The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
3 O n M a in t a i n in g a R is in g U .S . S t a n d a r d o f L iv in g I n t o t h e M id - 2 1 s t C en tu ry 1 7 T h e F O M C in 1 9 8 9 : W a l k i n g a T ig h tro p e 3 6 A M e t h o d o lo g i c a l A p p r o a c h to C h a o s : A r e E c o n o m i s t s M is s in g t h e P o in t? 4 9 T h e L eg acy o f th e M o n e ta rist C o n tro v ersy THE FEDERAL A RESERVE RANK of ST.lJtM IS 1 F ederal R eserve B ank o f St. Louis R e v ie w M arch/April 1990 In This Issue . . . As population grow th in the United States declines, the average age of the population is rising. In the first article of this Review , "On M aintain ing a Rising U.S. Standard of Living into the Mid-21st C entury,” Keith M. Carlson exam ines w hether this com bination of slowing population and rising age composition bodes ill for maintaining a continuing rise in the nation's standard of living into the next century. Using Bureau of Census population projections through 2050, along with historical relationships among population, employment, capital stock and output, Carlson calculates the grow th rates of the capital stock that would be required to achieve certain rates of increase in the standard of living as m easured by GNP per capita. He concludes that achieving a rising standard of living like the one we experienced from 1948 to 1989, 1.9 percent grow th in GNP per capita, does not appear possible. A rising standard of living in the 1 to 1.5 percent range does appear achievable, however. * * * In the second article of this issue, “The FOMC in 1989: W alking a Tightrope,” Michelle R. Garfinkel exam ines the econom ic factors that in fluenced the Federal Open M arket Committee’s deliberations and deci sions in 1989. In reviewing the FOMC’s long- and short-run policy deci sions, Garfinkel emphasizes how the FOMC sought to balance the risk of g reater inflationary pressures against that of a w eakening economy. At the beginning of the year, the th reat of a w orsening of the under lying trend in inflation drove the form ulation of policy. As evidence of a w eakening econom ic expansion accum ulated and the outlook for infla tion appeared m ore promising, the FOMC shifted its prim ary focus to the possibility of a future slowdown in econom ic activity. The effect of this shift on policy, however, did not em erge immediately, for the FOMC understood that its interpretation of econom ic data was subject to great uncertainty, and it did not want to jeopardize eith er the progress that had been made tow ard achieving its ultim ate goal of long-run price stability or its credibility as an inflation-fighter. * * * A new theory, called chaos, w hich offers an explanation fo r the seem ingly random behavior of econom ic and financial variables, has stirred up considerable interest recently. In the third article of this Review, “A Methodological Approach to Chaos: Are Econom ists Missing the Point?” Alison Butler looks at how MARCH/APRIL 1990 2 econom ics is traditionally modeled and discusses w hy many of these models exclude, a priori, the possibility of chaotic behavior. Chaos is defined and its properties illustrated. The advantages and pitfalls of in corporating chaos into econom ic modeling are then discussed. Butler finds that the study of nonlinear dynamics in general, and determ inistic chaos in particular, may eventually provide econom ists w ith a determ inistic explanation fo r variables that appear to be random. * * * In the last article of this issue, we print the Fourth Annual Homer Jones Memorial Lecture. This y ear’s speaker, David Laidler, a professor of econom ics at the University of W estern Ontario, reviews "T h e Legacy of the M onetarist C ontroversy.” Laidler’s paper has tw o broad themes. The first is that people working in m onetary econom ics in the 1960s and ’70s lost sight of the fundamental issues under debate. For example, while original m onetarist argum ents stressed long-run relationships b e tw een changes in the money stock and oth er variables, a generation of young econom ists re-estim ated traditional relationships with quarterly data, found instability, and concluded that m onetarist propositions had been rejected. Laidler views this conclusion to be misplaced because m onetarists had never argued they could (or should) model the com plicated short-run dynamics that lead to the long-run relationship of interest. Laidler’s second them e is that ideas from New-classical econom ics, which w ere seen originally as supporting m onetarist thought, actually subverted it. In particular, the transm ission m echanism betw een changes in m oney and oth er variables proposed by m onetarists was dif feren t in fundam ental ways from that proposed by New-classical analysis, and em pirical studies eventually rejected im portant Newclassical ideas. But by that time, Laidler argues, many econom ists had confused traditional m onetarist principles with those of New-classical analysis. Laidler concludes that the legacy of the m onetarist controversy is an uncom fortable one, w ith im portant issues yet to be resolved. FEDERAL RESERVE BANK OF ST. LOUIS * * * 3 Keith M. Carlson Keith M. Carlson is an assistant vice president at the Federal Reserve Bank of St. Louis. Thomas A. Pollmann provided research assistance. On M aintaining a R ising U.S. S tand ard of Living Into th e M id-21st C en tu ry I n EARLY 1989, the Bureau of the Census released its latest population projections for the United States through 2080, showing slow er growth in the total population from 1990 through 2035 and little grow th th e rea fter .1 At the same time, the median age of the U.S. population is projected to continue rising. The rising median age prim arily reflects the aging of the babyboom generation, those born betw een 1946 and 1964. Given these projections, it is natural to ask w hether the United States will be able to maintain its standard of living into the next cen tury. A typical example of such con cern recen t ly appeared in the "Labor L etter" colum n of the W all Street Jo u rn a l:2 GRIM FORECAST: Most baby boomers “will find retirement at age 65 unaffordable" when they get there starting in 20 years, a survey of North American actuaries finds. They blame inadequate savings, tax incentives and too few workers. 1Bureau of the Census (1989). 2Wall Street Journal (1989). For a general discussion of the challenges facing the U.S. economy in the future, see Council of Economic Advisers (1990), chapter 4. 3For a discussion of the forecast accuracy of population projections, see Bureau of the Census (1989), pp. 14-16. The purpose of this article is to explore the arithm etic that underlies the broad econom ic implications o f these population projections. This involves deriving projections of the civilian labor force and employment from the popula tion projections. Achieving certain rates o f in crease in the standard of living m eans achieving certain rates of grow th of the U.S. capital stock. The feasibility of these various capital stock re quirem ents is exam ined in light of historical ex perience to assess the prospects for future econom ic grow th. P O P U L A T IO N PRO JECTIO NS TO 2050 The Census Bureau has prepared population projections fo r the United States fo r the last 30 y ears .3 In its most recen t report, the Bureau constructed th ree basic projection series using The bureau reports that, for a projection period of 15 years, the root-mean-square error of previous projected growth rates is 0.40 percentage points. The difference be tween the growth rate for its highest (lowest) projection series and the middle series for 1990 to 2005 is 0.34 (0.28) percentage points. MARCH/APRIL 1990 4 Table 1 Historical and Projected (Middle Series) Population Growth: 1950-2050 Period Actual: 1950 to 1960 1960 to 1970 1970 to 1980 Projected: 1980 to 1990 to 2000 to 2010 to 2020 to 2030 to 2040 to 1990 2000 2010 2020 2030 2040 2050 Growth rate 1.72% 1.27 1.06 0.95 0.69 0.52 0.41 0.21 0.04 -0 .0 7 three d ifferent assumptions about fertility rates, life expectancy and net im m igration .4 The dif feren ces in the fertility assumptions are chiefly responsible fo r m ost of the d ifferences in the projections. T he highest, middle and lowest series reflect the highest, middle and lowest assumptions, respectively, fo r each of these components. T he Census Bureau em phasizes that its highest and low est population projections do not rep resen t estim ated confidence intervals based on statistical analysis. Instead, they are referred to simply as a “reasonable” high and low p rojec tion. T he num erical exercises in this article con cen trate on the middle series projection because the Census Bureau considers it the “m ost likely.” In addition, the horizon of the population pro jections is the year 2050; this enables the Study to focus on the retirem ent years of the babyboom generation. T he Bureau of the Census projections are sum marized in table 1 and figure 1. Declining population grow th is not something new; it has 4By combining the assumptions in all possible ways, there are 27 different population projections. In addition, the Census Bureau prepares three series with a zero netimmigration assumption. The middle series for net im migration starts with 600,000 persons per year and is reduced and kept at 500,000 persons per year after 1997. Compared with the zero net-immigration case, this in FEDERAL RESERVE BANK OF ST. LOUIS Medial age in years (end of period) Percentage of population 65 and over (end of period) 29.4 27.9 30.0 9.2% 9.8 11.3 33.0 36.4 38.9 40.2 41.8 42.6 42.7 12.6 13.0 13.9 17.7 21.8 22.6 22.9 been under way fo r a num ber of years. The total population grew at a 1.7 percent annual rate from 1950 to 1960, then slowed to a 1.3 p ercent rate in the 1960s, b efo re slowing even fu rth er in the 1970s and 1980s. From 1990 to 2040, the Census Bureau projects a positive, but steadily declining population grow th rate; after 2040 the population level itself is projected to decline. As table 1 shows, one notable feature of the projections is the expected aging of the popula tion. The median age reached a post-1960 low of 27.9 years in 1970 and is estim ated to reach 33 years in 1990, b efo re clim bing to 4 2.7 years by 2050. Hand in hand with this overall aging of the population is the rising proportion of the population that is age 65 and older. T he m ajor issue of con cern is w h eth er slow ing grow th in the population, com bined with the rising age composition, bodes ill fo r main taining a continuing rise in the nation's standard of living. creases the population estimate for 2050 by 49 million per sons, or 19.5 percent. Each 100,000 net immigrants per year has the effect of increasing the 1990-2050 average population growth rate by about 0.05 percent. After 1998, the low net-immigration assumption levels out at 300,000 persons per year, and the high assumption levels out at 800,000 per year. 5 Figure 1 Population Projections 1 Ratio Scale Millions of Persons 450 Ratio Scale Millions of Persons 450 150 1950 1970 ’ Total population 1990 PROJECTIONS OF L A B O R FORCE A N D EM PLO YM ENT T O 2050 Unfortunately, the Census Bureau and the D epartm ent of Labor do not provide labor force and em ploym ent projections as fa r into the fu tu re as the year 2050. T h erefore, such p rojec tions must be estim ated in some m anner. Such projections requ ire inform ation about the rate at w hich the population will participate in the civilian labor force and the rate at w hich this labor fo rce will be employed. These rates, in turn, depend on the age and sex composition of the population and the labor force. 5Fullerton (1989). 6The effects of an alternative set of participation rate assumptions based on a continuation of recent trends were also examined. In particular, labor force participation of females between the ages of 16 and 64 was assumed to continue rising throughout the 2000 to 2050 period, with 2010 2030 150 2050 Participation Rates Table 2 sum marizes historical and projected participation rates fo r 1 0 age-sex groups for 10-year periods from 1950 through 2000. The projected figures fo r 1990 and 2000 are based on estim ates made by the D epartm ent of Labor .5 For purposes of this analysis, these same agesex group projections for 2 0 0 0 are assumed to hold constant through 2 0 5 0 .6 Even though the participation rates are held constant fo r each age-sex group, the total p ar ticipation rate changes quite dramatically in the the participation rate of all females 3.7 percentage points above the basic assumption by 2050. These alternative assumptions increased the overall participation rate by about 2 percentage points by 2050 and reduced the re quired growth rate of the capital stock (for a given stan dard of living scenario) from 1990 to 2050 by 0.2 percent. MARCH/APRIL 1990 6 Table 2 Historical and Projected Paticipation Rates1: 1950-2000 1950 1960 1970 1980 1990 2000 Males: 16 to 19 20 to 24 25 to 54 55 to 64 65 and over 63.2% 87.9 96.5 86.9 45.8 56.1% 88.1 97.0 86.8 33.1 56.1% 83.3 95.8 83.0 26.8 60.5% 85.9 94.2 72.1 19.0 58.0% 85.4 93.6 67.3 16.4 59.0% 86.5 93.0 68.1 14.7 Females: 16 to 19 20 to 24 25 to 54 55 to 64 65 and over 41.0 46.0 36.8 27.0 9.7 39.3 46.1 42.9 37.2 10.8 44.0 57.7 50.1 43.0 9.7 52.9 68.9 64.0 41.3 8.1 54.4 72.9 74.3 45.4 8.3 59.6 77.9 81.4 49.0 7.6 Total, 16 and over 59.2 59.4 60.4 63.8 66.7 69.0 ’ Civilian labor force as a percent of civilian noninstitutional population, which equals total population minus armed forces, institutionalized persons and persons under 16 years of age. Figure 2 Participation Rate Projections 1 Percent 70.0 Percent 70.0 65.0 57.5 1950 1970 1990 2010 ''Civilian labo r fo rce as a percent o f civilia n n o n in s titu tio nal p o pulation Digitized FEDERAL RESERVE BANK OF ST. LOUIS for FRASER 2030 57.5 2050 7 Figure 3 Population and Employment Ratio Scale Millions of Persons 1950 Ratio Scale Millions of Persons 1970 1990 projection period because o f the aging trend in the population (see figure 2). This projected decline in the participation rate after 2 0 0 0 raises doubts as to w hether the general standard of living can be sustained by a w ork fo rce that is shrinking relative to population. Total Civilian Em ploym ent One m ust also m ake a projection of total em ployment, given the projections of population and the labor force. Neither the Census Bureau n or the Departm ent of Labor m ake employment projections by age-sex group. So, fo r purposes o f this analysis, 1989 employment rates