The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
FEDERAL RESERVE BANK OF ST. LOUIS DECEMBER 1969 1969 — Battle A gainst Inflation ............... 2 Selective Credit — N o Substitute for M onetary Restraint................................. 13 Interest Rates and Price Level Changes, 1952-69 18 Federal Reserve System Actions During 1969 ........................................ 39 Review Index — 1969 ............................... 40 1969—Battle Against Inflation by NORMAN N. BOWSHER ] \ INETEEN SIXTY-NINE has been the fifth year of intensifying inflation. Overall prices, after remain ing fairly stable in the early Sixties, rose 1.7 per cent during 1965, 3.5 per cent a year in 1966 and 1967, 4.0 per cent during 1968, and an estimated 5 per cent in 1969. Not since W orld War II has the American economy experienced such a sustained price upsurge. General Price Index * R a tio S c a le R a tio S c a le 1958=100 130 1958=100 130 of experts are neither complacent about inflation, nor so pessimistic about its cure. Although stopping infla tion is generally believed to involve hardships, it is felt that the necessary transition costs of following proper public policies would not be great compared with the inequities and inefficiencies resulting from continued reductions in the purchasing power of the dollar. This article traces the course of the inflation of the past five years. It sets forth some of the chief causes and effects, reviews the actions taken to resist inflation, presents an analysis of economic develop ments during 1969, and provides some observations on the outlook. Balanced Economic Expansion 1961 through 1964 1961 1962 1963 1964 1965 1966 1967 1968 1969 • A * used in No tio nal Income Accounts Source: U.S. Deportm ent of Commerce Percentages are a n nu a l rates of chan ge between periods indicated. They are presented to a id in com paring most recent developm ents with pa st "tren d s.” late st d ata plotted: 3rd quarter Inflation has been the nation’s most serious domestic economic problem in recent years. The extent of con cern over this problem has been evident from the numerous tough decisions the public authorities have made in an attempt to moderate it. Tax rates were raised, growth in Government spending was reduced, and monetary growth was restrained. Despite these actions there is as yet no firm evidence that the rise in prices has decelerated, or that inflationary expecta tions have moderated. Some believe that inflation has now become inevi table, and that the country should learn to live with it. They feel that inflation has some desirable features, and that even if undesirable on balance, the nation apparently cannot stop it, or the costs of doing so are likely to exceed the benefits. At the other extreme, some feel that inflation is intolerable, and that it must be stopped even though the necessary cost is likely to be a severe and prolonged recession. The majority Page 2 — Following the recession of 1960, the country ex perienced four years of pronounced economic expan sion with little inflationary pressure. Real output of goods and services increased at a 5.4 per cent annual rate from early 1961 to late 1964. Except for a pause in late 1962, this was a period of steady economic expansion. In the early Sixties real growth was faster than the estimated 4 per cent rate of growth in productive potential. As a result, unemployment was reduced from about 7 per cent of the labor force in early 1961 to less than 5 per cent in late 1964, and manufactur ing plant utilization rose from 75 per cent to 86 per cent of normal capacity. These gains were accom plished in an orderly fashion without great frictions, shortages, or imbalances, while average prices re mained relatively steady. Overall prices, measured by the GNP deflator, rose at a 1.3 per cent annual rate. However, because of the difficulty of fully taking account of changes in discounts granted and quality improvements, the index probably overstated the actual price increase. The economy demonstrated a great resiliency and ability to expand in the 1961 to 1964 period. Fiscal and monetary management caused no great shocks to the economy, and the free enterprise system responded admirably. The nation’s money stock grew at a 2.7 per cent annual rate from mid-1960 to mid- F E D E R A L R E S E R V E B A N K O F ST. LOUI S DECEMBER 1 9 6 9 Excessive Spending and Inflation 1965 through 1969 — Since 1964 total spending for goods and services has risen much faster than the country’s potential to supply them. Total outlays rose at an average 8 per cent annual rate from late 1964 to late 1969. With little excess capacity, increases in real output were constrained by the growth in the nation’s capacity to produce. The rise in spending was roughly double the estimated rate of expansion in potential output, given the growth in the labor force, capital equip ment and technology. 1964, faster than the average 2 per cent rate of the previous decade, but slower than the estimated 4 per cent rate of growth in productive capacity. Dur ing the early Sixties policymakers faced two major economic problems which apparently called for op posite courses, probably accounting for a compromise between moderate and relatively steady growth in money. Unused resources were a signal for monetary expansion while an adverse balance of payments and gold outflows called for restraint. Since the demand for money to hold was probably increasing less rap idly than either real production or productive poten tial, it was appropriate that money expand less rapidly than either of these magnitudes in order to avoid inordinate inflation. The influence of the Federal budget on total spend ing, as commonly measured, was moderate from 1960 to early 1964. The growth rate of Government spending and the net surplus or deficit of both the high employment and the national income accounts budgets remained in fairly narrow ranges. Despite the strong and balanced economic expansion without ex cesses during the early Sixties, some policy advisers held theories which indicated that the state of the budget was highly depressing to total spending, pro duction, and employment. After a substantial delay, taxes were cut in early 1964, with the objective of keeping the country moving by eliminating the al leged actual or potential “fiscal drag.” Most sectors of the economy have participated in the rapid increase in spending which began in late 1964. Government purchases of military goods have expanded at an 11 per cent annual rate since 1964, compared with a trend rate of 2 per cent from 1957 to 1964. Other Federal Government expenditures have also increased at an 11 per cent rate since late 1964. Business investment has risen at an 8 per cent rate since late 1964, after increasing at a 5 per cent rate from 1957 to 1964. Personal consumption expendi tures have increased at a 7.6 per cent rate since late 1964, following a 5 per cent rate of increase from 1957 to 1964. With spending on goods and services rising faster than their production, prices were bid up. The rate of inflation appeared mild at first, reflecting inflexibility of some prices in the short run, the moderate amount Demand and Production R a tio S c a le Q u arterly Totals at A n n u al Rates R atio S c a le Q G N P in current dollars. Source: U.S. Department of Comm erce [2 G N P in 1958 dollars Percentages are annual rates of change between periods indicated. They are presented to aid in comparing most recent developments with past "trends.” Latest d ata plotted: 3rd quarter 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 I S of excessive demand, and fuller use of resources. Overall prices went up 1.7 per cent from late 1964 to late 1965. About one-sixth of the rise in total spend ing was reflected in higher prices while five-sixths went for additional goods and services. As time passed, demand-pull intensified, and price increases accelerated. During both 1966 and 1967 prices rose 3.5 per cent, and in 1968 they rose 4 per cent. Only about half the total growth in spending during the 1966 through 1968 period went for addi tional real output; the other half was taken in higher prices. Effective prices may have risen even more than these figures indicate, for when demand becomes ex cessive, discounts and rebates are eliminated, sur charges abound, and there is a tendency to reduce quality standards. In the preparation of price indexes, some of these developments may have been missed, since producers are not likely to disclose their com plete discount policies or a deterioration in product quality. The inflation has been even more severe during 1969. The continued rise in total spending in excess of production, prompted in part by deterioration of the illusion that money has a relatively constant value, has contributed to the greater price increases. Overall prices have gone up about 5 per cent in the past year, even though there has been a little moderation of growth in total spending. As a result, nearly threefourths of the rise in outlays has been used to pay higher prices. Growth in real production declined to about 2 per cent in 1969, or to about half the trend rate. In recent months the rise in prices probably has been at about the same pace as the rise in spending, with little net change in total real output. Causes of Excessive Spending and Inflation In our free enterprise system each spending unit — household, business firm, or governmental unit — de termines whether to spend or save the funds available to it. Hence, one might conclude that the excessive total spending resulted from an unfortunate bunching of expansive individual decisions. Such a conclusion is only a half-truth, providing little insight into the basic forces determining total spending. Consumer spending is closely related to income, and business outlays are generally related to expected profits. Although these sectors account for a major portion of total spending, and changes in them do have considerable affects on total spending, they alone have seldom been the prime motivating forces in bringing about cyclical movements in total spending Page 4 DECEMBER 1 9 6 9 Two forces in the economy which are under the control of public policy and which are believed to have a great influence on private spending decisions are fiscal policy and monetary management. Most studies of economic stabilization have focused on them. Fiscal Actions — The results of decisions by the Fed eral Government which change its spending and tax ing programs are fiscal actions. It has commonly been believed that such changes expand or contract total public and private spending by some multiple. H ow ever, the aggregate influence of the Government on total spending is greatly diminished if the resulting deficits or surpluses are financed by the public out of planned saving rather than accompanied by changes in the money stock.1 Changes in Government activi ties tend to be offset by opposite movements in pri vate spending when the Government finances its deficits with debt paid for by the public out of cur rent planned saving. Among the commonly used measures of Govern ment fiscal actions are the national income accounts budget, the expenditure component of this budget, and the high-employment variation of it. The national income accounts budget summarizes the receipts and expenditures of the Federal Government sector as an integral part of the national income accounts. The high-employment budget is an estimate of the ex penditures and revenues in the Federal sector of the national income accounts at an assumed constant rate of growth of real economic activity (conventionally about 4 per cent unemployment). It attempts to ab stract from the impact of actual economic activity on the realized surplus or deficit. It is widely believed that the inflation since 1964 has been caused by Government fiscal mismanage ment. Forthcoming defense spending was greatly un derestimated, there was lack of restraint on non-de fense outlays, and there was delay in raising taxes. As a result, from 1963 to mid-1968 the Government cut tax rates and increased growth rates in both de fense and nondefense outlays. The high-employment budget, which was at a $13 billion surplus in 1963, declined to about a $14 billion annual rate of deficit in the first half of 1968. The national income accounts budget shifted from a small surplus in 1963 to a $9 billion deficit in the first half of 1968. From 1963 to 1Leonall Andersen and Jerry Jordan, “Monetary and Fiscal Actions: A Test of Their Relative Importance in Economic Stabilization,” this Review, November 1968, pp. 11-24; and Michael Keran, “Monetary and Fiscal Influences on Eco nomic Activity — The Historical Evidence,” this Review, November 1969, pp. 5-24. DECEMBER 1 9 6 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 I S rose at a 7.3 per cent rate until January 1969. The course of money reflected roughly parallel courses of Federal Reserve credit, the monetary base and member bank reserves. Effects of Inflation Inflation is a rise in the general level of prices or, otherwise stated, a decline in the purchasing power of the dollar, and tends to cause a redistribution of wealth and income.2 It affects holders of money adversely, reduces the relative value of bonds, savings accounts, and other dollar-denominated assets, and gives a windfall to debtors. For a savings account drawing 5 per cent per year interest, while prices increase at a 6 per cent annual rate, the saver re ceives a net yield before taxes of minus one per cent, since when the saver receives his principal plus in terest, the funds will buy less than his principal alone would have bought at the time the deposit was made. After-tax income is a yet greater net loss. early 1968 national income accounts expenditures rose at a 10 per cent annual rate, after increasing at a 6 per cent rate from 1957 to 1963. Monetary Actions — Another force under the con trol of public policymakers, which is believed to have a vital influence on private spending decisions, is monetary actions. The Federal Reserve, by determin ing the volume of Federal Reserve credit outstanding and thereby the amount of the monetary base and the reserves of the member banks, can manage the supply of money (demand deposits and currency) in the economy. By supplying more money than the public desires to hold, given current levels of income, wealth, and interest rates, the public’s demand for other financial assets and for real goods and services is stimulated. Businesses and households undertake to exchange excess money balances for assets which will provide more satisfaction. Conversely, by pro viding less money than the public wishes to hold, the central bank can place downward pressure on the rate of spending, since businesses and individuals will reduce outlays in an attempt to build up cash balances in relation to other assets. Monetary actions share responsibility for the over heating of the economy. From the end of 1964 to the end of 1968 the stock of money rose at an average 5 per cent annual rate. By comparison, money grew at about a 2 per cent rate from 1957 to 1964. By sub periods, from late 1964 to the spring of 1966 money rose at a rapid 6 per cent rate, remained on a plateau during the summer, fall and winter of 1966, and then Inflation has different effects on various individuals and businesses, depending on types of assets and liabilities held and sources of income, which, in turn, may have been affected by the extent to which infla tion has been anticipated. When inflation can be an ticipated and provided for, types of asset and liability holdings may be adjusted and the rate of price in crease built into contracts by cost-of-living or other escalators. If both borrower and lender expect a 5 per cent inflation, and funds are worth a real 4 per cent, the interest rate stated in the contract would be 9 per cent. Then, with the 5 per cent rate of inflation, neither party gains nor loses.3 Hence, with greater inflationary expectations market interest rates are driven up to progressively higher levels. Since there is much uncertainty about the future course of prices and all people are not capable of making contracts against contingencies, inflation causes a redistribution of wealth and income. Money is noninterest bearing, and so adjustments cannot be made for reduced purchasing power of money hold ings. Many mortgages, pensions, bonds, and other long-term contracts cannot be changed until they ma ture. Persons with small savings have been especially disadvantaged by inflation. 2See Albert E. Burger, “The Effects of Inflation (1960-68),” this Review, November 1969, pp. 25-36. 3Income tax considerations would make the stated rate even higher, since the borrower is able to deduct from his income the amount of interest paid, and the lender must include as income the greater amount of interest received. 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 Inflation has a tendency to cause inefficiencies and reduce output. For example, many find it advanta geous to devote more effort to holding cash balances to a minimum. Because some prices are not com pletely flexible, shortages develop, causing inefficien cies in production. Uncertainties regarding the rate of inflation tend to encourage speculation. DECEMBER 1 9 6 9 Federal Government Expenditures N a t i o n a l In c o m e A c c o u n t s B u d g e t R a tio S c a le 500 400 400 300 300 +3 9 % 200 Actions Taken to Resist Inflation Because inflation is such a serious problem, the Government has regretted the policy errors which produced it, and taken a number of actions designed to resist or eliminate it. Unfortunately, many of these actions have been based on poor economic analysis, and they have proven to be largely ineffective, at least until very recently. Chief actions presumed and intended to be anti-inflationary have been raising tax rates, reducing the rate of increase of planned Gov ernment spending, regulating credit, permitting high interest rates, using moral suasion, and slowing the growth in money. Some have suggested that since these measures as yet have been largely ineffective in stopping price acceleration, the country should resort to a broad set of wage and price controls in order to govern prices. Fiscal Actions — Because the rapid increase in the demand for goods and services and the resulting ac celeration of price increases were thought to stem from expansionary fiscal actions, a serious step in re sisting inflation was to reverse these actions. In fact, it was widely felt that the best, and perhaps only, way to reduce inflation was to raise taxes and reduce Government spending. After long deliberations filled with assurances of potency of fiscal actions, the Revenue and Expendi ture Control Act was passed and signed into law on June 28, 1968. The Act represented a significant move in fiscal policy for the express purpose of moderating Digitized for Page FRASER 6 +14.5 To al - The presumed benefits of inflation are dependent upon its continued acceleration and upon the losses of some to the benefit of others. Only if the rate of inflation were stabilized, with all the public fully an ticipating it and acting upon the anticipations intelli gently and costlessly, would the rate of inflation be immaterial. But, under present conditions of uncer tainty, nonuniform expectations, and lack of flexibility, inflation is unjust, inefficient, and undesirable, and it is the stated policy of the Government to eliminate it. R a tio S c a le illio n s o f D o ll a r s Quarterly Data at An n ual Rates B illio n s o f D o ll a r s __ — '— — _ 793.6 100 100 80 +20 9% ' 80.3 +1.1% /V . --- 60 ---------- 40 20 200 +4.6% ^ lslqlr. 