- 1 .3 4 ) .04 (2.22) 3.20 (4.15) — 3.10 (- 2 .6 9 ) United Kingdom^ 111/1954 - 11/1969 t— 8 — 1.03 (- 2 .3 1 ) .30 (3.22) - 1 .4 9 (- 1 .5 3 ) .84 2.51 ** .47 1.93 .86 .40 2.38 (1.11) N o te : Regression coefficients are the top figures; their “t ” statistics appear below each coefficient, enclosed by parentheses. R 2 is the per cent of variation in the dependent variable which is explained by variations in the independent variables. D-W is the Durbin-Watson statistic. t A Y is measured as a proxy for economic activity because quarterly GNP data for the U .S. are not available before 1947, and because quarterly GNP data for the U.K. do not give statistically significant results. The proxy is equal to the Indus trial Production Index, times the Consumer Price Index, times GNP in the base year of the indexes. Gross Domestic Product (GDP) was used for South A frica. ft The tax variable is computed as follows: Tx A tx = A ) *Y ♦ ** 1 Lags selected on the basis of minimum standard error adjusted for degrees of freedom. A transform ation of this equation was made to eliminate possible bias in reported “t ” values. For details, see text of this Appendix. Levels are used for interest rate data. United States — Industrial Produc tion Index, 1957-59 = 100: Board of Governors of the Federal Reserve Sys tem; Consumer Price Index, 1957-59 = 100: United States Department of La bor; Money Stock: Board of Governors of the Federal Reserve System and the Federal Reserve Bank of St. Louis; Government Receipts and Expendi tures: Daily Treasury Statem ent, Office of the Secretary of the Treasury of the United States and Federal Reserve Bank of St. Louis. In each case, the seasonally adjusted series for gross national product, gross domestic product, money stock, gov ernment expenditures and government receipts were used. The interest rate series are not seasonally adjusted. The seasonally adjusted industrial production index and the unadjusted consumer price index were used in the construction of the proxy variable (Y) which was used for the United States and the United Kingdom. Page 19 SUBSCRIPTIONS to this bank’s R e v ie w are available to the public without charge, including bulk mailings to banks, business organizations, educational institutions, and others. For information write: Research Department, Federal Reserve Bank of St. Louis, P. O. Box 442, St. Louis, Missouri 63166.