are as sumed to hold constant throughout the p rojec 7An age-sex adjusted employment (unemployment) rate is also derived. For the middle series population projection, the age-sex adjusted employment rate rises slowly from 94.7 percent in 1989 to 95 percent in 2050. The adjusted unemployment rate falls from 5.3 percent in 1989 to 5 per 2010 2030 2050 tion period for each of the 1 0 age-sex groups. Applying these employment rates to the chang ing age distribution of the labor force yields projections of total civilian em ploym ent .7 The result of these total em ploym ent projections is charted in figure 3 along w ith the middle-series population projection from figure 1. The ratio of total employment to total population is shown in figure 4. Continuing a trend that began in the 1960s, the w orking proportion of the popu lation is projected to rise until 2005; it then declines until 2030 b efore leveling off in 2050. Changes in the employment-population ratio after 2 0 0 0 reflect only the changes in the agesex distribution of the population. cent in 2050. The resulting implications for capital stock growth are not very sensitive to these employment rate assumptions. For example, a 97 percent employment rate assumption in 2050 would reduce the required growth in the capital stock by about 0.1 percent. MARCH/APRIL 1990 8 Figure 4 Employment-Population Ratio 1 .525 .500 .475 .450 .425 .400 .375 .350 1950 1970 1990 2010 2030 .350 2050 1Total civilian employment relative to total population A LT E R N A T IV E STAND AR D - OFLIV IN G ASSU M PTIO NS To assess the econom ic implications of these projections on the standard of living, one must define and m easure the standard of living. An admittedly crude, but commonly used, m easure o f the standard of living is GNP per capita. It does not capture changes in the distribution of real incom e among the population, but, for an econom ically advanced country like the United States, it is deemed useful in sketching the pro spects fo r future econom ic grow th. T h ree sce narios are developed: Scenario 1: what grow th in the nation’s capital stock would be required to keep real GNP/capita growing at about “ Note that graphing scenario 3 gets ahead of our story: with regard to GNP per capita, it is a result rather than an assumption. Digitized FEDERAL RESERVE BANK OF ST. LOUIS for FRASER 1.85 percent per year, its actual grow th rate from 1948 to 1989? Scenario 2: what grow th in the nation’s capital stock would b e requ ired if real GNP/capita rem ained constant at its 1990 value? In this case, th ere would be zero grow th (no change) in the nation’s standard of living. Scenario 3: w hat would the grow th in GNP/capita be if the capital stock grew at its 1948-89 tren d rate? This is an in term ediate case, betw een scenario 1 and 2 . Figure 5 depicts these alternative standard-ofliving assumptions graphically .8 9 Figure 5 Real GNP Per Capita Ratio Scale Thousands of 1982 Dollars 60 Ratio Scale Thousands of 1982 Dollars 60 1950 1970 1990 Capital Stock Implications 2010 2030 2050 nent of grow th equal to 0.81 percent per year because of an implicit time tren d .9 Deriving the capital stock or GNP per capita specified in the scenarios described above re quires the use of an estim ated production fun c tion. The function chosen was developed by Rasche-Tatom (1977) and updated in Tatom (1988). This production function is Cobb-Douglas in form and includes an implicit time trend (proxy fo r technical change) and the relative price of energy (for details, see appendix). Based on this estim ated production function, each standard-of-living assumption includes a com po The derived results fo r the capital stock are summ arized in table 3 and shown in figure 6 . Table 3 decomposes the grow th of GNP per capita in two ways: (1 ) its simple com ponents, GNP and population, and (2) the contribution of resource inputs and the employment-population ratio. The decomposition fo r the 1948-89 period shows that GNP per capita grew rapidly, in part, because em ploym ent grow th dramatically exceeded population grow th. For the 1990-2050 9This estimate of technical progress is of critical importance in deriving the estimates of capital stock requirements. The estimate of 0.81 per year is optimistic; estimating the production function with a zero-one dummy suggests that technical progress has been only 0.45 percent annually since 1968. Energy price shocks have important effects on GNP and the standard of living, but for these projections the relative price of energy is assumed to be unchanged from its 1989 value. Generally, such shocks cannot be predicted. MARCH/APRIL 1990 10 Table 3 Decomposition of Per Capita GNP Growth1 Projected: 1990-2050 Historical 1948-89 Real GNP growth Minus: population growth Equals: per capita GNP growth Or Real GNP growth Technical progress Plus: employment contribution Employment growth Times: elasticity Plus: capital stock contribution Capital stock growth Times: elasticity Minus: employment growth Plus: change in employment-population ratio Employment growth Minus: population growth Equals: per capita GNP growth3 3.27% 1.30 1.972 3.27 0.81 1.36 1.72 (0.79) 0.76 3.60 (0.21) 1.72 0.42 1.72 1.30 1.972 Scenario 1 Trend growth in GNP/capita Scenario 2 Zero growth in GNP/capita 2.17% 0.30 1.872 Scenario 3 Trend growth in capital stock 0.30% 0.30 0.00 2.17 0.81 0.21 0.26 (0.79) 1.14 5.45 (0.21) 0.26 -0 .0 4 0.26 0.30 1.872 1.78% 0.30 1.48 0.30 0.81 0.21 0.26 (0.79) -0 .7 0 -3 .3 3 (0.21) 0.26 -0 .0 4 0.26 0.30 0.00 1.78 0.81 0.21 0.26 (0.79) 0.76 3.60 (0.21) 0.26 -0 .0 4 0.26 0.30 1.48 1 Some components do not add to total because of rounding or production function error. 2 Rates calculated from initial year to terminal year, which differs from rate in text which is trend rate fitted to 1948-89. 3 X - P = (X-L) + (L-P), i.e., productivity growth plus change in employment-population ratio. The contribution of the change in the relative price of energy was zero historically, and zero by assumption for the projection period. Figure 6 Capital Stock Requirements1 Ratio Scale Trillions of 1982 Dollars Ratio Scale Trillions of 1982 Dollars 100.00 75.00 50.00 25.00 10.00 7.50 5.00 .50 1950 1970 1990 <Net real nonresidential fixed capital stock Digitized FEDERAL RESERVE BANK OF ST. LOUIS for FRASER 2010 2030 2050 11 period, the em ploym ent and population p rojec tions place a g reater burden on capital stock grow th in achieving GNP grow th. As figure 6 shows, the d ifferences in the necessary capital stock among the alternatives are huge by 2050. Expressing these requ ire m ents in term s of rates of change of the capital stock from 1990 to 2050 (also shown in table 3), the differences are m ore understandable. To achieve the 1948-89 trend grow th rate in real GNP per capita, given the population-employm ent projections, would requ ire a 5.45 percent average rate of grow th in the capital stock, m uch faster than the 3.60 percent tren d rate of grow th from 1948-89. The second alternative of no grow th in real GNP per capita implies a steady decline in the capital stock. This reflects most clearly the role of technical change in the production function. A com bination of declining capital stock and continuing grow th in technical change is quite implausible, how ever. If the rate of technical progress w ere zero, the capital stock would have to rise slightly at a 0.44 percent average rate, rath er than fall, to keep the standard of living constant at the 1990 level. Investment Ratio An alternative way of examining these results is to look at the annualized first d ifference of the capital stock relative to GNP, that is, the in vestm ent ratio .10 The annualized d ifference of the capital stock rep resen ts the flow of n et real nonresidential fixed investm ent in plant and equipment. The investm ent ratio averaged 3.06 percent during 1948-89, with a maximum value of 4.29 percent in 1966. The impossibility of maintaining a rise in the standard of living at the 1948-89 tren d rate becom es obvious in figure 7; it would require a sharp and continuing increase in the investm ent ratio after 2000. The investm ent ratio averages 16 percent from 1995 to 2050. The no-growth alternative, of course, implies a steady decline in the capital stock and a nega 10Since the projections were for every fifth year starting with 1995, investment was calculated as: AK, = ( (K,/K,_ 5 2- 1 ) • K, . )° This was divided by the level of GNP in year t to get the investment ratio. tive investment-GNP ratio. If technical progress w ere zero, how ever, the capital stock req u ire m ents would be higher. An average investm ent ratio of 0.42 p ercen t would be required for 1995-2050 under such assumptions. The final alternative—a capital stock that con tinued to rise at its 1948-89 trend rate—implies a slowly rising investm ent ratio that reaches 1 0 p ercen t by 2050, still well above any value it reached during 1948-89. For this scenario, the grow th of per capita real GNP varies over the projection period depending on the projected grow th rate of employment; how ever, it aver ages about 1.48 p ercen t from 1990 to 2050 (see figure 5). O TH ER A N A LY T IC A L R ATIO S Productivity o f Labor A nother way to assess the likelihood of alter native assumptions is to exam ine their implied productivity trends relative to historical ex perience. These trends are shown in figure 8 . Scenario 1 implies a grow th rate for productivi ty of 1.90 percent from 1990 to 2050, m uch faster than its 1.52 percent trend rate of grow th from 1948-89. The no-grow th scenario requires only a 0.04 percent rate of increase in produc tivity. Scenario 3 yields a grow th rate of p ro ductivity of 1.52 percent, or the same as the trend rate from 1948-89.11 Capital-Labor Ratio An examination of capital-labor ratios can also be used to assess the feasibility of the th ree standard-of-living alternatives. The implied capi tal-labor ratios for these scenarios, summarized in figure 9, m ust be interpreted with care. Over the 1948-89 period, the capital-labor ratio rose slowly, at a 1.85 p ercent annual rate. M ore sig nificantly, how ever, its rate of increase has de clined over this period, from a 3.2 percent grow th rate from 1948-58, to a 2.1 p ercen t rate from 1958-75, to a 0.6 percent rate of grow th since 1975. W ith this pattern in mind, calcula tion of capital-labor ratios for the alternative scenarios does not yield clear-cut conclusions. 11ln general, implied productivity trends shed little light on feasibility because they follow from the per capita GNP assumptions; productivity growth differs from per capita GNP growth by the difference between projected popula tion growth and employment growth. MARCH/APRIL 1990 12 Figure 7 Investment Ratio1 -.10 - .1 0 1950 1970 1990 2010 2030 2050 'C hange in net real capital s to c k divided by real GNP Figure 8 Productivity1 Ratio Scale Thousands of 1982 Dollars Ratio Scale Thousands of 1982 Dollars 110 110 20 1950 1970 'Real GNP per worker FEDERAL RESERVE BANK OF ST. LOUIS 20 1990 2010 2030 2050 13 Figure 9 Capital-Labor Ratio1 Ratio Scale Millions of 1982 Dollars Ratio Scale Millions of 1982 Dollars .800 .002 1950 1970 1990 'Net real capital stock per employed worker Scenario 1 requires a rise in the ratio to 5.18 p ercen t a y ear from 1990 to 2050, m uch faster than that which occu rred even during the 194858 period. The zero-grow th case does not m atch historical experience at all. T he trend-growth-ofcapital scenario requires a rise in the capitallabor ratio of 3.34 p ercen t per year. W hile this is in line with the early post-war experience, it is m uch faster than the m ore recen t experience. C A P IT A L REQUIREM ENTS FOR ALT E R N A T IV E P O P U L A T IO N PROJECTIONS The capital stock requirem ents associated with the Census Bureau’s high and low population projections are sum m arized in table 4. T he basic assumptions about participation rates, unem ployment rates and the production function are the same as used above fo r the middle-series population projection. The high-series popula tion projection is 38 p ercent higher than the 2010 2030 2050 middle projection in 2050, while the low-series projection is 23 p ercen t lower. Despite this wide variation in population, the capital requirem ents do not differ greatly from those derived with the middle-series projection. T he differences are relatively small because, as shown in figure 1 0 , the pattern of m ovem ent of the employmentpopulation ratio over the 1990-2050 period dif fers very little across the alternative projections. This ratio is the key factor underlying the cru cial im portance placed on capital stock grow th in achieving the particular grow th rates in per capita incom e that w ere examined. In fact, the high series population projection requires the largest capital stock, because its projected age distribution has the smallest proportion of the population actually working. CONCLUSIONS The grow th of the nation’s output depends on the grow th of its labor force, its capital stock MARCH/APRIL 1990 14 Table 4 Achieving Historical Trend Growth of Real GNP Per Capita in 1990-2050: Alternative Population Projections_____________ Low population series Population Total civilian employment Real GNP Capital stock Investment ratio1 -0.13% -0 .1 7 1.73 4.89 13.69 Middle population series High population series 0.30% 0.26 2.17 5.45 16.02 0.83% 0.74 2.70 6.17 19.93 1 Change in capital stock as a percent of real GNP: average for 1995-2050 Figure 10 Employment-Population Ratios: Alternative Population Projection Series FEDERAL RESERVE BANK OF ST. LOUIS 15 and technical progress. The Census Bureau has released projections of U.S. population into the 21st century. T h eir middle projection shows slowing grow th in the population through 2040, followed by a slight decline in the population itself for the rest of the outlook period. They also project that the median age o f the popula tion will rise steadily, implying that the w ork force as a proportion of the total population will eventually decline. Can a rising standard of liv ing be achieved in the face o f such projections? The conclusions derived from the study are as follows: (1) Achieving a rising standard of living at the rate experienced from 1948-89 does not appear possible. Based on an estimated technical pro gress of 0.8 percent per year, the nation’s capi tal stock would have to grow at a 5.5 percent average annual rate from 1990 to 2050, well above its 3.6 percent rate of growth from 1948 to 1989. Achieving this higher capital stock growth