2nd qtr 2nd qtr. t ... 1961 80 60 1962 1963 1964 1 1965 1966 t 1967 2nd qtr. t 1968 3rd qtr. t 1969 Source: U.S. Department of Comm erce Percentages are an nu a l rates of change between periods indicated. They are presented to aid in com paring most recent developm ents with past "trends." Latest d ata plotted: 3rd quarter growth of total demand, and thereby reducing infla tion. The major features of the Act were reductions in some proposed spending and a 10 per cent sur charge on corporate and individual income taxes. Reflecting the provisions of the Act and a few sub sequent decisions made in the same spirit, growth in total Federal expenditures slowed sharply. In the last half of 1968 outlays in the national income ac counts budget rose at an 8 per cent annual rate, and in the first three quarters of 1969 at only a 4 per cent rate, compared with a rapid 13 per cent per year pace from late 1964 to mid-1968. The larger tax receipts, flowing from the surtax and the higher levels of the public’s income, and the slower growth in Government spending led to an abrupt reversal in the Federal budgets. The national income accounts budget went from a $9 billion annual rate deficit in the first half of 1968 to a $10 billion rate surplus in the first three quarters of 1969. On a high-employment budget basis, the shift was from a rate of $14 billion deficit to an $8 billion surplus. Despite the predominant feeling that a tax increase and some expenditure controls were necessary, many analysts in the late summer and fall of 1968 felt that the steps actually taken were too vigorous, and ex pressed a fear of overkill. Most econometric models of the economy indicated a quick and drastic slow down in spending and inflation as a result of the fiscal actions. Arthur Okun, then the Chairman of the Coun cil of Economic Advisers, summed up this opinion by stating, “I know of no one who would say now that our worries are still those of expanding too fast. If any 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 1 9 6 9 thing, the balance has shifted a bit in the other direction.”4 Responding to the marked shift in sentiment and expectations after the tax increase and the cut of planned Government spending, monetary policy was ostensibly relaxed. The directive to the System’s op erating manager on July 16, 1968, stated in part, “The new fiscal restraint measures are expected to contri bute to a considerable moderation of the rate of ad vance in aggregate demands,” and he was instructed to accommodate “ . . . the tendency toward somewhat less firm conditions in the money market . . ”6 Not all early evaluations of the impact of the fiscal action agreed that the inflation would be stemmed by this action alone. For example, this bank’s study of the effects of the fiscal program published in the August 1968 issue of this Review (pp. 3-6) concluded: Fiscal authorities have adopted a program of Federal budget restraint in an effort to com bat excessive total demand. It is hoped that this action will moderate inflationary pressures while only slightly affecting output and em ployment. However, an inflationary psychology has becom e entrenched in the economy, as evidenced by large wage settlements and the rising costs of credit. If the Administration and Congress have finally assigned high priority to the task of reducing inflationary pressures, monetary actions to complement the fiscal pro gram are needed. The tax increase and the controls placed on Fed eral spending did not produce the results expected by their sponsors. Excessive growth in total demand continued at only a slightly reduced rate. Slower growth in spending by the Federal Government was largely offset by greater outlays by those who were able to attract the funds formerly flowing to the Gov ernment to finance its deficits. To some, fiscal action as a tool of economic stabilization became largely discredited by this experience, while the more de voted followers of this approach believe that the in flation would have been much worse without the higher tax rates and spending cuts, that the actions were not large relative to the size of either the econ omy or the inflation, and yet total spending growth stopped accelerating and slowed moderately after the action. Interest Rates — Market interest rates have risen greatly since the early 1960’s, but the increasing rates, 4“What’s Ahead for Business,” U. S. News and World Report, August 5, 1968, pp. 52-55. 5Federal Reserve Bulletin, October 1968, p. 866. excessive total spending and resulting inflation. Yields on highest grade seasoned corporate bonds, which averaged less than 4% per cent in the first half of the 1960’s, rose to 5.13 per cent in 1966, 5.51 per cent in 1967, 6.18 per cent in 1968 and over 7.30 per cent in the fall of 1969. Interest rates on other marketable securities also increased sharply. Greater costs of credit, it was argued, should mod erate inflation since higher rates make expansion of spending more difiicult and are a stimulus to increased saving. The high interest rates have made it more costly for businessmen to purchase plant, equipment, and inventories. Higher rates have made it more expensive for consumers to buy homes, purchase automobiles, and obtain other durable goods. Yet, the excessive growth in aggregate spending has continued, placing more and more upward pressure on prices. The higher interest rates have not stopped the in flation, but rather the inflation itself has been largely responsible for the rise in rates. Interest rates are a price for the use of funds, and as a price, they are affected by inflation. Price expectations, as indicated by recent actual price behavior, have been a major factor in the rise of market interest rates.6 With expected inflation, saving is discouraged unless inter6“ lnterest Rates and Price Level Changes, 1952-1969,” by William P. Yohe and Denis S. Karnosky on pp. 18-38 of this Review. Page 7 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 1 9 6 9 est rates paid are high enough so that expected future real purchasing power of funds is protected, while boiTowers offer higher rates since they expect that the prices of investment goods will increase. annual rate from 1957 to 1968 to a 4.9 per cent rate in the first half of 1969 and to a 1.8 per cent rate of decline from June to November. With fewer funds flowing to them, banks have had less to lend or invest. The higher market rates in 1969 may have had no more stimulative effect on savers or restraining in fluence on investors than the lower market rates did in the early 1960’s. In fact, recent rates paid on sav ings accounts have not even equalled the attrition in the purchasing power of the dollar, and many bor rowers have anticipated that equipment and building costs would rise faster than the interest rate charged to finance purchases. Income tax considerations have also dulled the effectiveness of interest rates, since lenders must pay taxes on interest “earnings” and borrowers receive deductions for interest “expense.” Regulation Q limitations, however, have had no demonstrable effect on the total amount of credit ex tended in the economy. Funds have failed to flow through commercial banks because they have been attracted to users through other channels, such as di rect loans, commercial paper, and the Eurodollar market. By diverting loan funds through a second best route, the financial system has become less effi cient. Other financial intermediaries subject to similar regulations have also been severely affected. Interest rate limitations have made it more difficult for home buyers and smaller businesses who must rely on local institutions to obtain credit. Larger businesses which can obtain funds in the central money markets have probably obtained funds cheaper and more readily than in the absence of disintermediation. Small savers have been penalized by the low rates received, while larger lenders who have more alternatives have re ceived higher returns. Despite these inequities and disruptions to the financial system, it appears that growth in total credit granted (bank plus nonbank) was largely unaffected by the limitations placed on interest rates paid by financial intermediaries. Controlling Bank Credit — During 1969, the rate of growth of commercial bank loans and investments has been sharply reduced. In the first six months of the year outstanding bank credit rose at a 3.5 per cent annual rate, and since June has declined at about a 2 per cent rate. By comparison, this credit increased at an average 7 per cent rate from 1959 to 1966, and at an 11 per cent rate in 1967 and 1968. The reduc tion in the growth of bank credit has been brought about largely by regulating the maximum interest rate banks are allowed to pay on savings and other time deposits (Regulation Q ). The marked slowing in time deposits and corresponding bank credit growth seems unlikely to have exercised any restraint on excessive total spending and inflation. Regulation Q was instituted following bank failures of the early Thirties, primarily as a device to keep maverick banks from establishing rates clearly out of line with market conditions. However, in recent years as market rates have risen above Regulation Q ceil ings, and most banks have found it increasingly diffi cult to compete for time deposit funds. Regulation Q apparently has been considered a tool for influencing total spending, on the grounds that a lower growth in time deposits causes a reduced growth in bank credit. Largely because of limitations on interest rates they can pay on time deposits, banks have been drained of a substantial amount of time funds and have been unable to attract a normal flow. Large certificate of deposit obligations of commercial banks have fallen by more than half in the past year, from about $24 billion to about $11 billion. Growth in other time and savings deposits slowed from an 11 per cent Page 8 Moral Suasion — Another popular suggestion for re sisting inflation, which is usually considered to be largely ineffective, is to ask the public to forego rais ing wages and prices. A set of guideposts was pro posed by the President’s Council of Economic Ad visers in mid-January 1969: wages were to be raised no faster than average productivity growth ( estimated at about 3 per cent a year), and prices were to be established so as not to raise profit margins. However, the guideposts and other appeals to the public to use restraint in setting prices and wages have never been effective. Workers and businessmen cannot be expected to forego returns which are avail able to them. If they did, the economy would become less efficient; incentives would be dulled; shortages would quickly develop; and resources would not be attracted into areas of greatest demand. Some have suggested that prices and wages should be rigidly controlled by law, with severe penalties for violations. Yet, attempts to control prices in the past indicate that such controls have been largely ineffec tive, because blackmarkets develop, quality deterio rates, and in other ways effective prices are raised. 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 1 9 6 9 Money Stock Ratio Scale Billions o f D o lla rs Ratio S cale Billions o f D ollars M onthly A v e ra g e s of D a ily Figures S e a so n a lly Adjusted 210 210 205 205 +4.0% 200 200 from 1957 to 1964 when total spending was restrained. Most of the recent slowing of the money supply oc curred in the demand deposit component. The cur rency component, which responds more to trends in spending than to current changes in member bank reserves, continued to rise rapidly in 1969. 199 r 195 195 190 190 185 185 180 180 175 175 170 170 | -0.3% 165 165 160 160 155 N •o s O' o S 4- 150 11 Iti 1 1 11 l . u l t i 1 1.1-I..I.. 1966 1967 O' ;0 c -3 Q O> o Z 1 1 1 1 1 1 l 1 1 l I It; I 1 I itl I I 1 ltl 196 8 155 150 1969 Percentages are ann ual rates of c h an ge between periods indicated. They are presented The reduced rate of money stock growth during 1969 reflected a marked reduction in the rate of ex pansion in other monetary aggregates. Federal Re serve credit outstanding, which had risen at a 10 per cent rate during 1967 and 1968, slowed to about 5.5 per cent rate in the first five months of 1969 and to a 4.7 per cent rate after May. The monetary base after rising 6.5 per cent in both 1967 and 1968, in creased at a 5.3 per cent rate in the first five months of this year, and has slowed to a 1.5 per cent rate since then. The deceleration in total member bank reserves growth was even sharper. to aid in com paring m ost recent developm ents with p ast "trend s." Latest d ata plotted: N o v e m b e r prelim inary Controls are costly to administer, impinge on freedom, create shortages, and misallocate resources.7 Controls interfere with the continuous price adjustments which' are essential for equating supply and demand in the myriad sectors of the economy and attracting re sources to the uses where they are most needed. In flation may be “bad,” but an inflation temporarily repressed by a broad set of arbitrary wage and price controls is worse. Unless controls are accompanied by policies which will reduce the excessive demand for goods and services, they provide no solution to inflation. If total spending is restrained, the controls are unnecessary. Continued growth in spending at an excessive rate during most of 1969 was consistent with the monetary actions of the past two years. Growth of the money stock was very rapid until early 1969, and the expan sionary effects of such growth have usually been M onetary Base and Federal Reserve Credit Monetary Actions — Growth in the nation’s money stock has slowed markedly in the past year. This ac tion has already limited total spending, and studies of past lags of the effect of monetary restraint indicate that this action will be effective in further reducing demand for goods and services. Reducing the stock of money relative to the demand for it causes con sumers and business to spend less than their incomes in an attempt to build up actual cash balances to desired levels. During 1967 and 1968, the money stock rose at a very rapid 7 per cent annual rate. In the first half of 1969, growth in money slowed to a 4 per cent rate, and since June money has risen only slightly. By comparison, money rose at a 2 per cent trend rate [H J s e s of the m onetary b a se a re m em b e r b a n k rese rve s a n d currency held b y the p u b lic a n d n o n m e m b e r ban ks. A djustm ents a re m a d e for rese rve requ ire m e n t c h a n g e s a n d shifts in d e p o sits a m o n g c la s se s of b an ks. D a t a a re com puted by this b an k. [2 Total F e d e ra l R e se rve credit o u tsta n d in g inclu d e s h o ld in g s o f securities, loans, float, a n d "o t h e r " a ssets. Adjustm ents a re m o d e for rese rve requirem ent c h a n g e s a n d shifts in d e p o sits a m o n g c la s se s o f b an ks. D a t a a re com puted b y this bank. Percentages a re a n n u a l rates of c h a n g e betw een p eriod s indicated. They a re p resented 7See “Controlling Inflation,” a speech by Darryl R. Francis, President, Federal Reserve Bank of St. Louis, printed in the September 1969 issue of this Review, pp. 8-12. to a id in c om paring m ost recent d e ve lo p m e n ts with p a s t "tre n d s ." Latest d a t a plotted: N o v e m b e r p re lim in a ry 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 I S DECEMBER 1 9 6 9 Reserves of M ember Banks * SELECTED M O N ET A R Y AG G REG ATES (Annual Rates of Change) R a tio S c a le Bill o n s o f D o lla r s Dec. 6 7 to Dec. 68 Dec. 6 8 to June 69 Ju n e 6 9 to N o v . 6 9 p. Stock 7 .2 4.4 0 .8 D e m a n d D e p o sits 7.1 3 .7 — 0 .5 C urren cy 7 .4 6 .6 5 .4 9 .4 0.1 -4 .0 10.2 5 .9 4 .2 M o n e t a r y Base 6.5 4 .0 2.3 25 Total 7.8 0 .7 — 5 .9 24 M oney M o n e y Plus Tim e D ep osits Reserves R a tio S c a le B illio n s o f D o ll a r s 29 29 28 28 I V 2 7.4 27 27 y 26 Federal Reserve C redit M o n t h ly A v e r a g e s of D a ily F igures 26 /J 25 ( 24 p — P r e lim in a r y strongest after a lag of about two quarters. Hence, spending in the first half of 1969 was being affected primarily by earlier expansionary monetary develop ments. Then in the first half of 1969, money expan sion, although dampened, continued at a rate in ex cess of the trend since 1957. As a result, one might have expected only a gradual moderation in the growth rate of spending during the summer and early fall. Since mid-year monetary restraint has intensified, and the initial major effects of this action would nor mally be expected to occur in late 1969 or early 1970. Economic Developments in 1969 Despite the many actions taken in the battle against inflation, price increases accelerated in 1969. The greater inflation has been caused by both a continued strong demand-pull effect from excessive spending and a cost-push effect from previous excessive de mands for goods and services. As the year progressed, however, there were increasing signs that the de mand excesses were waning. First Half — In the first half of 1969, growth in total spending continued to be excessive and moderated only slightly from a year earlier. Total spending on goods and services rose at a 7.4 per cent annual rate, down from the 9 per cent rate in the previous six quarters, but only a slightly less than the average 8 per cent rate that had prevailed since late 1964. Final sales (that is, total sales less changes in business in ventories) rose at an 8.3 per cent rate in the first half of 1969, virtually the same as in the previous year and a half, and in the entire period since late 1964. Growth in production, however, slowed in early 1969. Real output of goods and services rose at a 2.3 Page 10 23 23 22 22 21 20 £ s fc * * I a* ■*? 5 z . i . t ------- 1966 C O — t ------------------------------ . 1967 i i 1 i i 1 i i 1 i i tl 1 1 i f 1968 21 0 1 1 1 !