would require an investment-GNP ratio that reaches 33 percent by 2050. (2) Maintaining the 1990 standard of living at a constant level requires that the capital stock decline at a 3.3 percent rate. This case was not intended to be taken seriously; it is presented solely to show, given this analysis, why zero growth (or less) is an unlikely possibility. (3) Continued growth in the capital stock at its 1948-89 trend rate yields a growth rate in per capita real GNP of 1.5 percent. However, to achieve this would require a steadily rising investment-GNP ratio that reaches 10 percent 12Demographic projections should produce a rise in the saving rate as the baby-boom generation matures, before falling as they reach retirement age. Such projections are by 2050. While achieving this ratio would be unusual, the required ratio would be within the range of historical experience until 2015. Left unexam ined in this study is an analysis of the prospects fo r saving .12 Capital stock req u ire m ents will not be achieved unless resources are released for investm ent by abstaining from p re sent consum ption. By using historical trends in realized investment-GNP ratios, how ever, an ap proxim ate determ ination of feasible grow th paths was possible. Increasing the standard of living at the 1948-89 pace appears impossible. On the oth er hand, achieving a rising standard o f living in the 1 to 1.5 p ercen t range appears achievable in light of historical experience. REFERENCES Aaron, Henry J., Barry P Bosworth, and Gary T. Burtless. . Can America Afford to Grow Old? (Brookings Institution, 1989). Bureau of the Census, U.S. Department of Commerce. Pro jections of the Population of the United States, by Age, Sex, and Race: 1988 to 2080 (GPO, 1989). Council of Economic Advisers. Economic Report of the Presi dent (GPO, 1990). Fullerton, Howard N., Jr. “ New Labor Force Projections, Spanning 1988 to 2000,” Monthly Labor Review (November 1989), pp. 3-12. Rasche, Robert H., and John A. Tatom. “ Energy Resources and Potential GNP,” this Review (June 1977), pp. 10-24. Tatom, John A. “Are the Macroeconomic Effects of Oil-Price Changes Symmetric?” Stabilization Policies and Labor Markets, Carnegie-Rochester Conference Series on Public Policy, Vol. 28 (Spring 1988), pp. 325-68. Wall Street Journal, November 7, 1989. tenuous, however, as pointed out by Aaron, Bosworth and Burtless (1989), p. 139: “ Demographic trends appear to bear little relation to past rates of private saving . . . ” MARCH/APRIL 1990 16 Appendix Production Function The production function used in this study is a version of the one originally developed by Rasche and Tatom (1977) and modified and up dated in Tatom (1988). It is a Cobb-Douglas function w ith labor, capital and energy specified as resource inputs. GNP (in 1982 dollars) was estim ated as a function of total civilian employ ment, net real nonresidential fixed capital stock adjusted for capacity utilization and the relative price of energy. Using annual data for 1954-89 (omitting the early post-w ar years and the Kor ean W ar), the following estim ate was obtained w here din is the first d ifference of the loga rithm (t-values in parentheses): din ( = .0081309 + .787 din ( VKU; (3.88) (19.85) V K U; - .041 din P ( - 2 .3 5 ) RESERVE BANK OF ST. LOUIS FEDERAL R 2 = .930 SE = .010 DW = 1.952 e = .189 X = GNP in 1982 dollars K = net nonresidential fixed capital stock in 1982 dollars U = Federal Reserve capacity utilization rate (manufacturing) L = total civilian employment P = relative price of energy (ratio of produ cer prices of fuel, related products, and pow er to the GNP implicit price deflator) This estim ated production function was solved fo r In KU and used to derive estim ates of the capital stock for the alternative assumptions making use o f the GNP estim ates implied by those alternatives and the total employment estim ates derived from the population p rojec tions. The 1989 values of the relative price of energy ( = .577) and the capacity utilization rate ( = .839) w ere held constant during the p rojec tion period in the derivation of the n et real capital stock. 17 Michelle R. Garfinkel Michelle R. Garfinkel is an economist at the Federal Reserve Bank of St. Louis. Scott Leitz provided research assistance. The FOMC in 1 9 8 9 : W alking a T ig h tro p e 1 111 FEDERAL Open M arket Committee (FOMC) sought to balance the risk of infla tionary pressures against that of a weakening econom y in 1989, the seventh year of the cu r ren t econom ic expansion .1 The changing relative intensity of these risks, as perceived by the FOMC (hereafter, the Committee), influenced the course of m onetary policy throughout the year. Because the Committee believed that long-run price stability is necessary to prom ote maximum sustainable econom ic grow th over time, the perceived risks of inflationary pressures greatly influenced its decisions early in the year. As the year progressed, how ever, it becam e increasing ly apparent to the Committee that the econom ic expansion was weakening. At the same time, the Com m ittee’s perception of the tren d in infla tion becam e slightly m ore optimistic. According ly, the weight that the Committee attached to reducing the risks of a slowdown in econom ic activity increased somewhat throughout the se cond half of 1989. Nevertheless, the Committee’s con cern about future price pressures and the im portance of maintaining its own credibility as NOTE: Citations to the Record refer to the “ Record of Policy Actions of the Federal Open Market Committee” as published in various issues of the Federal Reserve Bulletin. Citations to “ Report” refer to the “ Monetary Policy Report to the Congress,” which is also published in the Federal Reserve Bulletin. an inflation-fighter rem ained in the fo refro n t of its deliberations. This article review s the form ulation of m one tary policy by the Committee in 1989. T he dis cussion focuses on how changing econom ic con ditions influenced the Comm ittee’s decisions as it balanced the risk of futu re inflation against that of a futu re slowdown in econom ic activity. LO NG-R UN OBJECTIVES The Board of Governors of the Federal Reserve System rep orts to Congress tw ice a year on its annual grow th rate targets for the m onetary and debt aggregates. The one-year target periods run from the fourth q u arter of the previous y ear to the fou rth q uarter of the cu rren t year. These rep orts are mandated by the Full Employment and Balanced Growth Act of 1978 (or the Humphrey-Hawkins Act). A fter its first m eeting of the year in February, the Committee submits a report on its m onetary and debt grow th objectives fo r the cu rren t year. In July, upon reviewing the progress it has made tow ard achieving its objectives for the 'See the shaded insert on pages 18 and 19 for a description of the Committee’s membership during 1989. MARCH/APRIL 1990 18 Organization Of The Committee T h e Federal Open M arket Committee (FOMC) consists of 12 m em bers, including the seven m em bers of the Federal Reserve Board o f Governors and five of the 12 Federal Reserve Bank presidents. The chairm an of the Board o f Governors is traditionally elected chairm an of the Committee. The president of the New York Federal Reserve Bank, also by tradition, is elected the Commit tee’s vice chairm an. All Federal Reserve Bank presidents attend Committee m eetings and present their views, but only those who c u r rently are m em bers of the Committee are perm itted to vote. In previous years, the four m em berships that rotate among the Bank presidents w ere held fo r one-year term s com m encing M arch 1 of each year. In 1989, how ever, the m em berships w ere only for 1 0 m onths so that, starting in 1990, the one-year term s would be on a calendar-year basis. The president of the New York Federal Reserve Bank is a perm anent voting m em ber of the Committee. M em bers of the Board of Governors at the beginning of 1989 included Chairman Alan Greenspan, Vice Chairman Manuel H. Johnson, W ayne K. Angell, H. Robert Heller, Edward W. Kelley, Joh n P. LaW are and M artha R. Seger. Mr. Heller resigned as a m em ber of the Committee during the year. His position was not filled during 1989. The following Bank presidents voted at the meeting on February 7-8, 1989: E. Gerald Corrigan (New York), Robert P. Black (Rich mond), Robert P. Forrestal (Atlanta), W. Lee Hoskins (Cleveland) and Robert T. Parry (San Francisco). In M arch, the Committee m em ber ship changed and the presidents’ voting posi tions w ere filled by E. Gerald Corrigan (New York), Roger Guffey (Kansas City), Silas Keehn (Chicago), Thom as C. M elzer (St. Louis) and Richard F. Syron (Boston). The Committee m et eight tim es at regularly scheduled m eetings during 1989 to discuss econom ic trends and decide the future course of open m arket operations .1 As in previous 1No meetings were held in January, April, June or September. FEDERAL RESERVE BANK OF ST. LOUIS years, telephone consultations occasionally w ere held betw een scheduled meetings. At the end of each scheduled meeting, a direc tive was issued to the Federal Reserve Bank of New York. Each directive contained a short summary of econom ic developments, the general econom ic goals sought by the Committee, its long-run m onetary grow th ob jectives, and instructions to the M anager for Domestic Operations at the New York Bank fo r the conduct of open m arket operations during the new interm eeting period. These instructions w ere stated in term s of the degree of pressure on reserve positions to be sought or maintained. The reserve conditions stated in the directive w ere deemed consis tent with specific short-term grow th rates for M2 and M3 w hich, in turn, w ere considered consistent with desired long-run grow th rates for these m onetary aggregates. T he Commit tee also specified interm eeting ranges fo r the federal funds rate. These ranges provided one m echanism fo r initiating consultations betw een m eetings w henever it appeared that the constraint on the federal funds rate was inconsistent w ith the objectives for the behavior of the m onetary aggregates. The M anager for Domestic Operations has the prim ary responsibility fo r formulating plans regarding the timing, types and amounts o f daily buying and selling of securities in fulfilling the Comm ittee’s direc tive. Each m orning the M anager and his staff plan the open m arket operations fo r that day. This plan is developed on the basis of the Com mittee’s directive and the latest develop m ents affecting m oney and credit m arket conditions, the grow th o f m onetary aggre gates, and bank reserve conditions. T h e Man ager also consults with the Board’s staff. Pre vailing m arket conditions and open m arket operations that the M anager proposes to exe cute are discussed each m orning in a tele phone con feren ce call involving the staff at the New York Bank, one voting president at 19 another Reserve Bank and staff at the Board. O ther m em bers of the Com m ettee may par ticipate and are inform ed of the daily plan by internal memo or wire. The directives issued by the Committee and a summary of the discussion and reasons for Committee actions are published in the "Record of Policy Actions of the Federal Open M arket Com m ittee.” T he “Record” fo r each m eeting is released a few days after the next scheduled Committee meeting. It subsequent ly appears in the F ed era l R eserv e Bulletin. In addition, “Records” fo r the entire year are published in the annual rep ort of the Board of Governors. T he record fo r each meeting in 1989 included: (1 ) a staff sum m ary of recen t econom ic developm ents—such as changes in prices, employment, industrial produc tion and com ponents of the national accounts—and projections of general price, output and employment developments for the year ahead; first half of the year, the Committee decides w h eth er to adjust or retain its target fo r the cu rren t y ear and establishes tentative targets for the following year. Shortly after the July meeting, a rep ort of the Com m ittee’s decisions is submitted to Congress. Table 1 sum m arizes the Committee's long-run m onetary grow th ob jectives for 1989 as reported to Congress. As was the case in the previous two years, the Committee did not establish a targ et range for M l in 1989. T he Committee believed that the unpredictable relation o f M l to econom ic activity and prices did not w arran t reliance on this aggregate as a guide to the im plem entation of m onetary policy .2 To u nderscore its com m itm ent to resist future inflationary pressures and to m ake progress tow ard reasonable price stability, the Committee reaffirm ed the 1989 targets for M2 and M3, 3 to 7 p ercen t and 3Vz to 7Vz percent, respective 2Record (May 1989), pp. 357-58. See Hafer and Haslag (1988), who discuss the Committee's decision to omit a target range for M1. (2 ) a sum m ary of recen t international financial developments and the U.S. foreign trade balance; (3) a sum m ary of open m arket operations, grow th of the m onetary aggregates and bank reserves and m oney m arket condi tions since the previous meetings; (4) a sum m ary of the Committee’s discus sion of the cu rren t and prospective econom ic and financial conditions; (5) a sum mary o f the m onetary policy discussion of the Committee; (6 ) a policy directive issued by the Commit tee to the Federal Reserve Bank of New York; (7) a list o f the m em bers’ votes and any dissenting com m ents; and (8 ) a description of any actions regarding the Committee’s oth er authorizations and directives, and reports on any ac tions that might have occu rred betw een the regularly scheduled meetings. ly, that had been set tentatively in July 1988.3 These target ranges for M2 and M3 w ere 1 and Vz percentage points, respectively, low er than those established for 1988. The Committee also decided to m aintain the 4 percentage-point spreads betw een the upper and low er bounds o f the target ranges for the tw o broad m onetary aggregates. Until tw o years ago, this spread had been 3 percentage points. The w ider ranges w ere adopted in 1988 w hen the Committee concluded that the relations of M2 and M3 grow th to econom ic grow th and in flation had becom e considerably m ore variable and, therefore, that estim ates of grow th rates fo r these aggregates that would be consistent with the Com mittee’s objectives for the econom y w ere subject to greater uncertainty. Forecasting the appropriate grow th rates was made more difficult by the Com mittee’s uncertainty about the impact of future developments on th rift in stitutions and the subsequent effect on the 3Report (March 1989), p. 108. MARCH/APRIL 1990 20 Table 1 The FOMC’s Long-Run Operating Ranges Ranges Date of meeting Target period M2 M3 June 29-30, 19881 February 7-8, 19892 July 5-6, 1989 IV/1988-IV/1989 IV/1988-IV/1989 IV/1988-IV/1989 3-70/o reaffirmed reaffirmed 3.5-7.5% reaffirmed reaffirmed July 5-6, 19893 IV/1989-IV/1990 3-7% 3.5-7.5o/o 1 Ms. Seger dissented. She wanted to retain the 4-8 percent target ranges for M2 and M3 at that time in light of the uncertainty about the economic outlook. 2 Mr. Hoskins dissented. He believed that money growth should be toward the low end of the ranges if any reasonable progress were to be made in reducing inflation. In his view, establishing lower target ranges would emphasize the System's perseverance in reducing infla tionary pressures. 3 Mr. Keehn dissented. He favored reducing the target ranges to communicate the System’s adherence to its anti-inflationary commitment. grow th of the m onetary aggregates .4 The wider target ranges w ere intended to give the Com m ittee leeway to achieve its long-run objectives of maximum sustainable econom ic grow th and, eventually, price stability. Table 2 Target Ranges and Actual Money Growth in 1989 By the early July m eeting, M2 and M3 w ere growing at rates at or below the low er bounds of their respective target ranges. Through June, M2 grow th was estim ated to be 1 percentage point below its range and M3 grow th was estim ated to be at the low er end of its range. Nevertheless, in the context of recen t declines in m arket interest rates, the Board’s staff forecasted that the stronger grow th exhibited by these aggregates since the middle of May would continue, thereby bringing the grow th rates of M2 and M3 com fortably w ithin their target ranges by the fou rth quarter. Because it believed that these expected grow th rates would be compatible with its objectives fo r econom ic grow th and progress tow ard price stability, the Committee reaffirm ed its 1989 targets for M2 and M3 in the July R eport .5 4Record (May 1989), p. 357. The wide range of possible ef fects of future developments on thrift institutions and subsequent effects on money growth made the staff’s money growth forecasts subject to considerable uncertain ty. See Garfinkel (1989) for a discussion of the Commit tee’s motivation for adopting the wider M2 and M3 target ranges. 5Report (August 1989), pp. 528-29. Also see Record (October 1989), p. 693. It should be noted that all money growth rates reported here are revised figures, as of February 21, 1990, incorporating the February 1990 ben chmark and seasonal revisions. The reported numbers for M2 growth also incorporate the redefinition of M2 that now includes thrift overnight repurchase agreements. Digitized FEDERAL RESERVE BANK OF ST. LOUIS for FRASER Aggregate M2 M3 Target range1 Actual 3-70/0 3.5-7.5 4.6% 3.3 1 The target period for M2 and M3 growth is from IV/1988 to IV/1989. :r 21 Figure 1 Money Stock (M2) B illio ns o f do llars 3320 S easonally A djusted M onthly Averages o f D aily Figures B illio n s o f do llars 3320 3280 3280 3240 3240 3200 3200 3160 3160 3120 3120 3080 3080 3061.8 3040 3040 Nov 0 Dec 1988 The July Report also stated that the 1989 target ranges fo r M2 and M3 would be extended tentatively to 1990. W hile the Committee recog nized that a fu rth er reduction in the grow th of the m onetary aggregates would be m ore consis tent w ith attaining price stability over time, many m em bers believed that m ore rapid M2 grow th might be necessary to prom ote reasonable grow th in econom ic activity in 1990. Reductions in the ranges would increase the like lihood of making policy appear unpredictable by increasing the possibility of a reversal later or of having to tolerate grow th rates exceeding the low er target ranges. The targets for 1990 could be adjusted in February, if appropriate .6 Table 2 indicates that the actual rate of grow th in M2 during 1989, 4.6 percent, was close to the middle of its target range. The actual rate of grow th in M3, 3.3 percent, how ever, was slightly below the low er bound of its target range. T he grow th rates varied considerably Nov Dec 1989 during the year as illustrated in figures 1 and 2 , w hich show the monthly averages of (revised) daily figures, respectively, for M2 and M3. As dis cussed below , these variations had some in fluence on the Com mittee’s short-run policy decisions. SH O RT-R UN PO LIC Y OBJECTIVES Each year, the Committee holds eight regular ly scheduled m eetings to review incoming data and assesses the cu rren t econom ic environm ent and the prospects fo r the future course of the economy. Based on this inform ation, the m em b ers determ ine w hat changes, if any, should be made in short-run m onetary policy to achieve the Committee’s long-term goals. At the close of each meeting, the Committee issues a domestic policy directive to the Federal Reserve Bank of New York. This directive serves as the basis for day-to-day im plem entation of policy in the inter- 6Record (October 1989), pp. 693-94. MARCH/APRIL 1990 22 Figure 2 Money Stock (M3) Billions of dollars 4250, Seasonally Adjusted Monthly Averages of Daily Figures Billions of dollars 4250 7.5% 1 4200 - 4200 7 4150 4150 4100 4100 3.5% 4050 —■■ ■■ 4050 4000 4000 3950 3950 3900 3900 3901.9 3850 3850 •; m eeting period by the M anager for Domestic Operations, w ho is responsible for executing the directives. As usual, the directives issued during procedure, an instruction to increase the degree of pressure on reserve positions would imply a higher target for borrow ed reserves (adjustment 1 9 8 9 prim arily em phasized the d egree of r e plus seasonal borrow ings) and a h igh er fed eral straint on reserve positions (maintain, increase or decrease) that was considered by the m em bers to be consistent with the Com m ittee’s money grow th targets and goals for the economy. funds rate for a given discount ra te .7 Maintaining the approach used since O ctober 1982, the Federal Reserve System followed a borrow ed-reserves operating procedure. This procedure translates the degree of reserve re straint specified in the directive into a target for borrow ed reserves (reserves borrow ed from the Federal Reserve Banks). For example, under this H'he positive relation between borrowed reserves and the federal funds rate follows from economic theory. Specifically, the demand for borrowed reserves (the “ bor rowings function” ) is negatively related to the opportunity cost of borrowing from the discount window. This cost equals the spread between the discount rate (the interest rate paid by depository institutions for borrowed reserves RESERVE BANK OF ST. LOUIS FEDERAL Tow ard the end of 1988 and continuing into 1989, how ever, this operating procedure was com plicated by the unstable relation betw een the demand for borrow ed reserves and the federal funds rate. Specifically, the willingness of deposi tory institutions to borrow reserves, fo r given federal funds and discount rates, was declining unexpectedly. At the D ecem ber 1988 meeting, the Committee considered the possibility of ad justing the operating procedure to shift the focus from Federal Reserve Banks) and the federal funds rate (the interest rate paid for reserves borrowed from other depository institutions). For a discussion of the implemen tation of monetary policy under the borrowed-reserves operating procedure, see Gilbert (1985) and Thornton (1988). 23 of policy im plem entation to the federal funds rate, but the m em bers generally agreed that ad hering to the cu rren t procedure would be ap propriate given its advantages .8 N evertheless, the Committee believed that the u ncertainty about the relation of borrow ings to the federal funds rate w arranted flexibility in implementing m one tary policy. Hence, some of the directives issued in 1989 w ere w ritten with the understanding that flexibility would be perm itted in conducting day-to-day policy so as to achieve the Commit tee’s objectives, given the changing conditions in the m arket fo r borrow ed reserves. Fu rtherm ore, the Committee maintained the flexibility in short-run policy adopted in previous years because of u ncertainty about the relations of m onetary aggregate grow th to output grow th and inflation. T he Committee believed that shortrun policy should be decided not only on the basis of the behavior of the m onetary aggre gates, but on the basis of indicators of infla tionary pressures, econom ic grow th and the changing conditions in domestic financial and foreign exchange m arkets .9 In addition to the desired degree of reserve pressure, the directives indicated potential modifications in the interm eeting period, and the expected grow th rates of M2 and M3 condi tional on the desired reserve restraint. Each directive also established a m onitoring range for the federal funds rate. T he Chairman could in itiate a Committee consultation if, during the in term eeting period, the federal funds rate w ere to move out of that range. Over the past several years, how ever, such consultations have been initiated because of unexpected econom ic and financial developments. The following discussion review s each FOMC m eeting chronologically. It focuses on the im portant econom ic developments of 1989, show 8Record (April 1989), pp. 295-96. Also see, for example, Record (May 1989), p. 359. The particular advantage of the current operating procedure mentioned in the Commit tee’s discussion was that it permits “ greater scope for market forces to determine short-term interest rates.” [Record (April 1989), p. 296.] Such a procedure, however, can be less effective in maintaining short-term control over the money stock if the borrowings function is subject to unexpected shifts that are not quickly identified and reflected in changes to the borrowings assumption. See Thornton (1988) for a discussion on the advantages and disadvantages of this procedure. 9Report (March 1989), p. 108. ing how they influenced the Com mittee’s form u lation of short-run policy objectives. Table 3 sum m arizes the directives issued in 1989. Table 4 shows the actual (revised) intra-year grow th rates in M2 and M3, as well as those rates ex pected by the Committee. February 7-8 Meeting Econom ic data review ed at this m eeting sug gested that, abstracting from the direct im pact of the previous y ear’s drought, econom ic grow th continued at a fast pace. The m arked increase in total nonfarm payroll employment in Jan uary w as widespread and the civilian unem ployment rate of 5.4 p ercent was only m arginal ly above D ecem ber’s rate of 5.3 percent. Indus trial production rose sharply in D ecem ber and Jan uary and the industrial capacity utilization rate in January exceeded its average rate over the fourth qu arter of the previous y e a r .10 Indicators of inflation at the beginning of 1989 showed hardly any change from 1988. Al though the producer price index rose sharply in D ecem ber, the behavior of the consum er price index, excluding food and energy, was perceived to be in line with its pattern in 1 9 8 8 .11 Labor costs, particularly wages and salaries, rose appre ciably faster than one year earlier, however. As instructed by the Committee’s last directive in 1988, the degree of pressure on reserve posi tions was increased at the beginning of the p re vious interm eeting period. During this period, the average level of adjustm ent plus seasonal borrow ings was slightly above $500 million and the federal funds rate rose to about 9 percent from about 8 V2 percent. As oth er m arket in ter est rates rose, the grow th of the broader m onetary aggregates, particularly M2, weakened in Jan u ary .12 in December and January. The Board’s measure of the total industry capacity utilization rate in January was 84.3 percent, up from the previous quarter’s average of 84.1 percent. "Ibid., pp. 353-54. The annual growth rate of the seasonally adjusted producer price index for finished goods was 3.3 percent in December, while the seasonally adjusted con sumer price index for all urban consumers rose at an an nual rate of 4.1 percent. Excluding food and energy, the latter index rose at an annual rate of 4.9 percent. 12lbid., p. 354. In January, M2 rose sluggishly at an annual rate of 0.5 percent and M3 rose at an annual rate of 2.4 percent. 10Record (May 1989), p. 353. Industrial production rose at annual rates of 4.4 percent and 3.5 percent, respectively, MARCH/APRIL 1990 24 Table 3 The FOMC’s Short-Run Operating Ranges for 1989 Expected growth rates M2 M3 Degree of reserve pressure1 Intermeeting federal funds range Date of Meeting Target period February 7-8, 19892 December 1988March 1989 2% 3V2°/o maintain (+) 7-11% March 28, 19893 March-June 3 5 maintain (+) 8-12 May 16, 19894 March-June V/2 4 maintain 8-12 July 5-6, 1989= June-September 7 7 slightly reduce 7-11 August 22, 19896 June-September 9 7 maintain (-) 7-11 October 3, 19897 SeptemberDecember 6V2 4V2 maintain (-) 7-11 November 14, 19898 SeptemberDecember 71 /2 41/2 maintain (-) 7-11 December 18-19, 1989^ November 1989March 1990 81 /2 51 /2 slightly reduce 6-10 1A ‘ + ’ indicates an expectation that during the intermeeting period developments were more likely to warrant an adjustment toward restraint than toward ease. The opposite is true for a 2Messrs. Hoskins and Parry dissented. Mr. Hoskins thought that an immediate move toward greater monetary restraint would be appropriate to put policy on a course toward price stability in the longer run. Mr. Parry stressed that, since economic growth had exceeded its long-run, noninflationary rate, inflationary pressures were already increasing. In their view, without an immediate increase in restraint on reserve positions, inflationary pressures would intensify and thereby make the task of achieving the Committee's anti-inflationary goal more difficult. 3Ms. Seger dissented. Although maintaining the existing degree of reserve pressure was ap propriate in her view, she believed that the bias toward monetary restraint was undesirable in light of the lagged effects of appreciable tightening that had been undertaken earlier along with current indications of slower economic growth. 4Mr. Melzer dissented, advocating prompt action to ease the degree of reserve pressure slight ly. Pointing to the past two years of slow money growth, he stressed that the current high in flation would be reduced eventually and that, in the absence of an easing action, the risks of a recession would be augmented. Reaching the System’s non-inflationary goal would be hampered if, in response to a recession, monetary policy were to aim at a quick recovery. 