■ itl 20 1969 ‘ D a t a before M a y 1969 h av e been a d justed fo r estim ated effect o f reserve requirem ent c h a n g e s . C u rre n t d a ta exclud e in cre a se in re q u ire d re se rv e s d u e to c h a n g e s in R e g u la t io n s M a n d D, effective O c to b e r 16,1969. P e rc e n ta g e s a re a n n u a l rates of c h a n g e betw een p e rio d s in d ica te d . T he y a re p re se n te d to a id in c o m p a rin g m ost re c e n t d e v e lo p m e n t s w ith p a st "tre n d s ." L atest d a t a plotted: N o v e m b e r prelim inary per cent annual rate during the first six months of 1969, about half the rate of the period from late 1964 to late 1968. Despite cutbacks in total output, both industrial production and employment continued to rise at rapid rates, and the level of unemployment remained unusually low. The upward trend in total production was probably restrained as the economy approached capacity and could not physically main tain the earlier growth pace. Growth of productivity slowed reflecting inefficiencies resulting from the inflation. With spending continuing to rise rapidly and with production expanding at slower rate, the pace of in flation accelerated in early 1969. Overall prices went up at a 5 per cent annual rate in the first half of the year after rising 4 per cent in 1968 and 3.5 per cent in 1967. Consumer prices rose at a 6.4 per cent rate in the first half of 1969 compared with 4.7 per cent in 1968 and 3.1 per cent in 1967. Wholesale prices increased at a 6.3 per cent rate in the first half of 1969 following a 2.8 per cent rise in 1968 and a 0.8 per cent increase in 1967. Second Half — Evidence of some real progress in combatting the economic ebullience was provided by a number of sensitive data series during the third quarter of 1969. Although the overall measure of F E D E R A L R E S E R V E B A N K O F ST. LOUI S DECEMBER 1 9 6 9 Prices Ratio S c a le Ratio Sc a le 1957-59=100 135 1957-59=100 135 Aug 66 1961 1962 1963 1964 1965 1966 Aug.67 ♦ 1967 1968 1969 Source: U.S. Department of Labor Percentages are a n nu a l rates of chan ge between periods indicated. They are presented to aid in comparing most recent developm ents with past "trends." Latest data plotted: Consum er - O ctober; W h o le sa le - Novem ber total spending expanded slightly faster than in the first half, analysis of spending by major categories indicated much less strength. Spending was supported by an unusually large catch-up Government pay raise in July. Nevertheless, final sales rose at only a 6.3 per cent annual rate, down from the 8.3 per cent rate in the first half. Total sales were bolstered by a sub stantial, probably largely unintended, buildup in in ventories, as final sales apparently did not measure up to earlier expectations. An unplanned rise in business inventories while sales fall below expectations is fre quently a sign of developing weakness in total demand. Real output of goods and services grew at a 2.2 per cent annual rate from the second to third quarter of 1969, about the same as in the first half. Other measures of performance, however, indicated weak ness. Real final sales were about unchanged from the second to the third quarter after rising at a 3.3 per cent annual rate in the first quarter. Industrial pro duction declined at a 5 per cent rate from July to November after rising at a 6 per cent annual rate from last December to July. Payroll employment rose at a 1.1 per cent annual rate from June to November, fol lowing a 4.2 per cent increase in the first half. Total civilian employment rose at a 1.6 per cent annual rate from August to November, after increasing at a 2.8 per cent rate in the first eight months of the year. Private nonfarm housing starts were at a 1.4 million rate from July to October, down from a 1.6 million rate in the first half. Prices have risen since June at about the same pace as during the first half. The overall measure increased at a 5.4 per cent annual rate from the second to third quarter, up from a 5.1 per cent rate earlier in the year. Consumer prices rose at a 5.1 per cent annual rate from July to October, compared with a 5.9 per cent rate in the first seven months of the year. W hole sale prices increased at a slower pace from June to November, but the improvement provided little en couragement to those seeking to control inflation since it reflected a decline in farm prices resulting primarily from a jump in supplies. Personal income grew at a 4.5 per cent annual rate from August to October after growing at an 8.8 per cent rate in the first eight months of the year. Rusiness spending may have been moderated by rising inventories, declining corporate profits, and less op timistic expectations. Consumers found personal in come rising at a slower rate after mid-summer, a development which, according to surveys, was ac companied by deterioration of consumer confidence. State and local governments have reduced the growth rate of their outlays in response to higher interest rates, in some cases to levels above ceilings permitted, and to public discontent with ever rising tax burdens. Summary and Outlook The past year has been the worst of five successive years of inflation. Overall prices have risen about 5 Industrial Production Ratio S c a le R atio Scale 1957-59=100 200 1961 1957-59=100 Se ason a lly Adjusted 1962 1963 1964 1965 1966 200 1967 1968 1969 Percentages ore an nu al rates of chan ge between periods indicated. They are presented to aid in comparing most recent developm ents with post trends." Latest data plotted:October preliminary 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 I S per cent while consumer prices have increased 5.6 per cent. The inflation has caused a substantial re distribution of real wealth and income and has created inefficiencies. The inflation has resulted from an unduly large demand for goods and services which was nurtured by expansive monetary developments in 1967 and 1968. In addition, prices have been forced up in creasingly by cost-push forces resulting from the de layed effects of previous excesses. The rapid increases in spending and inflation have continued despite higher tax rates, cuts in Government spending pro grams, higher interest rates, a reduction in the growth of bank credit and exhortations by public officials for business and labor to use restraint in raising prices and wages. The end of inflation is not clearly in sight. The re cent record of economic advice and business forecast ing has not been distinguished for its accuracy. A slowdown in activity and inflation widely predicted for the last half of 1968 and then for the first half of 1969 as a result of the restrictive policies adopted failed to materialize. This unimpressive record should engender caution, if not humility, for those who ven ture judgments as to current economic prospects. Monetary actions continued very expansive through 1968, and projections based on these developments have not been misleading, except when based on short-run movements in preliminary data which were later revised. Monetary expansion was moderate in the first half of 1969, and since midyear the money stock has been virtually unchanged. Experience indi cates that spending usually slows in about two quar ters after a marked reduction in the growth rate of money. After mid-1969, evidence began accumulating that the economic environment was changing, that the ex pansionary forces were weakening, and that a slack ening in the quarter-to-quarter growth rate of spend ing was in prospect. It now (early December) ap pears that further reductions in the growth of total spending and production appear to be in prospect for early 1970 as a result of the monetary restraint in the summer and fall of 1969. Even if growth of total spending continues to slow and is moderate in 1970, inflationary forces will prob ably remain serious throughout the year and perhaps for some time afterward. Expectations of rising prices are strong. Price increases usually continue for an ex 12 Digitized for Page FRASER DECEMBER 1 9 6 9 tended period after growth in overall demand for goods and services moderates, reflecting cost-push forces generated by the earlier excessive spending. It took about seven years of restrained growth of spending in the 1950’s to eliminate the inflationary pressures created during the Korean conflict. Some prices, such as negotiated wages and those set in other contracts, which have been relatively inflexible during recent periods of rapid price increase, will probably be adjusted upward later at time of rene gotiation. Other prices have been held back by a money illusion, lack of knowledge of costs, public opinion, and inertia. As these wages and prices ad vance toward equilibrium levels, the increase in pro duction costs will place upward pressure on other prices. Nevertheless, the campaign against inflation seems to be yielding results in the last half of 1969, as infla tionary pressures probably passed their peak in in tensity. The results of great economic imbalances are likely to be felt more keenly in 1970 than they were in the late 1960’s when they were generated. Assessment of economic prospects suggests that the country faces a very difficult period. Spending is likely to be sluggish, with corollaries of reduced pro duction, rising unemployment, and a squeeze on busi ness profits. At the same time significant deceleration of price increases may be dishearteningly slow. Suc cess will require great perseveranoe which is more likely to be achieved if extreme erratic stabilization actions can be avoided. The crucial consideration for stabilization policy makers in the coming year will be the determination of how rapidly the excessive growth of total spending can and should be reduced. Because of past errors, the choice now is to determine the lesser of evils. If monetary and fiscal actions are followed which will slow the rise in total demand for goods and services abruptly, inflationary pressures may be extinguished sooner than otherwise. However, the costs in terms of lower production, employment and incomes would be great, and the temptation would be strong to restimulate the economy before the task is completed as was the case in 1967. On the other hand, if de mand grows so rapidly as to permit growth in produc tion, employment and income to continue at near their long-run maximum trends, moderation of inflationary pressures may not be achieved and we are likely to have a continuation of the inefficiencies and inequi ties caused by a continuous erosion of the value of the dollar. A middle course is advisable. Selective Credit —No Substitute for Monetary Restraint A speech given by DARRYL R. FRANCIS, President, Federal Reserve Bank of St. Louis, at the 23rd Annual Conference of Bank Correspondents, First National Bank of St. Louis, November 13, 1969 I HE USE of selective credit controls for economic stabilization is not of recent origin. The eligible paper provisions of the Federal Reserve Act passed in 1913 are a form of selective credit control. They provide for easier Federal Reserve Bank credit terms to bor rowing member banks who offer short-term commer cial paper as collateral. This provision implies that, if bank credit is limited to short-term commercial loans, monetary conditions will approach an optimum. Selective credit controls were given a major boost in 1934 with the passage of the Securities Exchange Act, which delegated authority to the Board of Gov ernors to control margin requirements (Regulations G, T and U ) on securities traded on the major ex changes.1 This control was provided for the purpose of preventing the “excessive” use of credit for pur chasing or carrying securities. It was generally be lieved that a large volume of bank credit led to a major increase in stock prices and the ultimate col lapse of the stock market in the late 1920’s and early 1930’s.2 It was assumed that the expansion and con traction of stock market credit was; a major factor contributing to the Great Depression of the 1930’s. 'Securities Exchange Act, 1934, Federal Reserve Act as amended through Oct. 1, 1961, p. 222. 2See John T. Abele, “Black Tuesday, ’29: Forgotten Lesson?” New York Times, Oct. 26, 1969 for a discussion of this view. Another boost to selective credit controls was pro vided by the passage in 1933 and 1935 of interest rate restrictions on demand and time deposits. Under authority granted by this legislation both member and nonmember banks were prohibited from paying interest on demand deposits. This legislation also au thorized the Federal Reserve Board and the F.D.I.C. to limit the rates that banks could pay on time and savings deposits. At that time it was contended that these restrictions would tend to reduce the rates charged to bank customers, slow the movement of funds from smaller to larger communities where it was believed that they were used for “speculative” purposes, and prevent the excessive bidding up of rates, thereby reducing bank failures. More recently, time and savings deposit rate controls have been in tended to reduce competition between banks and savings and loan associations, and thereby to speed the flow of funds into the housing market. This move away from free competition implies that it is better for the nation to have larger credit flowsi to the housing industry than to bank creditors, who may use the funds for plant and equipment expenditures and other purposes judged to be less worthy. In 1941 the Board of Governors of the Federal Reserve System was authorized by an Executive Order of the President to restrict the use of credit for Page 13 F E D E R A L R E S E R V E B A N K O F ST. LOUI S consumer purchases (Regulation W ). This was done by requiring a minimum down payment and a maxi mum time to maturity for such credits. The immedi ate reason for the restriction was to reduce the infla tionary danger during the war by restricting the de mand for consumer goods.3 Actually, however, the legislation was associated with the thought that ex cessive consumer credit in the 1920’s had contributed to the 1929-33 depression. Restrictions on real estate credit were imposed in 1950 as part of a program to check inflation after the start of Korean fighting (Regulation X ). These restrictions followed the general pattern of consumer credit controls by requiring minimum down-payments and maximum maturities on new one- and two-family houses. Credit restrictions were later extended to other real estate. As an indication of the wide interest in the use of selective credit for controlling inflation, Time maga zine, in discussing the reluctance of the Administra tion to impose price controls, reported that “ credit controls, which were last imposed on the U. S. during the Korean War, might work more selectively to re strain lending, and in turn, demand for some kinds of goods” .4 In recent testimony before a Senate sub committee, the President of the National Association of Home Builders stated that “we are now convinced that some type of credit controls must be undertaken”.5 Selective Credit Controls — Pros and Cons The argument for selective credit controls rests primarily on the assumptions that restrictive mone tary actions cause high interest rates, and that high interest rates have an unduly harsh impact on some sectors of the economy such as residential construc tion. Proponents of selective credit controls believe that the “undesirable” impact of aggregate monetary controls can be eliminated by controlling the volume of credit used for specific purposes. The view that credit controls in a specific sector will have an impact on total credit outstanding and total demand for goods and services may be looked upon as a variant of the Income-Expenditure ap proach to economic activity. For example, if one 3Charles R. Whittlesey, Arthur M. Freedman and Edward S. Herman, Money and, Banking: Analysis and Policy (New York: Macmillan Co., 1963), p. 256. *Time, Oct. 10, 1969, p. 87. 5U.S., Congress, Senate, Subcommittee on Financial Institu tions of the Committee on Banking and Currency, Hearings on S-2499 and S-2577, 91st Congress, 1st session, September 9, 1969, p. 17. Digitized forPage FRASER 14 DECEMBER 1 9 6 9 viewed total economic activity as resulting from au tonomous expenditures in each of the separate sectors of the economy, a variation of expenditures in any one sector would have an impact on total expendi tures and total income. This is consistent with the fiscal view of economic stabilization, which asserts that an increase in Government spending on goods and services results in an increase in total expendi tures and total income. However, I do not mean to imply that all proponents of fiscal stabilization would recommend the use of selective credit controls for stabilization purposes. In contrast to the Income-Expenditure approach to economic activity, I hold to the view that the stock of money is a major influence on total demand and the course of spending. Preferences for specific goods and services determine amounts that consumers will spend in each sector. If specific credit controls or other nonmarket controls are applied in any one sec tor, I believe that any reductions achieved will be offset by higher expenditures for other goods and services. Despite the merits attributed to selective credit controls, I believe that they are socially undesirable for the following reasons: they are useless in con trolling inflation; most of the hardships attributed to general monetary restraint are actually caused by selective controls or self-imposed rigidities; they are difficult to enforce; they are biased in favor of the wealthier groups; to the extent that they reduce de mand for particular goods and services, they also reduce national welfare; and they are a restriction on individual freedom. Selective Credit Controls — for Quantitative Controls No Substitute The argument that selective credit controls will restrain inflation is not consistent with the functioning of our monetary mechanism and the factors which determine total demand. With our fractional reserve system of banking, the volume of bank credit is strongly influenced by total reserves and the reserve ratio requirements. Within limits, total credit is de termined by the Federal Reserve System simultane ously with the determination of bank reserves and the stock of money. With fixed reserve ratios, other monetary multipliers, and a given level of bank re serves, credit restrictions on some purchases are fully offset by credit expansion for other purchases. Thus the total stock of money or credit remains unchanged after application of the restrictions. Money created 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 by one type of credit expansion is equal in quality to money created by another type of credit, and total demand is unchanged. Even if one subscribes to the idea that monetary management exercises its influence primarily through the credit market rather than through the money stock, selective credit controls are not a solution to the problem of excess demand. Credit restrictions on some purchases will cause rising demand for uncon trolled goods and services. Prices for uncontrolled goods will rise faster and resources will, in turn, flow from the production of controlled to uncontrolled goods. Output will be enhanced in the uncontrolled sectors and reduced in the controlled sectors, but prices for all goods will continue up, assuming ex cessive credit and monetary expansion has been permitted. The argument that quantitative controls are per verse and create undue hardships in some sectors implies that the judgment of the controller is superior to market decisions. To the extent that the controls work, the controller is essentially transferring to others his own values as to what the nation should produce and sell. I contend that the market place can de termine with least welfare loss which goods and serv ices should be produced at slower or accelerated rates as a result of stabilization actions. Consumer pur chases of goods and services under aggregate mone tary controls are determined by utility at the margin, thus providing greater welfare than purchases arbi trarily determined for specific products through selective credit restrictions. Furthermore, appropri ate monetary policies will reduce the wide swings in demand of recent years. A reduced amplitude of demand fluctuations would eliminate the unevenness of the effect of controls on total money and credit. Credit used for purchasing and carrying common stocks has been given major attention because of the widespread belief that stock market credit and stock prices tend to trigger major swings in economic activ ity. Proponents of this view contend that speculators, when borrowing to make stock purchases, bid up stock prices to excessive levels and the following sharp declines tend to produce recessions. I believe that both the impact of credit on stock prices and the impact of stock prices on economic activity have been greatly overestimated. Much of the fluctuation in the stock prices can be traced to the unevenness of aggregate monetary controls. Stock prices may be influenced temporarily by the volume of credit extended for security purchases. Reverse DECEMBER 1 9 6 9 causation, however, is more likely the case. In other words, movements in stock market credit are influ enced by changes in stock prices. Statistical analysis tends to demonstrate this reverse causation.6 Causation likewise runs from economic