5Ms. Seger dissented. She favored a greater degree of easing that, in her view, would be necessary to promote reasonable economic growth in the following years. 6Mr. Guffey dissented. He could accept an unchanged policy. But he believed that a directive that was biased toward ease was not appropriate given that the chances of a weakening of the economic expansion appeared to be essentially the same as the chances of a strengthen ing of the expansion while, in his view, the current and expected future inflation rates were not acceptable. Furthermore, such a bias toward ease could lessen confidence in the Commit tee’s commitment to achieve long-run price stability. FEDERAL RESERVE BANK OF ST. LOUIS 25 Table 3 continued The FOMC’s Short-Run Operating Ranges for 1989_________ 7Mr. Guffey and Ms. Seger dissented. Mr. Guffey believed that the bias toward easing reserve pressure in the inlermeeting period was not appropriate, for he was concerned that inflation in the future would remain unacceptably high. Ms. Seger pointed to signs of a weakening economy and indicated that some immediate easing would be necessary to sustain the expan sion. She believed that such a policy would be consistent with the Committee’s long-term ob jective of price stability. eMs. Seger dissented. In her view, signs of a weakening in economic activity, particularly in the manufacturing sector, warranted an immediate further easing of reserve pressure. 9Messrs. Angell and Melzer dissented, favoring an unchanged policy. Mr. Angell believed that a policy of easing reserve restraint based on recent indications of a weakening in economic ac tivity was not warranted. To the extent that the impact of monetary policy is realized with a lag, in his view, policy should be based on forward-looking indicators of economic activity. Further, he believed that an easing of policy at this time, given past policy, could have an un favorable impact on the System’s credibility in pursuing a goal of price stability. Mr. Melzer noted that policy had been eased considerably over the past six months and, given the suffi cient liquidity in the economy, he did not believe that further easing was desirable. He also was concerned that, in light of the current and projected growth of the monetary aggregates, easing of policy at this time would hamper the System’s ability to make progress toward its long-term goal of price stability. Table 4 Actual and Expected Rates of Money Growth M2 Period M3 Expected Actual Expected Actual December 1988March 1989 2% 1.9% 31 % /2 4.0% March 1989June 19891 1V2-3 1.9 4-5 2.9 June 19897-9 September 19892 8.2 7 2.7 September 19896V2-7V2 December 19893 7.7 4V2 3.2 1 the May 16 meeting, the Committee lowered its ex At pectation for M2 and M3 to 1V2 percent and 4 percent, respectively, from 3 percent and 5 percent. 2 the August 22 meeting, the Committee’s expecta At tions for M2 growth were revised from 7 percent to 9 percent. 3 The Committee’s expectation for M2 growth was revised from 61/2 percent to IV i percent at the November 14 meeting. T he Board’s staff expected that econom ic acti vity would continue expanding in 1989 at a m ore modest pace than it had in 1988. Assuming that m onetary policy would attem pt to contain the inflationary pressures predicted by the staff, the staff’s forecast pointed to pressures in financial m arkets that would tend to dampen grow th in domestic spending. In addition, the staff ex pected that foreign demands would provide less of an impetus to dom estic output grow th than in 1988. Given the relatively low m argins of unutilized labor and high capacity utilization rates, the staff predicted additional price pres sures in the cu rren t y e a r .13 In the Committee’s discussion of policy imple mentation, many m em bers indicated that eco nomic grow th appeared to be balanced, but most m em bers expressed con cern about the prospect of greater inflation. A m ajority of the m em bers, however, saw no need for an immediate change in policy in light of the data available for review, the appreciable recen t tightening in policy and their perception that the credibility of the Com m ittee’s com m itm ent to fight inflation was high. These m em bers w ere willing to wait for addi tional evidence confirm ing their fears of greater inflation b efore tightening policy fu rth e r .14 13lbid., pp. 354-55. 14lbid., p. 358. MARCH/APRIL 1990 26 Some m em bers advocated an immediate tight ening of reserve conditions to contain any future inflationary pressures. In their view, w ithout im mediate action, the task of achieving price stability could becom e m ore difficult. O ther m em bers expressed con cern that additional pressure on reserve positions could aggravate the financial conditions of many th rift institutions and highly indebted firm s. In addition, fu rth er re straint might add to the recen t unusual strength of the dollar .15 Although the recen t slow grow th of the m onetary aggregates was thought to indi cate future restrain t on price pressures, some m em bers cautioned that a shortfall from targeted ranges would be a m atter of co n cern .16 At the end of the meeting, the Committee adopted a directive that called for an unchanged degree of pressure on reserve positions, with a possible increase or decrease depending on fo rth coming inform ation about inflationary pressures, the strength of business expansion, grow th in the m onetary aggregates and developments in foreign exchange and domestic financial m arkets. As table 3 indicates, how ever, th ere was a “bias” tow ard restraint, and the m em bers called fo r "rem aining alert to potential developments that might requ ire some firm ing during the in term eeting period .”17 In light of the continuing u ncertainty about the relation of the demand fo r borrow ed reserves to the federal funds rate, the directive was issued w ith the explicit under standing that flexibility would be needed in implem entating m onetary policy. Given the con templated reserve conditions, the Committee ex pected the annual grow th rates for M2 and M3 to be around 2 percent and 3Vz percent, respec tively, from D ecem ber to M arch. The directive left the range for the federal funds rate un changed at 7 to 11 p ercen t .18 March 28 Meeting During the interm eeting period, additional pressure was placed on reserve positions, in 15lbid., pp. 358-59. As a measure of the relative strength of the dollar in foreign exchange markets, the Federal Reserve Board constructs a trade-weighted index using the currencies of Belgium, Canada, France, Germany, Ita ly, Japan, the Netherlands, Sweden, Switzerland and the United Kingdom. The trade-weighted index rose about 6 percent over the intermeeting period. 16lbid. 17lbid., p. 359. 18lbid., pp. 359-60. 19Record (July 1989), p. 503. M2 rose at annual rates of 1.8 percent and 3.5 percent, respectively, in February and RESERVE BANK OF ST. LOUIS FEDERAL light of incoming data indicating g reater infla tionary pressures. Also, on February 24, the Board of Governors approved a 50 basis-point increase in the discount rate to 7 percent. From the time o f the increase in the discount rate to this meeting, the federal funds rate rose nearly 75 basis points to slightly above 9% percent. O ther m arket interest rates, especially those on shorter-term securities, also rose. As the de mand for borrow ed reserves appeared to fall, the borrow ings assumption was lowered as a technical adjustm ent. The average of adjustm ent plus seasonal borrow ings during the six-week period just b efore this meeting fell to about $450 million. Although the m onetary aggregates appeared to gain some strength in February and M arch, their grow th was viewed as sluggish re lative to that in the previous y e a r .19 T he inform ation available fo r review at this m eeting suggested that econom ic activity ex panded considerably in the first quarter. Total nonfarm employment advanced sharply in February and the civilian unem ploym ent rate fell to 5.1 percent. Only part of the employment gain was attributed to the unusually mild w eath er during the first two m onths of the y e a r .20 T h ere w ere, how ever, indications of a slight w eakening of the econom ic expansion. For ex ample, prelim inary data suggested that industrial production rem ained flat and capacity utilization rates fell slightly in Feb ru ary .31 Furtherm ore, grow th in consum er spending slowed in the first two m onths of the year from its vigorous pace during the last q uarter of 1988. Although the data indicated that the nom inal U.S. m er chandise trade deficit improved in January, the value of exports fell. But the observed net rise of the trade-weighted value o f the dollar over the interm eeting period was attributed largely to the rise in m arket in terest rates stemming, in part, from restrictive actions taken during that period .22 March. M3 rose at annual rates of 3.3 percent and 6.2 percent, respectively, during those same two months. 20lbid., p. 502. The civilian unemployment rate in February has been revised upward to 5.2 percent. 2 Ibid. Revised data indicate that industrial production ac 1 tually fell in February at an annual rate of 2.5 percent. 22lbid., pp. 502-03. The trade-weighted index of the value of the dollar in foreign exchange markets rose approximately 1.5 percent over the intermeeting period. 27 Price pressures appeared to gain some strength in the first two m onths of 1989. The observed increases in price indexes w ere thought to be due chiefly to energy and food prices. Even excluding these com ponents, how ever, the producer and consum er price indexes rose sharply in Jan uary and Febru ary .23 The Board’s staff predicted that the pace of econom ic expansion would slow considerably from the pace in 1988. The forecast assumed that m onetary policy would restrain the infla tionary tendencies in the econom y. Such a policy could put additional pressure on financial m ar kets and involve slow er grow th in consum er spending and business fixed investm ent .24 T he m em bers generally agreed that, in light of mixed evidence about the strength of econom ic expansion and the uncertain prospects for the future, an unchanged policy would be accept able. Some m em bers, who p referred an immedi ate tightening of reserve conditions, believed that inflationary pressures could intensify given the apparent m om entum in econom ic activity. These m em bers, how ever, w ere willing to wait fo r additional inform ation that tended to confirm their fears. Most m em bers believed that the unusual strength of the dollar in foreign ex change m arkets would dampen price pressures. Furtherm ore, as suggested by earlier ex perience, the sluggish grow th in the m onetary aggregates lessened the likelihood that inflation could gain m uch strength in the fu tu re .25 As table 3 shows, the Committee adopted a directive that did not call fo r a change in policy but that perm itted a policy adjustm ent during the interm eeting period m ore readily toward restrain t than ease. Open m arket operations w ere to be conducted w ith some degree of flex ibility because of the continuing uncertainty about the relation of the demand of borrow ed reserves to the federal funds rate. Growth in M2 and M3 w ere expected to b e around 3 p er cent and 5 percent, respectively, from M arch to 23lbid. The seasonally adjusted producer price index for finished goods rose at annual rates of 13.9 percent and 7.8 percent, respectively, in January and February; ex cluding food and energy prices, it rose at annual rates of 6.2 percent and 7.2 percent, respectively, in January and February. Similarly, the seasonally adjusted consumer price index for all urban consumers rose at annual rates of 7.2 percent and 5.1 percent, respectively, during the first two months of the year; excluding food and energy, this index rose at annual rates of 5.9 percent and 4.8 percent. 24lbid., p. 504. Ju n e. The interm eeting range fo r the federal funds rate was increased 1 percentage point to 8 to 1 2 percent, "in light of the tightening of reserves since the February m eeting and the related increase in the federal funds ra te .”26 M ay 16 Meeting Aside from the slightly firm er reserve condi tions due to the greater-than-expected reserve flows related to April tax payments, reserve conditions hardly changed in the interm eeting period. During this period, the average of adjust m ent plus seasonal borrow ing rose to about $565 million, while the rate at w hich federal funds traded rose slightly to around 9-7/8 p er cent. O ther m arket interest rates, especially short-term rates, fell over the interm eeting period, however. Estimated grow th of the m one tary aggregates was sluggish, with the cum ula tive grow th o f M2 since the fou rth q uarter of 1988 well below the low er bound of the Com m ittee’s target range and that of M3 just above the lower limit of its target ran ge .27 Econom ic data reviewed at this meeting sug gested that the expansion of econom ic activity had m oderated in recen t months. Growth in total nonfarm employment edged downward in M arch and April, while the civilian unemploy m ent rate climbed from 5.0 percent in M arch to 5.3 percent in April. In addition, grow th in con sum er spending maintained the m uch slower pace established in the first part of the year relative to that in 1988. Industrial production grew in April, but, from D ecem ber to April, it grew m ore slowly than it had in 1988. Much of the April grow th was attributed to an increase in automobile assem blies after a w eak first qu arter and a rebound in the output of other consum er goods. Although the total capacity utilization rate rose slightly in April, it rem ained below its January ra te .28 T h ere w ere indications, how ever, that the momentum in econom ic activity had not been en- 25lbid., p. 505. 26lbid., p. 506. 27Record (September 1989), pp. 626-27. In April, M2 and M3 grew at annual rates of 1.0 percent and 2.6 percent, respectively. 28lbid., p. 625. The annual growth rate in the industrial pro duction index rose from 1.7 percent in March to 8.9 per cent in April. During that month, the total capacity utiliza tion rate rose 0.4 percentage points from the previous month to 84.2 percent. MARCH/APRIL 1990 28 tirely lost. Capital business spending rebounded after falling in the last q uarter of 1988. F u rth er m ore, although the nominal U.S. m erchandise trade deficit widened in February, the average deficit fo r the first tw o m onths of 1989 rem ain ed below that fo r the fou rth q uarter of 1988, w ith the value of exports grow ing m ore rapidly than that of imports. Despite the slight deteriora tion in the external trade balance in February and the general downward m ovement in interest rates m ore recently, the dollar gained fu rth er strength in the interm eeting period .29 w ere sufficiently restrictive to contain futu re in flationary pressures without precipitating an ex cessive slowing of econom ic grow th rem ained unclear. Although one m em ber believed th at an immediate easing of reserve pressure would be both necessary and desirable to improve the prospects for adequate m onetary grow th to sus tain the econom ic expansion, others feared the risks associated with such a policy—that is, of having to reverse the easing if the m onetary ag gregates w ere to accelerate unduly and price pressures w ere to intensify later .32 The recen t behavior of the price indexes did' not ease the m em bers’ fear of future inflation. Rather, price level m ovements w ere interpreted by the m em bers as an indication that inflation ary pressures w ere rooted deeply in the econo my. Although the producer price index grew m ore slowly in M arch and April than in the earlier two m onths of the year, the consum er price index grew at a slightly faster pace in the first q uarter of 1989 than in the previous quarter. Increases in food and energy prices contributed to the observed increases in m ea sured inflation, but w ere not the sole driving fo rce of the perceived upward tren d in infla tion .30 In the discussion about possible adjustm ents to m onetary policy in the interm eeting period, most m em bers agreed that no bias—eith er toward restrain t or ease—would be appropriate. W hile one m em ber believed that policy should be par ticularly alert to behavior of the m onetary aggre gates that could w arran t some easing, others believed that the deeply rooted inflationary pressures called for a bias tow ard restraint. A n um ber of m em bers expressed con cern that the absence of a bias tow ard restrain t might give an in correct signal that the Committee was moving away from its anti-inflationary com m itm ent .33 At the end of this m eeting, the Committee issued a directive that called fo r an unchanged degree of pressure on reserve positions. D epen ding on forthcom ing inform ation, a move to some restrain t or ease would be acceptable dur ing the interm eeting period. T he Committee believed that continuing u ncertainty about the relation of the demand fo r borrow ed reserves to the federal funds rate w arranted continuing flexibility in the im plem entation of m onetary policy. The Committee expected that the contem plated reserve conditions would be consistent with M2 and M3 growing at IV 2 and 4 percent annual grow th rates, respectively, from M arch to June. The interm eeting range for the federal funds rate was kept at 8 to 1 2 p ercen t .34 The sta ffs projection changed little from that prepared fo r the previous meeting. Grow th in econom ic activity was expected to b e slow er than in 1988. T he forecast indicated that prices at both the consum er and producer levels would increase at somewhat faster rates in 1989. In the staff's view, m onetary policy that at tempted to contain such inflationary pressures, should they m aterialize, would imply greater pressure on financial m arkets. In addition to a continuation of sluggish grow th in consum er spending, the staff expected that grow th in busi ness capital spending would retreat from its fast pace in the first q u arter .31 U ncertainty about the im pact of previous re strictive policy actions on inflation and the pace of econom ic grow th dominated the discussion at this m eeting. W h ether m onetary conditions “ Ibid., pp. 625-26. The value of the dollar relative to the other G-10 currencies appreciated about 4 percent over the intermeeting period. 30lbid., p. 626. The seasonally adjusted consumer price in dex for all urban consumers rose at annual rates of 6.1 percent and 8.1 percent, respectively, in March and April. Excluding food and energy prices, this price index rose 4.8 percent and 2.9 percent. The producer price index for FEDERAL RESERVE BANK OF ST. LOUIS July 5-6 Meeting Late in the interm eeting period, incoming in form ation tended to confirm earlier indications finished goods, excluding food and energy, rose at annual rates of 2 percent and 1 percent. 3 Ibid., p. 627. 1 “ Ibid., pp. 628-29. 33lbid., p. 629. ^Ibid. 29 that the econom ic expansion had slowed so that the prospect of w eakening inflationary pressures seemed m ore promising. Fu rtherm ore, the m on etary aggregates continued to exhibit slow grow th and the dollar had gained considerable strength earlier in the interm eeting period. Ac cordingly, a slight lowering of the pressure on reserve positions was sought. Before this easing, however, a technical upward revision had been made to accom m odate unusual strength in sea sonal borrow ing. Over the six-week period end ing Ju n e 27, the average of adjustm ent plus seasonal borrow ings was around $550 million, and the federal funds rate edged down to 9% percent. O ther m arket in terest rates, especially those on long-term securities, also fell. The ob served decline in the level of the broader m one tary aggregates during May was interpreted as a reversal of the tem porary rise in transaction accounts related to April tax paym ents .35 Confirming earlier evidence, the inform ation available fo r review at this meeting suggested that the econom ic expansion had slowed consid erably from its pace in 1988. W hile the civilian unemploym ent rate fell to 5.2 p ercen t in May, grow th in total nonfarm em ploym ent was rela tively weak. Prelim inary data indicated that, in May, grow th in industrial production was modest and the total capacity utilization rate had fallen back to its M arch level .36 W hile business capital spending appeared to m ake fu rth er gains in the second quarter, grow th in consum er spending rem ained sluggish. Further, the significant im provem ent in the nominal U.S. m erchandise trade balance during April stem m ed chiefly from a considerable drop in im ports with only a slight increase in exports .37 Price pressures persisted despite the indica tions of slowing econom ic expansion. Increases in food and energy prices, how ever, made large contributions to the increases in the producer price index and, to a lesser extent, in the con 35Record (October 1989), p. 691. Revised data indicate that in May, M2 declined at an annual rate of 1.6 percent and M3 rose sluggishly at an annual rate of 0.2 percent. In June, the annual growth rates in M2 and M3 rebounded to 6.5 percent and 6.0 percent, respectively. 36lbid., p. 689. Revised data indicate that the industrial pro duction index fell at an annual rate of 0.8 percent in May, while the total capacity utilization rate in May, which fell to 84.0 percent from 84.2 percent in April, was above the rate of 83.8 percent in March. 37lbid., pp. 689-90. Despite the improvement in the external balance, the value of the dollar relative to the other G-10 currencies fell on net about 3 percent over the in sum er price index .38 Nevertheless, the grow th in labor costs appeared to have maintained its momentum from the middle of 1988. The Board’s staff revised its forecast for econom ic grow th in the second half of the year downward from that made earlier in the year. The staff's forecast now suggested less inflation than was previously expected, though m ore in flation than had been experienced in 1988, and continued grow th in labor costs in 1989. This inflation outlook took account of the persistent strengthening o f the dollar that was expected to dampen inflationary pressures. T he forecast, how ever, also pointed to slightly m ore favorable inflationary conditions in 1990 than w ere previously expected .39 In the context of a w eaker outlook for econom ic grow th, the m em bers generally believ ed that a fu rth er reduction in the degree of pressure on reserve positions would be ap propriate. Although th ere was some disagree m ent about the timing and the extent of such easing, most m em bers agreed that they could accept an immediate slight reduction in reserve pressure. In the view of many m em bers, a greater move tow ard ease could have an unde sirable effect on inflationary expectations, th e re by putting upward pressure on long-term in ter est rates. A substantial move tow ard ease might have to be reversed if inflationary pressures subsequently intensified. Nearly all believed, how ever, that the easing should be implem ented immediately given the slowing pace of econom ic expansion and the sluggish grow th o f the broader m onetary aggre gates. Although some m em bers p referred a direc tive that was biased tow ard restrain t to main tain the credibility of the Com mittee’s anti-infla tionary com m itm ent despite the easing of policy, others advocated a bias tow ard ease to com m unicate the Com mittee’s belief that the risks termeeting period. After having risen sharply in the first half of the period, the trade-weighted index of the value of the dollar declined appreciably over the second half. Mlbid., p. 690. But, even excluding food and energy prices, the seasonally adjusted producer price index advanced sharply, rising at annual rates of 7.2 percent and 8.2 per cent, respectively, in May and June. Similarly, the seasonally adjusted consumer price index, excluding food and energy components, rose 5.8 percent and 2.8 percent, respectively, in May and June. Mlbid., p. 691. MARCH/APRIL 1990 30 of an undesirable shortfall in econom ic grow th w ere substantial. Most m em bers agreed that, given the prevailing uncertainty, they could ac cept an unbiased directive .40 As table 3 shows, the directive issued by the Committee at the close of this meeting called for an immediate and slight reduction in the degree of reserve pressure. Fu rth er easing or some tightening was considered to be ap propriate depending on future developments. Conditional on the contem plated reserve condi tions, the Committee expected that both M2 and M3 would grow at an annual rate of 7 percent from Ju n e to Septem ber. Given the easing of reserve pressure in early June and that specified in this directive, the m onitoring range for the federal funds rate was lowered 1 percentage point to 7 to 11 p ercen t .41 August 22 Meeting As instructed by the Committee at the close of the previous m eeting, the degree of pressure on reserve positions was reduced at the beginning of the interm eeting period. Tow ard the end of July, the degree of reserve restraint was eased fu rth er, in light of incoming data that indicated a continued w eaker econom ic expansion and a slight reduction in inflationary pressures. At the beginning of the interm eeting period, however, the assumed level of adjustm ent plus seasonal borrow ing was increased as a technical revision prompted by a projected rise in seasonal b o r rowing during the sum m er months. Hence, the average of adjustm ent plus seasonal borrow ings over the six-week period ending August 22 rose to approxim ately $600 million despite the easing actions taken during this period. Nevertheless, the federal funds rate fell 50 basis points to around 9 percent. Prelim inary data indicated that, in July, grow th in the m onetary aggregates gained considerable strength, w hich appeared to continue into August .42 40lbid., p. 695. 4 Ibid., p. 696. In this instance, there was no mention of the 1 uncertainty revolving around the relationship between bor rowings and the federal funds rate and, therefore, no reference to flexibility in monetary policy implementation. 42Record (December 1989), p. 813. Growth in M2 ac celerated from 6.5 percent in June to 10.3 percent in July. The annual growth rate in M3 rose less dramatically from 6.0 percent in June to 6.9 percent in July. 43lbid., p. 812. Revised data, however, indicate that in dustrial production rose at an annual rate of 3.4 percent in FEDERAL RESERVE BANK OF ST. LOUIS The data reviewed at this meeting reinforced the earlier evidence of a m oderate econom ic ex pansion. The data, how ever, suggested less w eak ness in the expansion than they had tow ard the end of July. Nonfarm payroll em ploym ent made considerable advances in Ju n e and July. T he ci vilian unemployment rates for these months, 5.3 percent and 5.2 percent, respectively, w ere close to the average unem ploym ent rate during the first five m onths of the year. In addition, prelim inary data indicated that industrial produc tion rebounded in July after having fallen in May and Ju n e .43 Industrial capacity utilization maintained its high rate, although the rate for m anufacturing in July was well below that in January. M oreover, grow th in consum er spend ing in the second qu arter was stronger than ori ginally estimated, and the observed narrow ing of the nominal U.S. m erchandise trade deficit reflected not only a notable decline in the value of imports, but a m arked jump in the value of exports .44 T he recen t behavior of price indexes sug gested somewhat less inflation prim arily because of appreciable declines in food and energy prices. Prelim inary data indicated that, while the consum er price index rose in both Ju n e and July, the increases w ere modest, and the producer price index for finished goods fell .45 W age grow th over the past several m onths did not appear to deviate from previous ly established trends. The staff expected that, during the rest of the year, grow th in the nonfarm econom y would maintain its pace from the first h alf of the year and then grow m ore slowly in 1990. W ith in ter est rates falling since the spring and the recen t ly observed substantial job gains, consum er spending was expected to exhibit greater strength in the coming months. The forecast in dicated that business capital spending would continue to make a large contribution to econom ic grow th. Partly because of the earlier strength- June and fell at an annual rate of 0.8 percent in July. The July civilian unemployment rate has been revised upward to 5.3 percent. “ Ibid., pp. 812-13. 