activity to stock prices. Rather than being an important factor contributing to the Great Depression as some contend, the sharp decline in the stock market in the late Twenties and early Thirties was mainly the result of a decline in earnings and in earnings expectations. This outlook for earnings can be traced to a decline in demand for goods and services, which, in turn, can be traced to a sharp reduction in the stock of money. Carrying the securities market analysis a step fur ther, we can assume that business capital will expand according to profit incentives and the cost of capital to entrepreneurs. If margin restrictions restrain the opportunity for raising capital through security sales, business will likely resort to borrowing directly from financial institutions to meet capital demands. I can see little difference between making loans to corpora tions and making them on securities which represent ownership of corporations. Furthermore, like other credit restrictions, if margin requirements alter credit or monetary flows, they also reduce national welfare as indicated by individual expenditure preferences. The Apparent Need for Selective Credit Controls — A Result of Other Restrictions Most of the hardships attributed to general mone tary restrictions by advocates of selective credit con trols would disappear if other useless impediments to credit flows were eliminated. Quite often the alleged victims of financial market imperfections are actually the victims of other controls. First, let us look at the argument that restrictive monetary actions discrimin ate against residential construction. Recent studies indicate that relatively slow rates of monetary growth '’'Statistical analysis of the Standard and Poors 425 Industrials and of the 500 Stocks gives better results for the hypothesis that changes in stock values cause changes in credit for stock purchases than the reverse-causation hypothesis. The hypotheses were tested by regressing first differences of monthly data (stock price indices and credit extended to margin customers by banks plus customers’ net debit balances at member firms of the New York Stock Exchange) on current and three lagged periods. The time period used was February 1953 to July 1969. The response of credit to changes in stock values was positive and significant for the current plus the first and second lagged month for each index. In contrast, the hypothesis that stock market credit causes stock prices to rise gave positive and significant results only in the current period. The coefficients were nega tive in the first and second lagged month for both indices. Page 15 F E D E R A L R E S E R V E B A N K O F ST. LO U I S do not cause excessive cutbacks in spending on homes. Conversely, all marked and sustained declines in hous ing starts began in periods of rapid monetary expan sion after excessive demands for goods and services had driven up prices and interest rates. The sharp rises in costs and interest rates were the major factors in reducing housing demand, and the reduced flows of funds into housing during these periods of high interest rates can be traced in part to such market impediments as usury laws, interest rate ceilings, and other regulations on financial intermediaries. When market rates reached these imposed ceilings, funds were diverted to other uses where rates could move to market-determined levels. An additional side effect of this diversion of credit from its normal flows through intermediaries is its bias against small savers and borrowers. Small savers are relatively unable to place their savings in funds and securities and are relatively limited to controlled rates. These rates are lower than market-determined rates and small savers are the losers. Similarly, small borrowers are limited to the use of financial agencies for credit, and when this credit dries up because of rate regulations, small borrowers are, for all practical purposes, banned from the credit market. In contrast, large borrowers can participate in the capital mar kets through either new common stock offerings or other securities. Self-imposed rate restrictions by state and local governments and rate restrictions on public utilities likewise reduce their expenditures during periods of high market rates. Rather than serving to reduce in terest costs, the restrictions simply serve to postpone the expenditures until supply and demand conditions for credit cause market rates to return to levels that these spending units are willing to pay. Enforcement of Selective Credit Controls is Difficult Selective credit controls are extremely difficult to enforce uniformly among all groups. Enforcement of ficials can apply restrictions on a basis of the collat eral offered as security, on the borrower’s declaration of intended use of proceeds, and on the indicated use of proceeds. None provides a sure test of the use of loan proceeds. If proceeds from loans are com mingled with other income such as salaries, wages, commodity sales, or other funds, one cannot deter mine which funds were used for the various expendi tures. For example, if an individual wants to use his salary, wages, or other sources of income to purchase Page 16 DECEMBER 1 9 6 9 securities, he may borrow funds to make home pay ments or provide for other living expenses. The bor rowed funds are similar to the other funds and the destination of the respective funds is difficult to trace. The collateral offered is likewise a poor indicator of how loan funds are used. Proceeds from loans on common stocks or real estate may be used for medical payments, to purchase automobiles, or for numerous other purposes having no connection with stocks or real estate. Trade-ins are also a means of by-passing down payment requirements of selective credit controls. With anything that can be considered a trade-in item, the prospective buyer and seller can get together and make a mutually satisfactory deal by setting up a fictitious down-payment which meets both legal and personal requirements. Selective Credit Controls Biased in Favor of the Wealthy Selective credit controls are biased in favor of wealthy groups and against those with limited assets. Real and financial assets can always be used as col lateral for loans. From such proceeds down-payments on purchases of controlled items can be made unless each dollar can be traced to its ultimate use. Further more, those with assets for collateral can avoid instal ment credit restrictions altogether by obtaining com mercial credit and purchasing consumer items with the proceeds. In addition, those with wealth are also likely to have sufficient cash flows to mingle with borrowed funds to make fund-tracing almost impos sible. Control officials cannot determine whether the borrowed money was used for down-payments on controlled items or whether other cash flows were used for such purchases. On the other hand, those without assets are forced to use purchased items as collateral. Thus, to the ex tent that selective credit controls serve to retard de mand in a particular sector of the economy, they are a boon to persons with assets, providing them with products at lower prices, while those without assets for collateral are frozen out of such markets until the necessary down-payments can be accumulated. Selective Credit Controls Distort Resource Use and Reduce Welfare Fortunately, most selective credit controls are not readily enforceable. To the extent that they reduce demand and production of goods and services in any 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 sector, they tend to reduce welfare. The welfare of an individual is maximized, other things being equal, when he can spend both proceeds from loans and funds from other sources without restrictions or im pediments. Each expenditure then provides him with maximum satisfaction at the margin. On the other hand, the capricious use of restrictions to alter in dividual spending, either in the form of higher down payments or shorter terms, requires consumers to make less desirable choices. Fewer goods and services are purchased in terms of their usefulness for the same level of expenditures. An economy operating under selective credit restrictions fails to produce an opti mum amount of some goods and overproduces other goods, with a consequent loss in total welfare. Selective Credit Controls Reduce Freedom Equally as important as the economic considera tions is the useless infringement of selective credit controls on freedom. W e can easily moralize as did the Medieval rulers that the poor should stay out of debt or that someone should set limits on their loan terms in order to assure that their credit is used for “appropriate” purposes. It is my belief, however, that man is happier when subject to the market forces rather than to arbitrary decisions of one or a few individuals. Freedom did not come easily to mankind, but we tend to take it for granted. Yet, in most of the periods since man’s early history, he has been forced to bow in both thought and action to harsh taskmasters. In my view, we should not take losses of freedom lightly, despite the fact that controls are imposed upon us only a fraction at a time. Summary The current inflation has brought to the forefront proposals for using selective credit controls to stabilize prices and prevent the allegedly perverse effects of aggregate monetary restraint. Despite the possibility that selective controls may reduce the volume of credit used for specific types of purchases, they do not achieve the announced stabilization goals of the controllers. They do not restrain total demand for goods and services. Dollars created by one type of bank credit have the same purchasing power and the same impact on demand as dollars created by an other type of credit. It is the total amount of credit and money created that determines average prices for all goods and services. Actually, to the extent that selective controls cause misallocation of resources, they have an inflationary impact. DECEMBER 1 9 6 9 Selective credit controls are almost impossible to enforce with equal results among all groups of in dividuals and businesses. Their use imposes much greater credit restraint on the small borrower, who is without other assets for collateral or large cash flows which can serve to disguise the actual use made of borrowed funds. To the extent that selective credit controls are ef fective in changing credit flows, they reduce total output of goods and services and national welfare. Furthermore, they are an infringement on freedom because they impose restraints on how one can utilize his credit resources. Last but not least is the fact that selective credit controls tend to provide their own breeding grounds. One control must ultimately lead to another as market forces tend to bypass each new regulation. The pro liferation of ceilings on time and savings deposits is a good example. First, the ceilings were established uniformly on all accounts. Then there was a “need” to segregate deposits by size because of major losses of larger deposits. Smaller depositors could be paid less under monopolistic pricing because of lack of alterna tive investment opportunities. Rates to small savers were seldom changed. They remain at levels insuf ficient to cover the rate of inflation. Permissible rates on the larger accounts were raised at intervals as market rates continued up. Observing an opportunity to capture funds, bank holding companies and bank affiliates began to issue savings-type instruments which were not covered by the regulations. Now controls have been proposed in this leakage area. These at tempts to cover all bypasses are like the man who built a dam to curtail the normal flow of a stream and discovered a leak. He used his finger to plug the hole. As water backed up, however, other leaks oc curred requiring more fingers. The process of leak springing and finger plugging continued until the builder ran out of fingers. Inflations can be controlled, but not through the use of specific controls on arbitrarily selected goods or services. The solution to inflation lies in the adoption and maintenance of appropriate monetary policies, which attack the cause of inflation. It requires an appropriate rate of growth in the stock of money. W e moved toward a reduced rate of monetary growth near the end of last year and a still slower rate last June. I am confident that these actions will soon re duce the excess demand which was created by overly rapid monetary expansion in 1967 and 1968. Page 17 Interest Rates and Price Level Changes, 1952-69* by W ILLIAM P. YOHE and DENIS S. KARNOSKY Our economy has been experiencing an accelerating inflation during the past five years. At the same time, market interest rates have risen to extremely high levels. The causes of infla tion are relatively well-known, but the reasons for high interest rates accompanying inflation are not. The following article investigates primarily this latter situation. Inflation develops when, at a high level o f resource utilization, total spending on final goods and services ( GNP) rises at a rate faster than the rate at which productive potential grows. Such has been the case in this country since early 1965. Total spending has risen at an 8 per cent annual rate and real product at a 4.4 per cent rate. As a consequence, the over all price level has risen at a 3.6 per cent annual rate. The major cause of the current inflation has been the stimulus to total spending provided by an excessive rate of expansion in the money stock. From early 1965 to the end of 1968, the money stock, on balance, grew at a 5.2 per cent annual rate, compared with a 2 per cent trend rate in the preceding decade. Rapid growth in the money stock accompanied by high and rising market interest rates has appeared a paradox to many observers. According to modern Keynesian economic theory, an acceleration in the rate of monetary expansion will provide lower market interest rates. However, this apparent paradox can be explained by the economic theory developed by Irving Fisher around the turn of the century. According to Fisher, nominal ( observed) interest rates consist of two components — the “real” rate of interest, to which real saving and investment respond, and a premium based on expected changes in the price level. The following study uses this Fisherian analysis to quantify the effect of inflation on movements in interest rates from 1952 to 1969. The prin cipal finding is that past price movements exert a major effect on nominal interest rates, with the effect largely manifested within two years. Consequently, most of the rise in market inter est rates since 1965 can be attributed to the current inflation. This finding has an important implication for market interest rates as an indicator of the thrust of monetary actions on economic activity. High market interest rates do not neces sarily indicate monetary restraint. Instead, they most likely indicate excessive monetary ease, ( as measured by rapid expansion of the money supply) which results in rapidly expanding total spending and eventually inflation. William P. Yohe is currently a visiting scholar with this bank and is also Professor of Economics and Director of Graduate Studies in Economics at Duke University. H e is the author of numerous publications, primarily in monetary economics. Denis S. Karnosky is an economist with the Federal Reserve Bank of St. Louis. Page 18 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 1 9 6 9 X n SUMMARIZING his many years of work on the subject, Irving Fisher cited four empirical relation ships between interest rates and price levels:1 in order to stimulate sufficient investment spending for the price level to be high, while empirically this has not been observed. (1 ) Interest rates tend to be “high” when prices are rising and “low ” when prices are falling. The present study is an examination of the second and third of Fisher’s propositions, making use of modern data sources and statistical techniques. There is, at present, a major controversy over (1 ) the ad vantages and disadvantages of using monetary ag gregates as opposed to using interest rates as indicators of the effect of monetary policy actions on the econ omy, and (2 ) the adjustments, if any, which must be made to an indicator to “neutralize” it with respect to changes that are not directly the result of policy actions.3 Previous studies of the effect of price level changes on interest rates, some of which will be re viewed below, have found the lags to be so long that recent price behavior could be ignored in evaluating changes in observed interest rates. In contrast, results will be presented here based on the 1952-69 period which indicate that the lags are very short, with most of the effect of price level changes on both longand short-term interest rates occurring within two years. Interest rates adjusted to remove the ap parent influence of price changes have sometimes moved contrary to movements in observed rates. Furthermore, price changes have had a greater effect on interest rates in the 1960’s than in the 1950’s, and indeed, price changes in the latter period account for nearly all of the movement in interest rates. (2 ) Interest rate movements lag behind price level changes, which obscures the relationship between them. (3 ) There is a marked correlation between interest rates and a weighted average of past price level changes, reflecting effects that are distributed over time. (4 ) “ High” interest rates accompany “high” prices, and “low ” interest rates accompany “low ” prices. The first of these relationships derives from the fact that, if lenders and borrowers could perfectly foresee future price level movements, the former would hedge against changes in the real value of their loan principal by adding the percentage change in prices over the life of the loan to the interest charge; the latter, expecting money income to change in pro portion to prices, would readily accept the higher rate. Fisher attributed the second and third relationships to imperfect foresight about future prices and the resulting inclination to extrapolate past price changes into the future in order to adjust interest rates for expected changes in prices. He devised the concept of the “distributed lag” to explain the way informa tion about the past affects expectations of the future. Fisher thought the fourth relationship, frequently called the “Gibson paradox,” was an accidental con sequence of the other three.2 What is paradoxical is that the theory prevalent in that period presumably led to the conclusion that interest rates must be low “ The authors are grateful to Christopher T. Babb and H. A. Margolis for advice on the statistical problems of this study, to Shigeyuki Fukasawa and James B. Greene for making available the results of their unpublished studies, and espe cially to Keith Carlson, Michael Keran, Thomas Havrilesky, and Edward Kane for helpful suggestions on earlier drafts of this paper. 1Irving Fisher, The Theory of Interest (New York: Macmillan, 1930), p. 438. Fisher first discussed these relationships in Appreciation and Interest (New York: Macmillan, 1896), pp. 75 and 76. 2The term “ Gibson paradox” was coined by J. M. Keynes in A Treatise on Money, Vol. II (London: Macmillan, 1930), pp. 198-208. A. H. Gibson had studied the high correlation between levels of interest rates and prices in England throughout the 19th and early 20th centuries. The henomenon was earlier called the “ Ricardo-Tooke conunrum,” after the leading antagonists in the Currency SchoolBanking School controversy in England in the first half of the nineteenth century. For a concise exposition of the controversy, see Knut Wicksell, Lectures on Political Econ omy, Vol. II (London: Routledge and Kegan Paul, 1935), pp. 168-190. Previous Studies of Price Expectations ( Fisher) Effects Tests for Fisher effects have generally been based on two hypothesized relationships: (1 ) mt (2) P* = 2 WiPt—i t = PF t + rrt i= o The first equation states that the nominal interest rate (rn) prevailing at time t for a particular debt instrument is equal to the annual rate of change in prices (Pe) expected at time t to occur over the life of the instrument plus its “real” rate of interest (rr) 4 3See, for example, Leonall Andersen, Michael Keran, and Emanuel Melichar, “The Influence of Economic Activity on the Money Stock,” this Review, August 1969, and Patric H. Hendershott, The Neutralized Money Stock (Homewood, Illinois: R. D. Irwin, 1968). 4Fisher used “real” rate in the sense of “virtual” or “true” rate. Technically, he also included a third term, rrtPj, on the right side of equation (1 ). This is the interest that would be earned on the price adjustment to the nominal rate. The term is ordinarily so small that it is customarily Page 19 F E D E R A L R E S E R V E B A N K O F ST. LOUI S Equation (2 ) is an application of the theory o f “ adaptive expectations,” “error-leaming,” or, alternatively, “extrapolative forecasting.” Faced with uncertainty about the future, an economic de cision-making unit is presumed to base its pre dictions about future price m ovem ents on a weighted average of current and past changes in prices. Thus, in equation (2 ) the rate of price change expected at time (P?) for some future period is the weighted sum of actual past price changes ( P t- i ) , where the importance of each past change is re flected in the weight xvu and where n indicates how many periods in the past are relevant in forming expectations.5 The approach is “adaptive” in the sense that in each period expectations are adjusted (or forecasting errors are corrected) for actual price changes. The approach is “extrapolative” in that past changes are extended (extrapolated) into the future. Substituting equation (2 ) into equation (1 ) yields the form of the equation that is usually estimated: n (3 ) 2 rnt = wiPt-i + rrt i= o The unmeasurable price expectations are not ex plicitly considered, but instead it is assumed they can be approximated by the observable pattern of past changes in actual prices (or in some other variable that may be critical to the formation of expectations about prices). Fisher assumed that the weights in equation (3 ) declined arithmetically as one goes backward in time. His procedure was first to posit a time interval over which the entire effect of price level changes would omitted. For the complete derivation of equation (1 ), see Appreciation and Interest, pp. 8-11, 66 and 67. Some studies have also been concerned with the effect of changes in the rate of price change (i.e., price level ac celerations) on changes (rather than levels) in interest rates. To see how this may be done, it is necessary to expand P“: dPe pe — __ 1 Pt Substituting this term in equation (1 ), differentiating, and manipulating the result yields: / d 2Pe d(mt) = dP° \ - p rJ be reflected in a nominal interest rate series, for example, ten years. Ignoring the current period price change, he then computed for each year the weighted average of past price level changes, using a weight of nine for one year earlier, a weight of eight for two years back, and so forth. The weighted price changes divided by the sum of the weights (9 -)- 8 — | — . . . . + 0) yielded the weighted average of past rates of price change. Fisher then observed which of these weighted averages best correlated with the nominal interest rate.6 The “best fit” would be obtained where the correlation was highest or where further length ening of the interval would not add appreciably to the correlation.7 A useful statistic for comparing the results of many distributed lag studies is the mean (or average) lag, that is, the time that elapses until half of the effect of a change in the independent variable is reflected in the dependent variable.8 Using annual and quar terly data for the United States, Fisher found very long mean lags for the effect of price changes on long- and short-term interest rates. For example, the highest correlation between commercial paper rates and rates of change in the wholesale price index from 1915-27 was obtained when the latter was lagged over 120 quarters (30 years), implying a mean lag of about 40 quarters (10 years). 6Within the framework of equation (3 ), Fisher calculated the correlation coefficient corresponding to the following re gression equation: mt = P 2 i=l (n—i ) 5For a concise survey of the theoretical literature on adaptive expectations, see Zvi Griliches, “ Distributed Lags: A Survey,” Econometrica, January 1967, pp. 42-45. • —----- -Pt_i -f. rrt +ut n(n— 1 )/2 where n(n— 1 ) /2 is the sum of n terms ranging from zero to ( n— 1 ). 7His procedure was directly related to the present-day prac tice of choosing an estimated equation with the highest R2 ( coefficient of determination or square of the correlation ratio). 8The mean lag is simply the weighted-average lag, where the coefficients [wi’s in equations ( 2 ) and (3 )] are used for the weights. When all of the weights are positive, the for mula for the mean lag is as follows (Griliches, p. 31): n 2 i • wi 2 wi i=o n p T + d (rrt) The term within the large parentheses represents price ac celeration. See, inter alia, Allan H. Meltzer, “The Appro priate Indicators of Monetary Policy, Part I,” Savings and Residential Financing: 1969 Conference Proceedings (Chi cago: U.S. Savings and Loan League, 1969) p. 14. Page 20 DECEMBER 1 9 6 9 i=o that is, a weighted sum divided by the sum of the weights. In Fisher’s calculations, the denominator is unity (his weights necessarily sum to one), so the formula for his mean lags is 2 ( i i= l\ • n_i n(n l )/2 / which simplifies to (n—1 )/3 . Fisher estimated his mean lags as n/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 I S In recent years there has been a considerable re vival of interest in the study of Fisher effects, osten sibly the result of the reappearance of substantial variability in interest rates and price levels and me thodological developments in the estimation of dis tributed lags. Two studies have attempted to measure “real” rates directly and then to relate the spread between various nominal rates and the estimated “real” rates to historical time series for price level changes, with inconclusive results.9 Most of the published studies have regressed nom inal rates directly on current and past rates of price changes (or changes in nominal rates on price ac celerations).10 Data intervals have ranged from quarters (Gibson) to business cycle phases (Fried man and Schwartz). The time span has ranged from as early as 1873 to as late as 1966. Lagged rates of change in various price level indexes and even nom inal income (Gibson) have been tried as indicators of price expectations. The forms of the distributed lags estimated have generally been either “uncon strained” lags or “geometrically decaying” lags.11 Without exception, the mean lags of interest rates behind price changes were found to be very long. For example, Friedman and Schwartz found mean lags for short-term rates of about ten years and for 9Suraj B. Gupta, “Expected Rate of Change in Prices and Rates of Interest” (unpublished dissertation, University of Chicago, 1964), and Phillip Cagan, Determinants and Ef fects of Changes in the Stock of Money, 1875-1960 (New York: National Bureau of Economic Research, 1965), pp. 305-309. Gupta’s work is summarized and his empirical work extended in William E. Gibson, “Effects of Money on Interest Rates,” Staff Economic Studies, No. 43, Board of Governors of the Federal Reserve System, March 1968, pp. 45-48 and 88-89. Preliminary work along similar lines was reported in David Meiselman, “Bond Yields and the Price Level: The Gibson Paradox Regained,” in Deane Carson ( ed.), Banking and Monetary Studies ( Homewood, Illinois: R. D. Irwin, 1963), pp. 119-122. 10Meiselman, pp. 112-133; Milton Friedman and Anna Jacob son Schwartz, “Trends in Money, Income, and Prices, 18671966” (unpublished manuscript, National Bureau of Ecoconomic Research, November 1966), chapter 2, pp. 110-143; and Gibson, pp. 44-66 and supplementary tables. In their multiple regression study, Michael J. Hamburger and Wil liam L. Silber ( “An Empirical Study of Interest Rate De termination,” Review of Economics and Statistics, August 1969, pp. 369-373) rejected the rate of change in prices as insignificant. 11Both forms will be discussed later. To estimate uncon strained lags, one merely regresses the current value of the dependent variable on the current and a predetermined number of lagged values of the independent variable — there is thus no a priori constraint on the time shape of the coefficients. Geometrically decaying lags impose a geomet rical decay on the coefficients, that is, part of each coeffi cient is a constant decay term less than one which, when raised to higher powers as the lag recedes into the past, decays (asymptotically approaches zero). See Griliches, pp. 16-49, and Lawrence R. Klein, “The Estimation of Dis tributed Lags,” Econometrica, October 1958, pp. 553-565. DECEMBER 1 9 6 9 long-term rates of 25 to 30 years, which they at tributed to the “slow and gradual adjustment of anticipations of price changes to the actual behavior of prices.”12 A Search for Fisher Effects This study is based upon earlier work but departs from previous studies in ways that appear to have significant effects on the results, in particular: (1 ) Monthly, instead o f exclusively quarterly or annual, data are used for short-term and long-term interest rates (dependent variables) and for price level changes and other independent variables. Further, the interest rate series have been seasonally adjusted. (2 ) A variety o f kinds o f distributed lags are estimated, in order to investigate the effect o f lag form on the length of the lags. (3 ) The monthly data are aggregated into quarterly and annual series to determine the effect of aggre gation over time on the lag estimates. (4 ) The study is purposely confined to the period following the Treasury-Federal Reserve A ccord of 1951 in order to avoid having to contend with the constraint on interest rate movements imposed by the Federal Reserve’s “par pegging” o f Govern ment securities prices. Further, the 1952-69 period is divided into two sub-periods to see whether there has been any apparent change in the m ech anism relating past price changes to the formation of price expectations and any clues to the reasons for earlier findings of very long lags. (5 ) A model will also be tested to see what happens to the explanatory power o f past price level changes when variables assumed to affect “real” rates of interest are added to the regressions. ( 6 ) Experimental “ real” rate series will be generated and their movements compared with nominal rates to see whether there have been times when nomi nal rate movements might have been misleading indicators o f changes in “ real” interest costs. Seasonal Movements in Interest Rates A number of economists have observed not only seasonal movements in monthly and quarterly inter est rate series, but also the influence on the seasonal 12Friedman and Schwartz, chapter 2, p. 139. Gupta, esti mating geometrically distributed lags for the nominal rate — “real” rate spread behind price changes, found a mean lag of 16 years for long-term rates. Gibson estimated un constrained lags for relatively short lag intervals (ten quar ters and four years), so it is not possible to calculate mean lags for the total effect of price changes on interest rates. In Meiselman’s study, the geometric decay coefficients came out very close to one, implying a long mean lag (nearly twenty years, for example, with a decay coefficient of 0.967, which he found in regressing bond yields on price changes over the 1873-1960 period). Page 21 F E D E R A L R E S E R V E B A N K O F ST. LOUI S of changes in Federal Reserve operating strategy for open market purchases and sales.13 Since some of the data used for independent variables in the regres sions were seasonally adjusted, it was advisable to seasonally adjust the short-term and long-term inter est rate series, so that the results could be compared with those generated using unadjusted series.14 As expected, stronger seasonals were detected in the short-term than in the long-term interest rates. The finding of pronounced seasonals in both for the 1952GO period and the virtual elimination of seasonal move ments for the 1961-65 period, probably the conse quence of the Federal Reserve strategy to assist the balance of payments, confirms the conclusions of earlier studies. The resumption of pronounced sea sonals is apparent in the calculations for the 1966 to mid-1969 period. The explanation may lie in the insertion of “proviso clauses” in the Federal Open Market Committee directives over the later period and the implementation of such directives by the Trading Desk at the Federal Reserve Bank of New York.15 Empirical Results — Interest Rates Regressed on Rates of Price Change Data for the period 1952-69 were used to test the hypotheses about the effect of price expectations on the level of nominal interest rates. Several measures of both prices and interest rates were used in the estimation, and various lengths for the total lags were tested. In addition, several estimation techniques were employed. The results were very similar across the many combinations of data, length of the lag dis tribution, and estimation procedures, all suggesting a much shorter time horizon in formation of price expectations than had previously been found. The interest rates used in this study are yields on securities issued by the private sector.16 Short 13Leonall C. Andersen, “ Seasonal Movements in Financial Variables — Impact of Federal Reserve and Treasury,” Business and Government Review, University of Missouri, July-August 1965, pp. 19-26; “A Closer Look at InterestRate Relationships, The Morgan Guaranty Survey, April 1961, pp. 3-5; Gibson, pp. 30-32 and Tables 3 and 4; and Hamburger and Silber, pp. 370-371. 14Data have been seasonally adjusted using the X -ll Variant of the Census Method II Seasonal Adjustment Program, U. S. Department of Commerce, Bureau of the Census. Source: Board of Governors of the Federal Reserve System, Federal Reserve Bulletin. 15Jan Warren Duggar, “The Proviso Clause and Bank Credit Proxy” (unpublished manuscript, Federal Reserve Bank of New York, 1969). 16Since there are many factors in addition to price expecta tions that affect the level of interest rates, die dependent variable used in the regressions should be the one least 22 Digitized forPage FRASER DECEMBER 1 9 6 9 term interest rates ( rns) were approximately by the yield on four- to six-month commercial paper. The yield to maturity on Aaa-,rated corporate bonds was used as the measure of long-term interest rates ( rnL). Price expectations were approximated by the rate of change of the consumer price index for all items, p c 17 Using monthly data for the period January 1952 to September 1969, the function mt = ao -|- ajP't 4- a2Pct_ 1 -f- a3P\—2 + — + “ n + lP ° t -n was estimated first by least squares regression of rnt on current and lagged values of price changes for n = 24, 36 and 48 months. The coefficients of the regressions are presented in Chart I. These regressions were run with both seasonally adjusted and nonseasonally adjusted interest rate series, and in each case the results using seasonally adjusted data traced quite closely those using un adjusted data. The introduction of the seasonal factor decreased the unexplained variance (increased the adjusted coefficient of determination, R2) only slightly. Chart II presents the coefficients of the regressions: m s = t do + m L = t a0 + a i P ct + a 2P ct _ i + ----- + “ 2 5 P ct— 24 a i P ° t -)- a 2P ct_ i + ----- + a 2 5 P ct— 24 The coefficients using seasonally adjusted interest rates are quite similar to those using unadjusted data. influenced by those other factors. Yields on private securi ties were selected, instead of rates on Government debt, because they are more free of the direct influences of debtmanagement and monetary actions. However, Fukasawa obtained similar results using yields on Government securi ties. Greene found that price expectations were somewhat easier to identify using interest rates on private debt. 17Mortgage costs are included in the consumer price index and might contribute to some degree of spurious correlation between interest rates and price movements. Since mort gage interest rates tend to move with other nominal rates, using the consumer price index as the measure of price movements would tend to result in a positive bias in the observed relationship between interest rates and price move ments. To test for this effect the consumer price index was purged of mortgage rate effects. Data on the mortgage component of the CPI were available from the Bureau of Labor Statistics only for the period 1954-64. Thus, nominal interest rates were regressed on the rate of change of the CPI and the adjusted CPI.for this period only. The regres sions using this adjusted Pc series were still quite close to those using the index inclusive of mortgage costs. Gibson’s procedure of using changes in national income as a proxy for price expectations was also treated, using however, per sonal income, which is available on a monthly basis. The results, summarized in the appendix, were quite similar to those using the consumer price index. F E D E R A L R E S E R V E B A N K O F ST. LO U I S 0 t-4 t-8 t-12 t-16 DECEMBER 1 9 6 9 t-20 t-24 t-28 Months Since this close relationship was observed in all of the tests, only the results using unadjusted data will be explicitly considered. The regressions show that price movements ac counted for about 50 per cent of the variance in interest rates between 1952 and late 1969. The pat tern of coefficients is consistent with the adaptive expectations hypothesis, that is, they are generally declining. The presence of small negative coefficients in the “tails” of the distributions could be explained theoretically by the eventual domination of positive “extrapolative effects” by negative “regressive effects” (see page 32 below ). Although the t-test is suspect in dealing with a distributed lag regression,18 the coefficients tend to be small beyond t-24 months and generally insignificant. Increasing the length of the 18Multicollinearity ( correlation between independent variables) is a possible source of difficulty in estimation of this type of distributed-lag relationship. In the presence of multicollinearity, the ordinary least squares regression technique is unable to identify the exact parameter associated with each independent variable. See J. Johnston, Econometric Methods (New York: McGraw-Hill, 1963), pp. 201-207. t-32 t-36 t-40 t-44 t-48 t-52 lag from 24 to 48 months had little effect on the distribution of coefficients. The sum of the coefficients increased as the lag was extended, however, sug gesting that, although great weight in the formation of price expectations comes from quite recent ex perience, the total adjustment procedure is probably somewhat longer, with only relatively small weight given to price movements in the distant past. In other words, the “true” distribution probably has a “tail” of small declining coefficients. These results sug gest a much shorter time horizon information of price expectations than had been found in the investiga tions cited earlier. Due to multicollinearity, direct estimation of an unconstrained distributed-lag function tends to re sult in wildly fluctuating coefficients. In order to re duce this fluctuation, the relationships were estimated using the Almon lag technique.19 This procedure re sults in a much smoother distribution, which is more 19Shirley Almon, “The Distributed Lag Between Capital Appropriations and Expenditures,” Econometrica, January 1965, pp. 178-196. 