45lbid., p. 813. The seasonally adjusted producer price index for finished goods rose at an annual rate of 1.1 percent in June and fell at an annual rate of 4.1 percent in July. But, excluding the food and energy prices, this index rose at an annual rate of 8.2 percent in June and fell at an annual rate of 1.9 percent in July. 31 ening of the dollar in foreign exchange m arkets, how ever, foreign trade was not expected to be a significant source of econom ic grow th. In ad dition, although expected fu rth er declines in food and energy prices suggested that price pressures could w eaken in the coming quarter, the staff expected no substantial im provem ent in the inflationary trend through 1990.46 3 indicates. Despite some m em bers' reserva tions, the directive included a bias tow ard ease. The Committee expected M2 and M3 to grow at annual rates of about 9 percent and 7 percent, respectively, from Ju n e to Septem ber. T he intermeeting range for the federal funds rate was kept at 7 to 11 p erce n t .49 W ith evidence that the econom ic expansion had stabilized at a “provisionally acceptable pace” and that inflationary pressures w ere not gaining strength, the m em bers generally believed that the cu rren t degree o f reserve pressure should be maintained, at least in the early part of the interm eeting period. An unchanged course for policy was also justified by the observation that grow th in M2 and M3 recently had gained suffi cient strength to place these aggregates in their target ranges .47 O ctober 3 Meeting Discussing possible adjustm ents in policy dur ing the interm eeting period, many m em bers ex pressed the belief that, if futu re developments w ere to w arran t a change in policy, the direction of change would m ost likely be tow ard some ease. Some m em bers, how ever, p referred not to incorporate such a presum ption in the directive. In their view, the "risks to the econom y w ere m ore evenly balanced.” That is, the direction of change in policy justified by developments in the interm eeting period was just as likely to be tow ard restraint as it was tow ard ease. Further, these m em bers believed that a bias tow ard ease could “lead to a m isreading of System policy in the context of an unacceptably high rate of in flation .”48 The directive issued at the end of this meeting specified no immediate change in policy, as table 46lbid., p. 814. Over the intermeeting period, the value of the dollar relative to the other G-10 currencies rose ap proximately 2.7 percent, almost offsetting the previous net decline. Even so, the trade-weighted value of the dollar was below the highs reached in June. 47lbid., pp. 815-16. Although the staff predicted that M2 and M3 growth would slow considerably from the current pace, the growth in the aggregates was expected to remain com fortably within their target ranges. These forecasts for money growth as well as those made subsequently in 1989, however, were subject to great uncertainty as a result of the uncertainty revolving around the resolution of thrift institution insolvencies and the responses of thrift in stitutions to recently enacted legislation. These factors were expected to dampen growth in the broader monetary aggregates, particularly that in M3. Thus, any observed weakness in the growth of these aggregates would not be interpreted as evidence of a slowing economy. Ibid., p. 816. Over the interm eeting period, reserve condi tions displayed no noticeable change. The aver age o f adjustm ent plus seasonal borrow ing dur ing the fou r weeks ending Septem ber 20 fell slightly to about $550 million, and the federal funds rate fluctuated within a narrow range centered around 9 percent. Although M2 grow th was strong, M3 grow th had unexpected ly lost some of its strength in August and prelim inary data suggested that this slow er grow th had continued into Septem ber .50 T he data available fo r review at this meeting reaffirm ed earlier projections, that the econom ic expansion had continued at a m oderate pace in the third quarter. Nonfarm payroll employment generally made considerable advances after al lowing fo r the effects of strike activity. Neverthe less, there w ere hardly any job gains in m anu facturing industries, and the civilian unemploy m ent rate in August and Septem ber was close to 5 V4 percent. Further, after increasing m oderate ly in August, industrial production fell slightly in Septem ber .51 T he industrial capacity utiliza tion rate, how ever, rem ained relatively high. W hile grow th in business capital spending seemed to have slowed in the third quarter, con sum er spending continued to exhibit consider able strength. W ith the value of im ports declin- 48lbid., p. 816. 49lbid., pp. 816-17. 50Record (January 1990), pp. 18-19. The slowing of the growth of the monetary aggregates was especially evident in M3. M2 grew at annual rates of about 7.8 percent and 6.5 percent, respectively, in August and September; M3 grew at an annual rate of 1.4 percent in August and was flat in September. 5 Ibid., p. 17. The annual rate of growth of the industrial pro 1 duction index fell from 5.2 percent in August to -1.7 per cent in September. Revised data indicate that the civilian unemployment rate in August and September was 5.3 percent. MARCH/APRIL 1990 32 ing by m ore than the value of exports, the U.S. m erchandise trade deficit im proved fu rth er in Ju ly .52 The price indexes continued to indicate a low er rate of inflation. In August, producer prices fell and the consum er price index was unchanged .53 The upward trend in labor costs, however, did not appear to change on a year-toy ear basis. The staff’s forecast fo r econom ic grow th in the rem aining part of 1989 and 1990 w ere es sentially unchanged from those made for the previous m eeting. Growth in business capital spending was expected to slow from its pace in the first half of the year, how ever .54 Most m em b ers believed that, although econom ic activity would continue to expand in the coming quar ters, the pace of the expansion would m ore like ly slow than build m omentum. W hile the m em b ers generally expected some w eakening in in flationary pressures, only a few thought this w eakening might b e appreciable. A num ber of m em bers expressed con cern that progress would be constrained considerably if econom ic activity w ere to build m omentum. Furtherm ore, the m em bers believed that the recen t fall of the value of the dollar in foreign exchange m arkets would add to future upw ard pressure on p rices .55 Most m em bers thought that an unchanged policy would be appropriate in the near term . The focus of policy continued to be that of gradually reducing inflation over time and a steady policy course seemed consistent with that objective, at least for the time being .56 “ Ibid., pp. 17-18. 53lbid., p. 18. The decline in producer prices, however, was driven largely by a continued decline in energy prices. In August, the seasonally adjusted producer price index for finished goods fell at an annual rate of about 3.1 percent, but, excluding energy and food, this index rose at an an nual rate of 6.1 percent. The seasonally adjusted con sumer price index excluding energy and food, however, rose only 1.9 percent in August. In their discussion about the outlook for inflation, the members commented that the recent declines in food and energy prices that had dampened price inflation might be temporary. Ibid., p. 20. S'Hbid., p. 19. 55lbid., pp. 19-20. The trade-weighted value of the dollar relative to the other G-10 currencies fell 2.7 percent over the intermeeting period. A fall in the value of the dollar, holding all else constant, increases the attractiveness of U.S.-produced goods to foreign importers and U.S. in dividuals. The resulting shift in demand can create domestic price pressures. A fall in the value of the dollar FEDERAL RESERVE BANK OF ST. LOUIS Growth in the m onetary aggregates was ex pected to m oderate from the rapid pace since the middle o f the year, given an unchanged policy. Most m em bers believed, how ever, that future developments would m ore likely require ease than restrain t in the interm eeting period. Nevertheless, the recen t depreciation of the foreign value of the dollar w arranted caution in undertaking any easing adjustm ents .57 The directive issued at the end of this meeting was w ritten with the understanding that a downward technical adjustm ent to the b orrow ings objective might be appropriate, if, as ex pected, seasonal borrow ings w ere to drop in the interm eeting period. T he reserve conditions contem plated by the m em bers w ere thought to be consistent with M2 and M3 growing at an nual rates of 6 V percent and 4 V2 percent, 2 respectively, betw een Septem ber and D ecem ber. The monitoring range fo r the federal funds rate was unchanged at 7 to 11 p ercen t .58 N o v e m b e r 14 Meeting Reserve conditions w ere eased in mid-October. For a short period after the sharp drop in stock prices on O ctober 13, an accommodative provi sion of reserves was undertaken while financial m arkets rem ained highly sensitive and volatile. Around the same time, in keeping with the previous m eeting’s directive, a decision was made to implement some easing on a m ore p er m anent basis. Incoming data, indicating an in creased risk of a w eakening in the business ex pansion, also prompted additional easing early in November. Furtherm ore, in light of a p er ceived decline in adjustm ent plus seasonal bor- can also increase the costs of production for those U.S. firms relying heavily on imported intermediate goods, thereby creating additional price pressures. See Hafer (1989) for a detailed discussion of the link between infla tion and a dollar depreciation. 56lbid., p. 20. There was also a concern that, given the re cent G-7 meeting, an easing of policy would be mistakenly interpreted as an action to lower the value of the dollar. The Committee believed that monetary policy should not be used as an instrument for achieving a given objective for the dollar in foreign exchange markets if that objective were not compatible with domestic policy objectives. In the view of some members, if recent intervention by G-7 and other nations were to result in a lower value of the dollar, the inflationary consequences would hamper the Commit tee’s ability to achieve its long-run goal of price stability. Ibid., pp. 20-21. 57lbid., p. 21. 58lbid., pp. 21-22. 33 rowings, several technical adjustm ents in the borrow ing assumption w ere made during the in term eeting period. From early O ctober to this meeting, actual borrow ings fell from about $635 million to about $200 million. W ith most m arket interest rates falling, the federal funds rate de clined from about 9 percent to 8 V percent in 2 the interm eeting period and grow th in the m on etary aggregates gained strength in O ctober .59 The data review ed at this m eeting indicated that the econom ic expansion had continued at a m oderate pace. Nonfarm em ploym ent gains w ere considerable in O ctober and the civilian unem ploym ent rate did not budge at 5.3 p er cent. The data also suggested, however, that the strength of expansion was not evenly distributed throughout the econom y. For example, most employment gains occu rred in the service sec tor, while m anufacturing employment declined. In addition, industrial production dropped appre ciably in O ctober, though m uch of the decline was attributed to several incidents that tended to disrupt production tem porarily (the Boeing strike, the earthquake and the h u rrican e ).60 The data also showed that retail sales had fallen and the grow th of business capital spending had weakened. Fu rtherm ore, w ith the value of im ports rising and the value of exports falling in August, the U.S. m erchandise trade deficit had risen to its highest level thus fa r in 1989.61 The recen t behavior of price indexes w ere consistent w ith a slight reduction in inflationary pressures. The percentage rise in producer prices fell in O ctober and, excluding energy and food, had hardly changed .62 But the data did not suggest any slowing in the grow th of labor costs. In light of the tem porary disruptions to pro duction, the staff’s forecast pointed to a fu rth er slowing in grow th in the fourth qu arter and a rebound in the first qu arter of 1990. On net, 59Report (February 1990), p. 56. The annual rates of growth of M2 and M3 rose to 7.1 and 1.4 percent, respectively, in October. This acceleration was not as pronounced as that in M1 whose annual growth rate rose from 3.9 percent in September to 8.3 percent in October. 60lbid., p. 55. The industrial production index fell at an an nual rate of about 4.1 percent in October. 6 Ibid., pp. 55-56. 1 the staff predicted that econom ic grow th would continue at a sluggish pace in the coming quar ters. Although continuing grow th in consum er demand was expected to contribute to econom ic activity in the near term , consum er demand was expected to w eaken subsequently. Further, the forecast indicated that the sluggish pace in the grow th o f business capital spending would continue and that n et exports would not m ake a significant contribution to the econom ic expan sion. The sta ffs forecast did not suggest, how ever, any substantial im provem ent in the u nder lying trend in inflation .63 Most m em bers agreed that the data pointed, on balance, to a sustained econom ic expansion, although grow th had w eakened recently. But th ere was no strong consensus among the m em b ers about the future outlook. W hile some m em b ers expected that the risks of a stronger-thandesirable expansion and a w eaker expansion w ere evenly balanced, others expected a greater likelihood of eith er a stronger or considerably w eaker econom ic expansion activity, and still others believed that the chances of an econom ic expansion close to the econom y’s potential in the future w ere not rem ote .64 Similarly, some m em bers believed that progress on improving the underlying inflation trend might be achieved given the recen t behavior of the price level in dexes and oth er factors, though others saw that such progress, if any, would be small over the next several q u arters .65 Although the econom ic expansion appeared to be slowing, most m em bers advocated a steady policy with no immediate change in the degree of pressure on reserve positions. Such a policy was considered to be consistent w ith the Com m ittee’s goals of promoting a sustained econom ic expansion while making progress tow ard reduc ing inflation in the long run. M oreover, m em bers believed that the recen t and expected grow th in the m onetary aggregates did not w arran t any “ Ibid., p. 57. ^Ibid., p. 57. The recent depreciation of the dollar in foreign exchange markets was expected by some members to provide a source of improvement in the nation’s trade deficit, especially in light of the observed strength in economic activity experienced recently by other industrial nations. Such an improvement would provide additional strength to the U.S. economic expansion. 