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 1 9 6 9 C hart II Regression Results Using Se a so n ally Adjusted a n d Not S e a so n a lly Adjusted Interest Rates Coefficients consistent with the adaptive-expectations hypothesis, that is, expectations are a continuous function of past price movements. The Almon lag estimates are presented in Chart Iip o 'The distribution of the Almon coefficients fol lows the least squares estimate quite closely. For lags from 24 to 48 months, most of the effect on interest rates come from price movements over the previous year. The tails of coefficients beyond these points sum to nearly zero. The regression using 48 lags suggests that, if the annual rate of change of the consumer price index increased by one per cent in a given month (for 20The regressions presented here were generated using a sixth-degree polynomial. Other degree polynomials were tested and gave similar distributions. The sixth-degree was chosen because it best approximated the unconstrained estimates, in that it minimized the sum of the squares of the difference between the unconstrained and Almon es timates. The only constraint on the selection of the degree of the polynomial is that it must be less than or equal to the number of lagged coefficients. The sixth-degree poly nomial was the maximum which could be used in the pro gram available to the authors. Digitized forPage FRASER 24 Coefficients example, from a 3 per cent annual rate of increase to a 4 per cent annual rate) and prices continued to rise at that rate, the yields on four- to six-month com mercial paper would rise 72 basis points ( for example, from 4 per cent to 4.72 per cent) during the first year, if all other factors affecting interest rates were un changed. After 48 months, short-term interest rates would have risen by 69 basis points.21 n In the long run, the nominal rate of interest does not rise by the full amount of the change in price expectations. An increase in price expectations will increase the difference between the nominal and Wicksellian market rates. How ever, the change in price expectations will tend to lower the market rate. Assuming an equilibrium position with ex pected price changes equal to zero, then rnt=rmt. If price expectations increase by one per cent per year, after 4 years the nominal interest rate will rise by 69 basis points, thus (1 ) (2 ) mt+48 — rmt+48 = 1.00 mt+48 — rnt = 0.69 Since mt = rmt, equations (1 ) and (2 ) reduce to (3 ) rmt+48 — rmt = — 0.31 Thus the market interest rate falls by 31 basis points fol lowing the increase in price expectations. This result is consistent with findings of other investigators; for example, see Keith M. Carlson and Denis S. Karnosky, The Influence F E D E R A L R E S E R V E B A N K O F ST. LOUI S DECEMBER 1 9 6 9 C h a rt III S u m m a r y o f R e g r e s s io n R e su lts ( A lm o........ n L ag) __* . -......nSIT..— Coefficients .12 / .10 U - Short-Term Interest Rate (1952-1969) \ \ 48 Lags .08 Sum .06 .04 Constant R2 24 Lags .670 2.471 .607 36 Logs .771 2.277 .618 48 Lags .760 2.294 .617 .02 A ' -.02 ' ' -.04 Coe ficients .08 36 Lags Long-Term Interest Rate (1952-1969) X 48 Lags \ Sum Constant R2 24 Lags .457 3.403 .514 36 Lags .537 3.249 .531 48 Lags .549 3.222 .526 -.02 -.04 t-4 t-8 t-12 t-16 t-20 t-24 t-28 M onths of Fiscal and Monetary Actions on Aggregate Demand: A Quantitative Appraisal, Federal Reserve Bank of St. Louis, Working Paper No. 4, March 1969. In the aggregate de mand model developed there, an increase in expected prices ceteris paribus generates a Government budget surplus t-32 t-36 t-40 t-44 t-48 t-52 Fisher hypothesized that the time horizon in form ing price expectations was related to the term to which results in a decrease in the stock df wealth and re duces the “real” interest rate. The net result is an increase in nominal rates less than the increase in expected prices. Page 25 F E D E R A L R E S E R V E B A N K O F ST. LOUI S DECEMBER 1 9 6 9 C h a rt IV Quarters maturity of the instrument. Potential buyers and sell ers of long-term debt would be interested in how prices move over an extended period and would tend to look further into the past than would those people who were dealing in short-term instruments. Partici pants in the market for short-term securities are less likely to be concerned with long-term price move ments and might need less information in forming their expectations. The results in the present study are consistent with this idea. The long-term interest rate is relatively less re sponsive to changes in price expectations. Twelve months after the one per cent increase in prices, long-term rates would be 59 basis points higher than they were originally, as opposed to 72 basis points for short-term rates. The effect on long-term rates would be a total increase of 56 basis points after 48 months. Why Such Long Lags In Earlier Studies? — Three Hypotheses The present study has found mean lags for the effect of price level changes on both long-term and short-term interest rates of less than a year. In con trast, earlier studies yielded mean lags of anywhere Page 26 from seven to thirty years. It is important to try to explain this discrepancy and to defend the results presented here. The authors have explored three hypotheses that might reconcile the differences: (1) The “true” lags of interest rates behind price changes are short, so that biases arise in aggregat ing the interest rate and price change series over longer observation periods, which lead to system atic overestimates of the length of the lags.22 (2) The forms of the lags estimated in other studies, in contrast to the more flexible class of lags esti mated in this study, are biased toward yielding longer average lags. (3) Institutional changes have occurred over time in financial and real markets, with the result that price level changes have come to have prompter 22Griliches, pp. 45-46; Yair Mundlak, “ Aggregation Over Time in Distributed Lag Models,” International Economic Review, May 1961, pp. 154-163; and William R. Bryan, “ Bank Adjustments to Monetary Policy: Alternative Esti mates of the Lag,” American Economic Review, September 1967, pp. 855-864. Griliches summarizes the issue as fol lows: “ aggregation over time (e.g., from quarterly to annual data) will in general result in a misspecification of the model. It will also . . . cause us to overestimate the implied average lags.” 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 1 9 6 9 C h a rt V Years and larger effects on interest rates.23 T o put it differently, there has been considerable thinning o f the “molasses (long-lag) world,” particularly in the past decade. Aggregation of Data To test the first hypothesis, the monthly data for all of the interest rate series and the rate of change in the consumer price index were aggregated to quar terly and annual averages of monthly data for the 1952-69 period. Almon distributed lags over 16 quar ters with sixth-degree polynominals were estimated for the quarterly series. The results (see Chart IV) were virtually identical to the original monthly re gressions with 48 month lags and the same degree polynomials.24 23The post-war increase in the degree of financial “market perfection” and its consequent effect on interest rate flexi bility is the subject of James S. Duesenberry’s essay, “The Effect of Policy Instruments on Thrift Institutions,” in Savings and Residential Financing: 1969 Conference Pro ceedings, pp. 135-143. 24Fukasawa has run unconstrained lags extending back six quarters with quarterly data from IV/1951-IV/1968 for Treasury bill and bond rates regressed on the rate of change in the GNP deflator. His results are similar to those reported here. The quarterly regressions suggest that if the annual rate of change of prices increases by one per cent in any quarter and remains at the higher level the short term rate would rise by 84 basis points after 4 years. The long-term rate would rise by 66 basis points over the same period. Using the results of the monthly estimates, an increase by one per cent in the annual rate of change in prices, would yield an increase of 69 basis points in short-term rates and 56 basis points in long-term rates after four years. There were too few observations, given the length of the lags and the degree of the polynomials, to fit Almon lags to the annual observations, so only un constrained lags were estimated, ranging from one to five years (see Chart V ). For the unadjusted com mercial paper rate, the R2 was highest (0.709) with only the current rate of change in prices in the regres sion; in all cases, only the coefficient for the current price changes was significant. As might be expected, the R2 for the unadjusted corporate Aaa yield was highest (0.552) when the current and one year lagged price change term were included, although in every 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 I S DECEMBER 1 9 6 9 case only the coefficient on the current term was significant. The regressions using annual data gave somewhat larger total effects. If the rate of change of prices rises by one per cent per year, short-term rates would be 137 basis points higher after four years. Long-term rates would rise by 134 basis points. The discrepancy between this and earlier studies apparently cannot be explained on grounds of an aggregation bias in the latter, and the first hypothesis cannot be accepted. The reason probably lies in the fact that the adjustment of interest rates to price level changes is not so rapid that aggregation of monthly into quarterly and annual data leads to systematic overestimates of the underlying lags. Estimation Procedure The second hypothesis pertains to the nature of the lag distributions estimated in other studies. Since several of the studies have estimated geometrically decaying (K oyck) lags, the monthly data used in the earlier part of the present study were used to estimate such lags. The following regression was run for each of the yield series, using for Pt both the simple monthly rate of change and compounded annual rates of change:25 mt = Xmt-i + /3Pt + constant The decay coefficient X, presumably somewhere be tween zero and one, indicates the rate at which the weight of the past rates of price change declines backward in time ( that is, X = 1 means that the lagged terms never decay at all, while X = 0 means that only the current price change term has any effect). All of the initial regressions yielded decay coeffi cients slightly greater than one, which, taken at face value suggests that the lagged terms do not decay. ms = 1.012 ms t mL = 1.007 mL t + .057 P't — .024 R2 = .980 + .053 K R2 = .994 t— l — .020 t— 1 A danger in such estimates of the decay coefficients and the (3 parameter is that they are inconsistent, and 25This is the convenient form in which such lags are usually estimated. This equation may be expanded into the following: mt = X0/SPt + XijSPui + X2/3Pt-2 + . . . + Xoo/SPt-33 + constant or, more simply, OO • mt = /32 X‘P t_i + constant. Page 28 i= o the estimate of X is probably biased upward.26 Fol lowing procedures outlined by Griliches and by Goldberger, the decay coefficients were re-estimated, which reduced them by only a very small amount: ms = 1.005 ms + mL = 1.003 mL + .054 K t t t—i t—i .071 Pct — .024 — .020 Decay coefficients greater than one are clearly in consistent with the adaptive expectations hypothesis. It would not be unreasonable to expect decay co efficients only slightly less than one to result from tests using different sample periods or data defini tions than were used here. The monthly data were divided into two sub periods, 1952-60 and 1961-69, and separate estimates of the decay coefficients obtained. For the earlier period, the coefficients dropped below one, ranging (unadjusted for consistency) from 0.977 for commer cial paper rates to 0.996 for corporate Aaa yields. These results imply long mean lags for both interest rates, with longer lags for the long-term rates. The coefficients on the current rate of price change in the commercial paper rate regressions strangely became negative for the 1952-60 period.27 For 1961-69, the decay coefficients were nearly the same as for the entire 1952-69 period, that is, slightly greater than one, for which it is difficult to find any theoretical rationalization. To see what would happen to the decay coeffi cients, the monthly data were aggregated into quar terly data and the decay coefficients re-estimated for the 1952-69 period and for the subperiods mentioned above. All of the decay coefficients for the entire period declined, which would be expected if a monthly decay process were to be converted into an equivalent quarterly process, but the decay coefficients for short term rates fell to below one (0.968 for commercial paper rates, with a mean lag of 20 quarters or five years). For 1952-60 alone, all the coefficients were less than one, but the decay process was again nega tive for short-term rates. For 1961-69 the results were all plausible, and the decay coefficients were 26See Griliches, p. 41, and Arthur S. Goldberger, Econo metric Theory (New York: John Wiley and Sons, 1964), pp. 276-278, and Kenneth F. Wallis, “ Some Recent Develop ments in Applied Econometrics: Dynamic Models and Sim ultaneous Equation Systems,” Journal of Economic Litera ture, September 1969, pp. 774-775. 27This implies that the lagged price change effects are op posite in sign from those hypothesized; they could be inter preted as evidence for Sargent’s “regressive effects” of price changes on short-term rates (see p. 32 below). F E D E R A L R E S E R V E B A N K O F ST. LO U I S DECEMBER 1 9 6 9 all lower than corresponding coefficients for 1952-60. The effect of price level changes on commercial paper rates for the latter period decayed with a X of 0.834 (mean lag of about three quarters), while the decay factor was 0.919 (mean lag of seven quarters) for corporate Aaa yields.28 As was the case with the entire period, there was little difference in the total price expectations effect between different degree polynomials. The length of the lag distribution was crucial, however. The total effect on short-term interest rates tended to decline as the lag was extended beyond 24 months, and this was quite pronounced in the 1961-69 period. The effect on long-term rates, however, increased as the lag was extended up to 36 months. Beyond 36 months, the sum of the coefficients remained almost constant. None of the coefficients beyond 48 months were significant. These results suggest that the time horizon in forming price expectations increases as the term to maturity of the security increases. The preceding experiments with simple geometri cally declining lag structures suggest that such lags, requiring an exponential decay, may not be the most appropriate ones to impose on the interest rate and price level data for the period of this study in the attempt to capture the “true” underlying lag dis tribution. In every case the average lags obtained with this procedure were considerably longer than with either unconstrained or Almon lags, which pro vides some explanation for the differences between this and previous studies.29 The total price expectations effect is much larger in the 1961-69 period than in the earlier period. In the latter period the total effect on short-term rates is about 90 per cent of the annual rate of change in prices. The effect on long-term rates is about 80 per cent of the rate of price change. In the 1952-60 pe riod the sum of the coefficients range between 5 and 35 per cent of the price change for a lag of 36 months. Chart VI contains the lag coefficients for short-term rates ( second-degree polynomial and 24 lags for 1961-69, and sixth-degree with 36 lags for 1952-60) for the relationship between the commercial paper rate and the rate of change in the consumer price Institutional Changes The third hypothesis asserts that price level changes have come to have larger and prompter effects on interest rates because of institutional changes in the economy. As a preliminary test of this hypothesis, the 1952-69 period was again divided into two subperiods, 1952-60 and 1961-69, and various Almon lag struc tures estimated separately for each.30 Table I con tains the sum of the lag coefficients for 12 to 48 lags and second- to sixth-degree polynomials. 28James B. Greene, using quarterly data for 1961-68, obtained a decay coefficient of 0.824 for the commer cial paper rate regressed on the rate of change in the consumer price index, which implies a mean lag of about 2% quarters; for the corporate Aaa yield his decay co efficient was 0.919 implying a mean lag of about seven quarters. 29Experiments were also conducted for the whole period and the subperiods with simple second-order lags (in the regressions the dependent vari able was lagged one and two pe riods ). The results were not appreci ably different from those for the first-order lags. 30The “Chow test” was conducted to see whether there was a fundamen tal shift in behavior patterns within the 1952-1969 period. For both commercial paper rates and corpo rate Aaa yields the “F” statistics were significant at the one per cent level, which indicates a substantial differ ence in the anticipations forming mechanism in the two subperiods. For an explanation of the test, see Gregory C. Chow, “Tests of Equality between Sets of Coefficients in Two Linear Regressions,” Econometrica, July 1960, pp. 591-605. T ab le I SU M OF THE REGRESSION COEFFICIENTS* (monthly data) Short-term interest rates Length of Lag 1 9 5 2 -6 0 1 9 6 1 -6 9 D e g re e of P olyno m ial 12 24 36 2 .2 8 2 5 .2 2 6 5 .0 8 9 9 48 — .0 4 8 2 12 24 36 48 .921 1 .9 5 1 8 .8 0 3 9 .5 7 2 6 3 .2 7 6 0 .2 2 5 4 .3 0 5 5 .1 8 5 6 .9 1 0 5 .9 0 3 5 .7 2 3 5 .5 3 4 4 4 .2 8 3 7 .2 4 0 2 .341 8 .1 3 4 9 .91 18 .9 2 3 1 .7 3 7 3 .4 7 5 0 5 .2 8 3 6 .2 4 0 6 .3 3 7 8 .0 5 3 9 .9 1 3 4 .9 2 1 0 .7 1 2 4 .4 6 6 8 6 .2 8 3 4 .2 4 3 9 .3 3 4 0 .0 9 6 0 .9 1 3 1 .9 1 7 2 .7 1 8 0 .4 7 5 9 Long-term interest rates Length of Lag 1 9 5 2 -6 0 1 9 6 1 -6 9 D e gre e of P o lyn o m ia l 12 24 36 48 12 24 36 48 2 .1 4 3 2 .1 1 5 4 .0 5 3 7 .0 0 8 1 .5 8 5 4 .7 0 8 6 .8 4 0 6 .8 3 2 1 ” 3 .1 4 1 7 .1 1 2 2 .1 6 3 9 .1 5 8 0 .5 8 8 1 .7 4 0 5 .8 6 1 8 .7 9 4 5 4 .1 4 4 5 .1 1 4 0 .2 0 7 8 .1 3 7 4 .5 8 8 6 .7 5 3 1 .8 5 4 0 .8 2 1 0 5 .1 4 4 5 .1 1 9 3 .2 0 6 2 .0 7 9 8 .5 8 8 9 .7 5 3 3 .8 4 7 4 .8 2 2 7 6 .1 4 4 4 .1 2 0 0 .2 0 2 3 .0 7 5 6 .5 8 8 6 .7 5 2 4 .8 5 2 6 .8 3 0 3 ♦ A ll c o e ffic ie n ts in th e ta b le f o r th e 1961-69 p e r io d a r e s ig n ific a n t a t th e o n e p e r c e n t le v e l. T h o se f o r th e 1952-60 p e r io d w e re in s ig n ific a n t f o r 48 la gs. 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 I S DECEMBER 1 9 6 9 Chart Vt Sum m ary of Regression Results Coefficients Months index. The sum of the coefficients for the earlier pe riod was .344 and the mean lag, 1 to 2 months. For the latter period, the sum was .952, and the mean lag 4 to 5 months. While it is true that the smaller effect in the earlier period was exhausted more quickly (the mean lag was shorter), the peak in total effect for the earlier period was reached after eleven months, while the same level of effect was attained in the Page 30 latter period in only 2 to jumps from 0.255 to 0.901, rate of change in prices cent of the variation in 3 months. Further, the R2 so for the latter period the accounts for over 90 per commercial paper rates.31 31The highest R 2 (0.938) for the commercial paper rate was obtained for 1961-69 using sixth-degree polynomials and 48 month lags. 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 These results are thus consistent with the hypothesis of the effect of institutional changes.3Similar results were obtained for the corporate Aaa yield (Chart V I), and the jump in R- for the latter period is even more pronounced, from 0.164 to 0.973. The coefficients of the long-term rate were generated using a second-degree polynomial and 48 lags for 1961-69 and sixth-degree with 36 lags for 1952-60. All of the coefficients estimated for the 1961-69 period are significant at the one per cent level. A mean lag of 16 months is implied by this result, meaning more than half of the adjustments in interest rates to price changes in the period were attained in less than a year and a half. A summary of the 1961-69 regressions appears in the appendix. What factors might cause a shift in the framework for transmitting past price level changes, via price expectations, to nominal interest rates? A listing of plausible explanations might include the following: (1 ) According to Friedman and Schwartz, “ the period used in forming anticipations should depend on the characteristics o f price behavior,” particularly the “variability in the behavior of the general level of prices .”33 Thus, one could argue that prices have been more variable, at least in an up ward direction, in the 1950’s and 1960’s than over long, earlier historical periods. Further, the greater publicity given to price level movements, as well as the more rapid processing of data, could convey greater awareness of recent price level behavior and affect price level expectations and interest rates more substantially than once was the case. (2 ) Nominal rates may have come to reflect past price level changes more fully both because of a de crease in “money illusion” and because of de creased effects of price changes on real wealth over time .34 The former could be explained by the increased importance of large institutional in vestors in markets such as that for corporate bonds. For the latter to be a contributory factor, real wealth would have to be affected relatively less than before by price changes (because assets not 32Fukasawa has estimated unconstrained lags with quarterly data for five subperiods from 1951-68 and has obtained similar results. 33Friedman and Schwartz, p. 143. 34“ Money illusion” means that behavior is based on and di rected toward nominal magnitudes rather than “real” mag nitudes, for example, investment outlay in money terms would be related to money income and nominal interest rates. If real wealth influences the decisions of savers, the saving function would not shift upward by the expected rate of increase in prices because of expected decreases in the real value of assets fixed in nominal terms ( for ex ample, money), which dampen the effect of price expecta tions on nominal rates (see Robert Mundell, “ Inflation and Real Interest,” Journal of Political Economy, June 1963, pp. 280-283). DECEMBER 1 9 6 9 fixed in nominal terms may have becom e rela tively more im portant), thus reducing the “ drag” on upward shifts in the saving function by the amount of expected price level changes. (3 ) Interest rates are more flexible than in many past periods. According to Duesenberry, “ Restrictive monetary policy has in the past operated to a large extent through [nonprice] rationing . . . . Market forces and public policy have been working toward perfecting capital markets, and thereby reducing the effectiveness of ration ing . . . [and resulting in] a world requiring wide swings in interest rates for stabilization purposes . . ,”35 Thus, one would expect to find larger coefficients linking price changes to interest rates than in the past. (4 ) The frame o f reference for forming expectations may well have changed, particularly in the 1960’s. The relative absence of cycles in prices except for the very distant past deprives individuals of a suc cession of comparable reference points from which to extrapolate into the future and forces the use of heavier weights on the more recent past. Price Expectations In An Expanded Model A recent study by Thomas J. Sargent3" differs from earlier studies of the effect of price expectations on interest rates in two important respects. Besides relating past price changes to nominal interest rates, he sought also to decompose the “real” rate into components representing the equilibrium “real” rate and the deviation of current “real” market rates from the equilibrium rate. In addition, the shapes of the distributed lags he estimated were more general, that is, capable of fitting the data into a greater variety of geometrical configurations. Sargent devised a useful identity:37 (a) (b ) mt = ret + (rmt — ret) V ----------------- V-----------------' “Real” rate ( rmt = m ) + (c ) (mt — rmt) v------------ y------------ ' Fisher effect ( ) where r e t is the rate of interest at which real saving and investment would be in long-run equilibrium and rm t is the current market level of the “real” rate, that is, rm t is the same as rrt in equation ( 1 ) above. Movements in the nominal rate may then be attrib uted to changes in the equilibrium rate (a), to a deviation between the equilibrium rate and the “real” market rate ( b ) , and to a Fisher effect ( c ) . 35Duesenberry, pp. 136 and 140. 3eThomas J. Sargent, “Commodity Price Expectations and the Interest Rate,” Quarterly Journal of Economics, Feb ruary 1969, pp. 127-140. 37Sargent, p. 130. It is an identity, since it simplifies to mt = mt. Page 31 DECEMBER 1 9 6 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 I S Earlier studies (including that which we have re ported above) regarded either ( b ) or ( a + b ) as a residual and regressed nominal interest rates on past price changes only, but Sargent attempted to estimate each of the components of the level of nominal rates. The relationships among the components of the nom inal rate and how he sought to identify them statis tically are shown in Figure I. Assume that real investment (I/P) and real saving ( S / P ) are func tions of real income and “real” market rates of interest and that real income is given (so shifts in the saving and investment schedules do not have to be accounted for). The equilibrium “real” interest rate (re) is the rate at which real saving and invest ment would be in equilibrium. The market rate (rm) below the equilibrium rate indicates that some por tion ( A B ) of current investment is being financed from sources other than intended saving, for example, by newly created money from the banking system or through the drawing down of previously accumulated money balances. This is sometimes called the “Wicksell effect” on interest rates.38 Assuming savers and investors form the same price expectations and that neither are subject to “money illusion” (an important Fisherian concept39), both functions would be shifted upward by the expected rate of change in prices, P\ = rnt — rmt. Since the equilibrium rate cannot be directly ob served, Sargent used a reduced form proxy for it. He solved his real saving and investment functions simul taneously, so that the one market rate consistent with equilibrium (equality of intended savings and in vestment) is a function of the other determinants of real saving and investment, namely real income and (from an investment accelerator) the change in real 38Knut Wicksell, Lectures on Political Economy, pp. 190-198, and Interest and Prices (London: Macmillan, 1936). Wick sell assumed that savers and investors expected current prices to continue into the future, so he did not need to account for price expectations effects on interest rates. As Sargent points out, views similar to Wicksell’s were also held by Henry Thornton in 1802 and by Keynes in A Treatise on Money. Emphasis on the equilibrium rate-market rate relationship as the proper one in using interest rates as monetary policy indicators and rejection of price expectations effects on empirical grounds characterizes re cent work of Patric H. Hendershott and George Horwich (see, for example, “The Appropriate Indicators of Mone tary Policy, Part II,” in Savings and Residential Financing: 1969 Conference Proceedings, pp. 42-44). What is here called the “Wicksell effect” may also be interpreted as the "liquidity effect” or “impact effect” of changes in the money stock; similarly, the real GNP variables reflect the “income effect” or “feedback effect” on interest rates associated with changes in the money stock (see refer ences to works by Friedman and Schwartz, Gibson, and Meltzer cited above). 39See footnote 34 above. Page 32 Figu re I Illustration of Sargent's Identity interest income. This solution was then used to measure com ponent (a ) in the equations he estimated. Similarly, having no independent observations for the market rate, he used another proxy, the current rate of change in the “real” (deflated) money stock, for the .devia tion of the market rate from equilibrium.40 Finally, Sargent estimated geometrically distrib uted lags on past price changes as a proxy for price expectations. Using annual data for 1902-40 (tw o of the regressions were also run for 1902-54) and taking for nominal rates Durand’s one-year and ten-year basic yields, he obtained estimates implying very long mean lags (twenty years or more for short- and long-term rates). In several of his regressions he estimated two sets of decay coefficients. Both were positive for the long term rate; for the short-term rate one was negative and more quickly decaying, which Sargent ration alized as indicative of a “regressive effect” of price changes on short-term rates (as opposed to the posi tive “extrapolative effect” ), that is, price changes temporarily generate expectations of changes in the opposite direction (that is, that they will move back to a “normal” level). The sum of the regressive and 40In Figure I the gap between the equilibrium and mar ket interest rates will widen as the portion of real invest ment not financed by current real savings (A B) increases. The rate of change in the real money stock, on the other hand, should be positively correlated with the magnitude of AB. As a proxy for ( rm-re) the rate of money change should have a negative coefficient (that is, be positively related to an (re-rm) gap). The entire reduced form for “ real” rates should also capture the effects of other capital market disturbances, for example, Government surpluses or deficits and the ways they are financed (banking system versus nonbank public). 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 extrapolative weights did not reach a peak until eight years and declined even more slowly thereafter (since the negative component decayed more rap idly), so the mean lag would not be much different from his other results. The authors have subjected Sargent’s basic approach to a further test, with the following modifications: (1 ) Both real saving and investment are assumed to depend on both real GNP and “ real” market rates of interest. Thus, there is no a priori expectations as to the sign of the coefficient for the real GNP term in the regressions. A negative coefficient would presumably indicate that shifts in the sav ing function in response to a change in real GNP outweighed shifts in investment, so that nominal and “real” rates would tend to fall as real GNP rose .41 A positive coefficient would suggest just the opposite, while a coefficient near zero might indicate roughly offsetting effects of saving and investment shifts. (2 ) Quarterly and monthly instead o f annual data are used, and as before, the emphasis is completely on the entire post-accord period and the 1961-1969 subperiod. The regressions with monthly data necessarily use proxies for real GNP (personal income deflated by the consumer price index and, alternatively, the index of industrial production) and the GNP price deflator ( consumer price in d ex ). (3 ) The interest rate series and distributed lag forms are different; further, regressions were run with and without a constant term (Sargent did not suppress the constant term in any of his regres sions ). The equations estimated are of the following form: mt = ao + 2 arhPt-i + ftY* + ft>AY* + /S3AM“ i=o t t t where P is the annual rate of change in the GNP deflator (or a monthly proxy), Y° and AY° are the level and rate of change in real GNP (or a monthly proxy), and AM * is the average change in the real money stock (nominal money stock deflated by the GNP deflator or its monthly proxy). Nominal rates (rn) are again the four- to six-month commerical paper rate (rns) and the corporate Aaa yield ( m L), using quarterly averages of monthly data in the quarterly regressions. Only results for the 1961-69 subperiod will be reported here, in Chart VII, and in the appendix. The explanatory power of price level changes was changed little when the equations were more fully 4'Sargent obtained negative coefficients in all of his regres sions. In his theoretical model he assumed that only saving was functionally related to the level of real GNP. DECEMBER 1 9 6 9 specified, and the adjusted RJ’s rose by small amounts. For example in the equations for the long-term rate with second-degree Almon polynomials, total lags of 16 quarters (best statistical fit), and a constant term, the sum of the coefficients on current and lagged rates of price change actually rose from 0.80 to 0.86, and the R2 was unchanged at 0.977 when the current real GNP and real change in the money supply were added to the regression; further, the mean lag for the effect of price changes on nominal rates increased from 3.2 quarters to 5.5 quarters. In other words, recent price changes alone tend to overstate the necessary adjustment of nominal rates to account for the Fisher effect. As would be expected, the coeffi cients on current and lagged rates of price change were redistributed toward the past in the expanded equations, since the current and last quarters’ price levels implicitly enter into the other independent variables.42 Suppressing the constant term in the equation (that is, forcing “o to zero) forces a redistribution of its effects over the other coefficients. In the case of the long-term rate, the constant was not significant, and suppressing it enhanced slightly the .explanatory power of real GNP and the change in the real money supply, lowered the sum of the price change coeffi cients (to 0.80) and the acceleration coefficient ( (3a), left the R2 virtually unchanged, and lengthened the mean lag (b y three quarters with total lags o f 16 quarters). In the case of the short-term rate, the mean lag rose from zero to nearly one quarter. Otherwise, the effects on the coefficients were exactly opposite to what happened when the constant was suppressed in the equation for the long-term rate. Since the expanded equations contain variables not all measured in the same units, “beta” coefficients were computed in order to assess the relative con tribution of each independent variable to the de termination of nominal interest rates. In the equation for the long-term rate with various lag lengths the “beta” coefficient for price level changes is nearly three times as large as for real GNP, which ranks second in importance. 42The equation was also estimated for various lengths of total lags without the current rate of price level change (all of Sargent’s regressions were of this form) to try to reduce multicollinearity. With total lags of 16 quarters, the sum of the coefficients on the lagged price changes 2 a i+l I rose slightly, ft. and /32 remained about the i=° / same, £3 declined in absolute value by about 10 per cent, and the R2 and Durbin-Watson statistics rose slightly. Page 33 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 1 9 6 9 C h a rt V II Estimated Long-Term R eal Interest Rates* PerCent PerCent coefficients were smaller than in the quarterly regressions, suggesting that the response of the equilibrium “real” rate of interest to real income growth may occur over a substantially longer period than the current month.44 Thus, the findings reported in this section appear to support the specifica tion of variables in Sargent’s model. The use of quarterly and monthly data over post-accord period and the esti mation of Almon lags provide better statistical results than in his study. The importance of price level changes in explaining nominal interest rates is diluted very little by the expanded equations, and the mean lags are not sufficiently lengthened to alter the con clusions of the earlier sections of the present study. Experimental Time Series For The “Real” Rate of Interest Y ie ld s o n c o rp o ra te A a a b o n d s Latest d a t a plotted: O c t o b e r The Federal Reserve Bank of St. Louis began calculating and publishing an experimental monthly series for the expected “real” rate of return on Cor porate Aaa bonds in 1966.45 The procedure em ployed was to subtract from the actual Aaa yield a simple average of rates of change in the implicit GNP price deflator for the previous twelve months (quarterly price deflator data were interpolated to obtain an estimated monthly index). Such a proce dure necessarily implies that the mean lag is half as long (six months) as the total lag and that the co- Sargent’s expanded model was also tested with monthly data, using alternatively, personal income deflated by the consumer price index and the index of industrial production as proxies for the real GNP (a series derived from the regression using the latter appears in Chart VII as “real” rate 3 ). The results closely paralleled those for the equations using quar terly data. For example, in the equation for the long term rate with a total lag of 48 months, the index of industrial production as the real GNP proxy, and a constant term (which was significant in the monthly regressions), the sum of the price change coefficients fell from 0.87 to 0.82, the mean lag rose slightly from 15.6 to 16.4 months, and the R2 went from 0.968 to 0.971.43 The change in industrial production and the change in the real money supply had the correct signs but were not significant; one month is probably too short a period to capture the full “Wicksell (liquidity) effect.” The level of industrial pro duction turned out to be quite significant, but the 44It should also be noted that there is another possible source of mis-specification in all of the expanded equations, namely, the interrelationship between changes in the nominal money stock and both price levels and rates of change. In other words, the monetary authorities would be expected to re spond to departures from stable prices. One way around this problem is to make the policy variable endogenous in a simultaneously estimated model containing a “reaction func tion” for the Federal Reserve ( see Michael W . Keran and Christopher T. Babb, “An Explanation of Federal Reserve Actions (1933-68),” this Review, July 1969, pp. 7-20; and Raymond G. Torto, “An Endogenous Treatment of the Federal Reserve System in a Macro-Econometric Model,” unpublished dissertation, Boston College, 1969. 43Results using personal income deflated by the consumer price index were virtually identical to those using the index of industrial production as the proxy for real GNP. 45“ Strong Total Demand, Rising Interest Rates, and Contin ued Availability of Credit,” this Review, August 1966, pp. 3 and 4. Page 34 FEDERAL. R E S E R V E B A N K O F ST. L O U I S efficients are constrained to sum to one.48 Shortly afterward the lag for averaging price changes was extended to 24 months (mean lag of 12 months), and the resulting proxy for the “real” rate has been reported periodically ever since. As a testimonial to the intuition of the series’ creator, the distributed lag results in the present study yield estimates of the magnitude of effect and the mean lag which are remarkably close to the original “real” rate series. Chart VII contains the nominal corporate Aaa yield from 1960 to October 1969 and various estimated monthly “real” rate series. “Real” rate 1 is the original series, that is, the nominal rate minus the average of rates of price change over the preceding two years. “Real” rate 2 is obtained from the regression using monthly data for 1961-69, total lags of 48 months, and second-degree polynomials. “Real” rate 3 is de rived from the regression reported in the preceding section, which seeks to explain the contribution of “real” rates, as well as price expectations to nominal rates of interest; “real” rates here are assumed to be related to the level o f and changes in the index of industrial production and changes in the deflated money stock. Detailed analysis of the movements in these series will require a separate study.47 Only a few observa tions will be made here. The pattern of movement in all three “real” rate series is remarkably similar. The old “real” rate 1, however, appears to have over stated the price expectations component of the nom inal rate over most of the period. What is of particular interest are the occasions when changes in nominal rates gave apparently false signals about the nature of changes in “real” rates and the extent o f agreement about directions of movement among the three “real” rate series. All three “real” rates indicated that credit condi tions were progressively tighter during the first half of 1961, when the nominal rate was virtually un changed. The nominal rate was a reasonably good proxy for “real” rates 2 and 3 during 1962 but not for “real” rate 1, which rose for most of the year (the consequence of heavier implicit weights than the other two series on price changes two years n 1 . 48Mathematically, Pe = 2 — Pt_j, where n is the length 1 i= 1 n of the total lag, and there are exactly n coefficients, each of which equals 1/n (hence, the sum is n *1/n = 1). Moving averages with equal weights are discussed by Griliches, p. 25. 47A variety of other monthly and quarterly “real” rate series have been computed, including short-term “real” rates. DECEMBER 1 9 6 9 earlier and lighter weights on the past year). The gradual upward creep in prices from 1963-65 caused “real” rate 1 to creep smoothly downward, generally opposite in direction to the nominal rate. With the different pattern of weights, movements in the “real” rates 2 and 3 were more pronounced, indicating that underlying price level changes were not entirely smooth over the interval. “Real” rates 1 and 2 fell and “real” rate 3 oscil lated around a constant level during the first half of 1966, while the nominal rate rose. From late 1966 until early 1967, all rates moved down in step. From 1967-69, the original “real” rate 1 tended to drift downward and oscillate somewhat ambiguously, al though the three “real” rate series fell before nominal rates declined in the summer of 1968. “Real” rates 2 and 3 moved upward with the nom inal rate from late 1968 until early 1969. For several months thereafter, nominal rates did not rise by enough to offset the effects of rapid inflation, with the consequence that the monthly “real” rates actually fell from about February until late in the summer. Such movement in “real” rates could be used to ex plain, in part, the strength of the 1969 surge in in vestment spending. Conclusions Citing the findings by Gibson and Sargent of long lags in the forming of price expectations, Hendershott and Horwich recently argued: . . . Their experience contradicts the monetary voices in government, industry and the acad emy that proclaim, but do not demonstrate, that price level expectations, rather than real forces, are largely responsible for interest rate move ments in this decade .48 In contrast, the present study has shown that, unlike the earlier historical periods on which most of the previous studies have been based, price level changes since 1952 have evidently come to have a prompt and substantial effect on price expectations and nom inal interest rates. In addition, the total effect of price expectations on interest rates and the speed at which they are formed appear to have increased greatly since 1960. This conclusion is invariant to the form or the term of the flexible classes of distributed 48Hendershott and Horwich, “Appropriate Indicator,” p. 44. Criticizing the earlier “ St. Louis ‘ real” rate,” they con tinue, “The Fisherian zeal of that institution would shock no one more than Irving Fisher, who himself stressed the fantastically long lags in the formation of price level ex pectations and their impact on interest rates in this country.” Page 35 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 lags estimated. Most significant is the finding that price level changes, rather than “real” rates, account for nearly all the variation in nominal interest rates since 1961. Furthermore, the addition of variables to the regressions to account explicitly for the “real” rate components of nominal rates does not apprecia bly alter these findings. The causes of price level changes over the period of the study have not been investigated. The pri mary concern has been to determine the extent to which nominal rate movements may be attributed to expectations about future rates of change in prices, so that nominal rates may consequently be adjusted to yield information about movements in underlying “real” rates.49 The failure to make such an adjust ment and the sole use of changes in nominal rates as indicators of monetary ease or tightness may on occasion give misleading information about the direc49An interesting attempt to “ neutralize” interest rates with respect to the impact of movements toward or away from full employment was reported in Dennis R. Starleaf and James A. Stephenson, “A Suggested Solution to the Mone tary Policy Indicator Problem: The Monetary Full Employ ment Interest Rate,” Journal of Finance, September 1969, pp. 623-641. Unfortunately, the authors did not incorporate price level changes into their analysis, which is a serious deficiency in their work. Page 36 DECEMBER 1 9 6 9 tion and the extent of movements in “real” rates. The importance of the Fisher effect to the controversy over appropriate monetary policy indicators has been succinctly stated by David Fand: . . . As w e get closer to a world o f high employment, and especially if interest rates and prices are both rising, the money stock may be a better (less misleading) indicator or target variable than [nominal] interest rates. Para doxically, the current tendency to emphasize interest rates and to ignore changes in the money stock would seem more relevant to a society where interest rates and prices are falling while the money stock is constant, or rising at a lower rate than output .50 According to economic theory, changes in “real” rates should then reflect both shifts in the equilibrium relationship between real saving and investment and current capital market disequilibrium. Further, it is such “real” rate series that should be employed in studies of the term structure of interest rates and of the effects of international interest rate differentials on short- and long-term capital flows. 50David Fand, “ Keynesian Monetary Theories, Stabilization Policy, and the Recent Inflation,” Journal of Money, Credit and Banking, August 1969, p. 576. This article is available as Reprint No. 49. The Appendix to this article begins on the next page. F E D E R A L R E S E R V E B A N K O F ST. LOUI S DECEMBER 1 9 6 9 APPENDIX Nominal Income As a Proxy for Prices Gibson suggested that movements in nominal income might serve as a measure o f price behavior. His reason ing was that the CPI might be unable to accurately measure short-term price movements since it is a selec tive index of prices. Nominal income, on the other hand, contains an implicit general price index. Regressions using current and lagged monthly rates of change of nominal personal income (Py) were run and the patterns of coefficients are similar to those resulting from the runs using the CPI. To adjust for the difference in magnitude between the income and price index series, “beta coefficients” were computed in Table 1 below .1 T a b le I SU M O F THE BETA COEFFICIENTS (1952-69) Short-term rates • py Long-term rates . • py Pc • pc any deviations due to short-run changes in financial markets. The series 2 was generated from the following relationship: ( 1 ) m t = “ o + a l P ct + “ 2 P ct — 1 + . . . . + “ 4 9 P ct — 4 8 T a b le II REG RESSION COEFFICIENTS L m t = “o + • a l P ct + (J a n u a r y • “ 2P ct — 1 + .... + • a 4o P ct — 18 1 9 6 1 - Sep tem b er 1 9 6 9 ) a. i t-value i t-va lue i 1 .0 3 5 3 1 2 .0 8 5 8 26 .0 1 5 7 9 .7 3 5 8 2 .0 3 4 5 1 3 .1 2 0 1 27 .0 1 5 0 9 .0 3 6 0 3 .0 3 3 6 1 4 .3 5 2 1 28 .0 1 4 3 8 .4 2 4 6 4 .0 3 7 9 1 5 .8 4 2 1 29 .0 1 3 6 7 .8 8 6 0 i 5 .0 3 2 0 1 7 .6 7 6 5 30 .0 1 2 9 7 .4 0 8 1 6 .0 3 1 1 1 9 .9 8 2 2 31 .0 1 2 2 6 .9812 7 .0 3 0 3 2 2 .9 4 9 8 32 .0 1 1 5 6 .5 9 7 7 8 .0 2 9 5 2 6 .8 7 0 0 33 .0 1 0 8 6 .2512 9 .0 2 8 7 3 2 .1 7 8 6 34 .0 1 0 2 5 .9 3 6 8 5 .6 5 0 1 24 la g s 2 .0 2 0 1.301 1 .5 0 2 1 .2 5 9 36 la g s 2 .0 9 0 1 .5 2 6 1 .6 5 9 1 .5 3 2 10 .0 2 7 9 3 9 .4 3 8 9 35 .0 0 9 5 48 la g s 2 .4 8 9 1 .581 2 .2 3 0 1 .7 6 7 11 .0 2 7 1 4 8 .8 6 9 7 36 .0 0 8 8 5 .3 8 7 6 12 .0 2 6 3 5 8 .1 8 0 7 37 .0 0 8 2 5 .1465 13 .0 2 5 5 6 0 .2 6 7 8 38 .0 0 7 5 4 .9 2 4 3 14 .0 2 4 7 5 2 .6 5 0 0 39 .0 0 6 8 4 .7 1 8 7 15 .0 2 4 0 4 2 .4 8 1 2 40 .0 0 6 2 4 .5 2 8 1 4 .3 5 0 8 The expectational effects of prices on interest rates, as indicated by movements in nominal income, are larger than those suggested by movements in the CPI. In ad dition, use of nominal income results in somewhat longer lags, but almost all of the effect still occurs within two years. The Real Rate of Interest Series The “ real” interest rates (series 2 and 3 ) presented in Chart VII are actually the Wicksellian “market” rate L (rmt) or the long-run equilibrium interest rate plus 'Arthur S. Goldberger, Econometric Theory (New York: John Wiley & Sons, 1964), pp. 197-198. 16 .0 2 3 2 3 4 .1 4 1 5 41 .0 0 5 6 17 .0 2 2 4 2 8 .0 1 0 2 42 .0 0 4 9 4 .1855 18 .0 2 1 7 2 3 .5 2 6 2 43 .0 0 4 3 4 .0 3 1 1 19 .0 2 0 9 2 0 .1 7 4 2 44 .0 0 3 7 3 .8864 20 .0 2 0 1 1 7 .6 0 0 6 45 .0 0 3 0 3 .7506 21 .0 1 9 4 1 5 .5 7 4 3 46 .0 0 2 4 3 .6229 22 .0 1 8 7 1 3 .9 4 3 1 47 .0 0 1 8 3 .5 0 2 7 3 .3 8 9 2 23 .0 1 7 9 1 2 .6 0 4 8 48 .0 0 1 2 24 .0 1 7 2 1 1 .4 8 8 6 49 .0 0 0 6 3 .2 8 1 9 25 .0 1 6 5 1 0 .5 4 4 3 C o n sta n t 3 . 3 1 0 2 7 5 .0 6 3 5 R2 Page 37 F E D E R A L R E S E R V E B A N K O F ST. LOUI S T a b le DECEMBER 1 9 6 9 Using Sargent’s model, real output ( Y ° ), the change in real output ( A Y * ) , and the change in real balances ( A M ° ) were added to the function, and the following was estimated: III REGRESSION COEFFICIENTS rn J = “ o a i 1 a Pc + + (Ja n u a ry z " t— 1 + 1 9 6 1 -S e p t e m b e r a • pc 25' t — 24 m. = 181Y 0 + ft>AY* + ftiAM* + 2 “ i+ iP t—i t t t i=o 1969) a i .1 0 0 4 t-value i 8 .5 1 9 6 14 .0 2 7 0 i 5 .0272 t-value 2 .0 9 3 2 9 .5 9 5 1 15 .0 2 3 2 4 .1 0 4 4 3 .0 8 6 1 1 1 .0 7 3 1 16 .0 1 9 7 3 .3 9 3 2 4 .0 7 9 4 1 3 .2 0 8 0 17 .0 1 6 6 2 .8 2 9 1 5 .0 7 2 9 1 6 .4 7 5 1 18 .0 1 3 6 2 .3 7 1 2 6 .0 6 6 7 2 1 .6 7 0 9 19 .0 1 1 0 1 .9 9 2 3 7 .0 6 0 8 2 8 .6 4 7 7 20 .0 0 8 6 1 .6 7 3 7 8 .0 5 5 2 2 9 .5 4 6 7 21 .0 0 6 5 1 .4 0 2 3 9 .0 4 9 8 2 1 .5 1 0 8 22 .0 0 4 6 1 .1 6 8 2 10 .0 4 4 7 1 4 .7 5 5 7 23 .0 0 3 1 0 .9 6 4 3 11 .0 3 9 8 1 0 .6 1 7 3 24 .0 0 1 8 0 .7 8 5 2 12 .0 3 5 3 8 .0147 25 .0 0 0 8 0 .6 2 6 6 13 .0 3 1 0 6 .2 6 8 2 C o n sta n t 2 .4 1 1 1 2 5 .8 9 5 7 r 2 = The first two terms on the right determine the equilib rium interest rate (r e t ) . The real-balances term yields the deviation o f real market rates from equilibrium, and the last term captures the price expectation effect. Quarterly data were used, and real GNP served to measure real output. The annual rate of change of the GNP deflator was used to measure price movements and also to deflate the quarterly changes in the stock of money. Table IV compares these results with the earlier regressions which included only prices as arguments. The first set of regressions in Table IV apparently has not overstated the total effect of price movements on long-term interest rates, as the sums o f the “ j co efficients are unchanged in the more completely speci fied functions. Current and most recent price changes apparently captured some of the effect of contem po raneous output and changes in real balances, however, since the mean lag is more than twice as long in the second set o f regressions. .9006 which was estimated by the Almon lag technique, using data for the period January 1961 to September 1969. The series presented is the difference between the actual Aaa yield in each month and the cumulated effect of past prices, or (2 ) L L 49 2 rmt = mt — i=l There appears to be some merit in the more exten sive specifications. However, inclusion o f the additional variables did not drastically alter the conclusions of the original regressions. The mean lag, while longer, is still much shorter than other studies have found. In addition, the real market rates implied by each set of regressions are very similar and suggest that price expectations account for a great deal of the movement in nominal rates since 1961. • a i+ iP ct-i The “ i are the estimated coefficients in equation (1 ) above and Pct is the annual rate o f change of the con sumer price index: Table < » - = ( § b ) “ IV - SU M M AR Y O F R EG RESSIO N S Notice that the constant ( ao) in equation 1 does not appear in equation 2. The esti mated coefficients are presented in Table II. Estimated coefficients for the short-term interest rate are presented in Table III. ( Q u a rt e r ly d a ta , 1 9 6 1 - 6 9 ) rn L Page 38 + “ o “ iP t "t" °2 P t— 1 a n + lP t-n n Further Tests of Sargent’s Results The regressions presented so far at tempted to measure the effect o f price ex pectations by regressing nominal interest rates on current and past prices only. Thus the effects of all other factors affecting nominal rates were averaged into the con stant term. This approach carries with it the danger of misinterpretation, if the excluded factors affect both the dependent and in cluded independent variables. T o test for this possibility, Sargent’s approach of ex plicitly considering some of these other factors was applied. = t rt Z a + i= 0 m ean la g <* 16 .8 0 3 1 3.1 3 .3 5 3 9 20 .7 9 8 8 3 .0 3 .3 5 8 9 .9 7 6 7 24 .7 9 3 4 3.1 3 .3762 .9 7 6 4 r n L z — P iY * t R2 -f- (32A Y * t -) - p 3A M t -J- ° 1Pt - j - a2Pt— i Pi $2 t .9 7 6 8 + ■ • • • + “ n : ] P ;—n n n 1 “ i+1 m ean la g R2 i= 0 16 .7 8 5 1 8 .6 .0 0 5 8 .0 0 8 2 -.0 9 9 8 .9 7 3 6 20 .8 0 3 8 8.1 .0 0 5 6 .0 0 9 8 -.0 7 7 8 .9 7 4 4 24 .8 3 3 5 7 .7 .0 0 5 4 .0 0 9 1 — .0 6 7 4 .9 7 6 4 DECEMBER 1 9 6 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 I S FEDERAL RESERVE SYSTEM ACTIONS DURING 1969 Federal Reserve Credit Annual Rates of Change 12/68 to 11/69 p. 12/67 to 12/68 8.3 5.1 - 2 .3 3.2 -1 .7 Federal Reserve Holdings of Government Securities Federal Reserve Credit ___________________________ Monetary Base ________________________________ Reserves Available for Private Demand Deposits 7.4 10.2 7.8 6.5 7.0 p. - preliminary Discount Rate (FRB St. Louis) In effect January 1, 1969 _______________________ ______________________ April 4, 1969 __________________________ In effect December 15, 1969 ___________________ " 5%% 6 Reserve Requirements1 Percentage Required Net Demand Deposits _____ up to $5 Million Reserve City Other MemBanks ber Banks 16% 17 17 In effect Dec. 15, 1969 12 12% 12% Net Demand Deposits in Excess of $5 Million Reserve City Other MemBanks ber Banks 17 17% 17% Time Deposits up to $5 Million Time Deposits & Savings Peps, in excess of $5 mil. 12 13 13 3 6 3 6 M argin Requirements on Listed Stocks 80% 80% In effect January 1, 1969 ------------------------------------------------------In effect December 15, 1969 _________________________________ Maxim um Interest Rates Payable on Time & Savings Deposits Type of Deposit Savings Deposits __________ Other Time deposits: Multiple maturity: 90 days or m o re ___ Less than 90 days Single maturity: Less than $100,000 30-59 d a y s________ 60-89 days________ 90-179 days 180 days and over In effect Jan. 1, 1969 In effect Dec. 15, 1969 4% 4% 5 4 5 4 5 5% 5% 5 5% 5% 6 6 6Vi 6Yi 'Beginning October 16, 1969, a member bank is required under Regulation M to maintain, against its foreign branch deposits, a reserve equal to 10 per cent of the amount by which (1 ) net balances due to, and cer tain assets purchased by, such branches from the bank’s domestic offices, and (2 ) credit extended by such branches to U. S. residents exceed certain specified base amounts. Regulation D imposes a similar 10 per cent reserve requirement on borrowings by domestic offices of a member bank from foreign banks, except that only a 3 per cent reserve is required against such borrowings that do not exceed a specified base amount. Page 39 REVIEW IN D E X — 1969 Month o f Issue Title o f Article Jan. Saving Flows in the Current Expansion Growth — Metropolitan vs. Nonmetropolitan Areas in the Central Mississippi Valley Feb. Stabilization Policy and Inflation Operations of the Federal Reserve Bank of St. Louis — 1968 International Monetary Reform and the ^Crawling P eg” Mar. Restraining Inflation Relations Among Monetary Aggregates A Program of Budget Restraint The Relation Between Prices and Employment: Two Views Farm Income Prospects A pr. Monetary Actions, Credit Flows and Inflation Monetary & Fiscal Actions: A Test of Their Relative Importance in Economic Stabiliza tion — Comment — Reply Towards A Rational Exchange Policy: Some Reflections on the British Experience M ember Bank Income — 1968 Month o f Issue Title o f Article July Credit Flows and Recent Interest Rate Trends An Explanation of Federal Reserve Actions (1933-68) International Monetary Reform and the Crawl ing Peg’ ’ — Comment — Reply Aug. Inflation Continues The Influence of Economic Activity on the Money Stock — Comments on the “ St. Louis Position” Reply Additional Empirical Evidence on the Reverse-Causation Argument M eat Prices Sept. Adjustment of D em and Deposit Series Stabilization Actions in 1969 — How Much Restraint? Controlling Inflation A Historical Analysis of the Credit Crunch of 1966 Oct. Recent Interest Rate Developments Revision of the Money Supply Series Elements of Money Stock Determination May Restraining the Growth of Total Spending Federal Open M arket Committee Decisions in 1968 — A Year of Watchful Waiting Controlling Money N ov. Progress in Controlling Inflation Monetary and Fiscal Influences on Economic Activity — The Historical Evidence The Effects of Inflation (1960-68) June Real Growth and Prices Monetary Policy and Inflation The Case for Flexible Exchange Rates, 1969 Dec. 1969 — Battle Against Inflation Selective Credit — No Substitute fo r Monetary Restraint Interest Rates and Price L evel Changes, 1952-69 Federal Reserve System Actions During 1969 R eview Index — 1969 Page 40