65lbid., pp. 57-58. 62lbid., p. 56. Revised data indicate that the seasonally ad justed producer price index for finished goods rose at an annual rate of 6.5 percent in October. Excluding food and energy components, it rose at a 2.0 percent annual rate. MARCH/APRIL 1990 34 adjustm ent in policy. Hence, as table 3 shows, the directive issued at the close of this meeting did not call fo r any change in policy .66 Most m em bers, how ever, believed that the possibility of w eakening in the econom ic expan sion exceeded the possibility of excessive grow th and, accordingly, that future econom ic developments would m ore likely w arran t subse quent easing actions than tightening actions in the interm eeting period. Those m em bers who believed that the likelihood of excessive grow th was evenly balanced against the likelihood of w eakening in the expansion indicated that they could accept a directive containing a bias tow ard ease in the interm eeting period. But some em phasized the need for approaching possible eas ing adjustm ents w ith caution so as not to detract from any progress that could be made in eventually approaching the Com m ittee’s goal of reasonable price stability .67 The m em bers expected M2 and M3 to grow at annual rates of 7Vi and 4 V2 percent, respective ly, betw een Septem ber and D ecem ber. The m on itoring range for the federal funds rate was m aintained at 7 to 11 p ercen t .68 D ec em b er 18-19 Meeting During the interm eeting period, policy aimed to maintain a steady (or unchanged) degree of reserve restraint. Technical adjustm ents in the borrow ings assumption w ere made tw ice in the period in light o f ongoing declines in seasonal borrow ing. During the first tw o weeks of De cem ber, adjustm ent plus seasonal borrow ings averaged about $130 million, down from the average of about $400 million during the two previous w eeks. Meanwhile, the federal funds rate rem ained at about 8 V2 percent and other m arket interest rates changed little during most of this period. Prelim inary data indicated that the grow th in the broader m onetary aggregates picked up during November and rem ained ro bust in the first part of D ecem ber .69 66lbid., p. 58. 67lbid., pp. 58-59. 68lbid., pp. 59-60. 69Record (Federal Reserve Press Release, February 9, 1990), pp. 4-5. The annual rate of growth in M2 increased slightly to 7.5 percent in November, while the annual growth rate in M3 nearly tripled to 4.0 percent. 70lbid., p. 1. The annual growth rate of the industrial produc tion index rose to 3.4 percent in November. It should be noted that the November civilian unemployment rate has been revised to 5.3 percent. FEDERAL RESERVE BANK OF ST. LOUIS T he inform ation available fo r review at this meeting suggested that the econom ic expansion in the fourth q uarter had slowed from its pace earlier in the year. Although total nonfarm pay roll employment made considerable gains in November, these gains w ere concentrated in the service, trade and financial sectors, w ith conti nuing losses in m anufacturing. The November civilian unemploym ent rate, 5.4 percent, was at its highest level since January. Industrial pro duction in November rebounded from its p re vious decline driven by earlier strike activity among oth er facto rs .70 Upon adjusting for these factors, industrial production appeared to have fallen, on average, in recen t months. In ad dition, although nominal retail sales rebounded in November, sales had hardly changed from their average in the third quarter, and data in dicated a w eakening in business capital spend ing. W ith im ports up sharply and exports virtual ly unchanged, the nom inal U.S. m erchandise trade deficit rose considerably in O ctober after having fallen slightly in Septem ber .71 Estimated movements in price indexes contin ued to suggest a slight w eakening in inflationary pressures. For example, the producer price in dex, based on prelim inary data, fell in November. This decline, how ever, was partly attributable to sharp reductions in energy p rices .72 Although average hourly earnings had fallen in November, the underlying trend in labor cost grow th was not expected to change given the results of the recen t collective bargaining activities .73 The staff’s forecast had not changed substan tially from the previous meeting. It pointed to a slowing in the econom ic expansion in the fourth qu arter with a rebound in the first q uarter of 1990. The magnitude of the rebound was ex pected to be limited by anticipated declines in m otor vehicle production. Econom ic grow th for the rest of 1990 was expected to be driven pri marily by m oderate grow th in consum er spend ing. Net exports w ere expected to m ake a small 71lbid., p. 2-3. Total industry capacity utilization having not changed in November from October at 83.1 percent was 1 percentage point below its level a year earlier. 72lbid., pp. 3-4. Revised data indicate that the seasonally ad justed producer price index actually rose at an annual rate of 1.1 percent in November. Excluding food and energy, this index rose at an annual rate of 3.0 percent. 73lbid., p. 4. 35 contribution to the econom ic expansion in 1990. Further, while the staff anticipated that pres sures on labor and other resources fo r produc tion would lessen slightly, no large changes in the underlying trend of inflation w ere ex pected .74 The m em bers generally agreed that th ere was considerable evidence that the econom y’s grow th had w eakened and would likely rem ain at a sluggish pace in the n ear term . Although a num ber of m em bers thought that some strength ening in the econom ic expansion in 1990 was a reasonable expectation, most believed that the chances of a w eakening w ere "sufficiently high to justify an immediate move to slightly easier reserve conditions .”75 Those advocating this poli cy change believed that such a move would not jeopardize the System ’s credibility of adhering to its long-run goal of price stability, as price pressures and business conditions appeared to have weakened. Others less optimistic about the potential pro gress tow ard reducing inflation favored an un changed policy. Skepticism about this progress was partly driven by the recen t decline of the dollar and the possibility that, if econom ic activi ty w ere to rebound in the next year, inflationary pressures could gain considerable stren gth .76 Those advocating an unchanged policy em pha sized that m aintaining cu rren t reserve condi tions would be sufficient to ensure a continua tion of the expansion with an easing of pressure on productive resources, and that “fu rth er eas ing might overcom pensate fo r cu rren t w eakness in the econom y at the cost of delaying progress toward price stability .”77 Nevertheless, most of these m em bers, recognizing the risks of an ad ditional w eakening in the econom y, could ac cept a policy that sought an immediate but slight easing of the degree of pressure on re serve positions. In their view, given such a poli cy, it was highly unlikely that fu rth er easing would be w arranted during the interm eeting period .78 As indicated in table 3, at the close of this meeting, the Committee issued a directive call ing fo r a slight easing of reserve conditions. 74lbid., p. 5-6. This directive did not reflect a presumption about the direction o f possible adjustm ents in the interm eeting period. The Committee ex pected that the annual rates of grow th of M2 and M3 would be 8 V2 percent and 5 V2 percent, respectively, from November 1989 to M arch 1990. In addition, given the easing of reserve conditions in recen t m onths and the fu rth er easing stipulated in this directive, the Committee lowered the monitoring range for the federal funds rate 1 percentage point to 6 to 1 0 p er cen t .79 CONCLUSIONS During 1989, the econom ic data available for review at FOMC m eetings prompted Committee m em bers to shift th eir prim ary con cern from the risks of inflation to the risks of a slowdown in econom ic activity. At the beginning of the year, the th reat of a w orsening in the underly ing inflationary tren d drove the form ulation of policy. As the evidence of a w eakening economic expansion accum ulated and the outlook for in flation appeared to becom e less threatening, the Committee becam e m ore sensitive to the risks of a future slowdown in econom ic activity. Because the Committee understood that its in terpretation of the data was unavoidably subject to great uncertainty, how ever, it took what it perceived to be a conservative approach to re acting to this inform ation in an effort to balance these risks. This approach was also motivated by the Com mittee’s ultim ate goal of eventually achieving reasonable price stability and its desire to maintain its own credibility as an inflationfighter. REFERENCES Garfinkel, Michelle R. “ The FOMC in 1988: Uncertainty’s Ef fects on Monetary Policy,” this Review (March/April 1989), pp. 16-33. Gilbert, R. Alton. “ Operating Procedures for Conducting Monetary Policy,” this Review (February 1985), pp. 13-21. Hafer, R.W. “ Does Dollar Depreciation Cause Inflation?” this Review (July/August 1989), pp. 16-28. Hafer, R.W., and Joseph H. Haslag. “ The FOMC in 1987: The Effects of a Falling Dollar and the Stock Market Col lapse,” this Review (March/April 1988), pp. 3-16. Thornton, Daniel L. “ The Borrowed-Reserves Operating Pro cedure: Theory and Evidence,” this Review (January/ February 1988), pp. 30-54. 77lbid., p. 10. 75lbid., p. 10. 78lbid., pp. 11-12. 76lbid., pp. 9-10. Over the intermeeting period, the tradeweighted value of the dollar relative to the other G-10 cur rencies fell about 2.8 percent. 79lbid., pp. 13-15. MARCH/APRIL 1990 36 Alison Butler Alison Butler is an economist at the Federal Reserve Bank of St. Louis. Lora Holman provided research assistance. The author would like to thank William Barnett and Jennifer Ellis for helpful comments and suggestions. A M ethodological A p proach to Chaos: Are E con om ists M issing th e P oin t? "A very slight cause which escapes our notice determines a considerable effect which we cannot fail to see, and then we say that this effect is due to chance.” t —Poincare T h e r e IS INCREASING interest among econ omists in a new field of study that may offer an alternative explanation for the seemingly random behavior of many economic variables. This research, which originated in the physical and biological sciences, concerns a phenomenon called deterministic chaos.1 Contrary to the common usage of the word, chaos in this context describes the behavior of a variable over time which appears to follow no apparent pattern but in fact is completely deter ministic, that is, each value of the variable over time can be predicted exactly. In fact, one "chaologist” describes chaos as . . lawless behavior governed entirely by law."2 To demonstrate the difficulty in determining whether a variable is random or chaotic, figures la and lb show two time series of a variable; one series is a random variable, whose actual value cannot be known with certainty, and the 'The terms “ deterministic chaos” and “ chaos” are used interchangeably here, although deterministic chaos is the more precise description. FEDERAL RESERVE BANK OF ST. LOUIS other is a chaotic variable, whose value can be predicted with certainty. Even the most prac ticed observer, however, would have difficulty determining which of these series, if any, is not random. As a result, most economists would model or estimate both time series as random processes. The chaotic series is described by a very simple deterministic equation and identified later in this paper. Often, behavior that cannot be explained by standard theories and modeling techniques is at tributed to random forces, even when there is no theoretical reason to do so. This paper argues that economists are perhaps not using the ap propriate types of models and empirical techni ques to explain the behavior of some economic variables and that the choice of methodology needs to be more closely examined. The study of chaos is a recent phenomenon in the biological and physical sciences and is just 2Stewart (1989), page 17. 37 Figure 1 Random or Chaotic? b MARCH/APRIL 1990 38 now beginning to be applied to econom ics. Un fortunately, many of the em pirical tests for chaos are im precise and, because of m athem ati cal constraints, the theoretical models used to generate chaos are generally limited to systems w ith only one or tw o explanatory variables. Both of these factors restrict the usefulness of applying chaos to econom ic systems. N everthe less, the theory of determ inistic chaos has at tracted a great deal of attention, both in the popular press and in academ ic circles. The dis cussion that follows attem pts to clarify some of the issues and suggests some ways to incorporate chaos into econom ics. This article first review s how econom ic vari ables typically are modeled by describing and evaluating several techniques of econom ic m odel ing using a simple model of output and popula tion grow th .3 Next, chaos is defined and its pro perties dem onstrated. The advantages and pit falls of applying the theories o f chaos to eco nomics are then discussed and illustrated. ECONOM IC M ODELING T h ere are m any d ifferent ways to build econom ic models. Four such possibilities are ex amined h ere fo r the case in w hich all variables are com pletely determ inistic .4 The types of mod els exam ined h ere are static linear, static non linear, dynamic linear and dynamic nonlinear. A fu rth er distinction, w hich proves to be signifi cant, is also draw n betw een discrete and contin uous tim e dynamic models. A simple model of output, w h ere labor is the only input, is used to illustrate each approach to modeling as well as the restrictiveness of many com m on modeling techniques. In addition, focusing on econom ic modeling allows us to show that chaotic dynam ics can only arise in certain types of models that have often been excluded, a priori, by econom ists. Static M odels T he simplest type of econom ic model is a static linear model, in w hich variables do not 3There is a growing body of theoretical literature incor porating chaos into many different types of economic models. These models include Benhabib and Day (1981), Deneckere and Pelikan (1986), Grandmont (1985), De Grauwe and Vansanten (1990), Kelsey (1988), Day and Shafer (1985) and Stutzer (1980). For surveys of the theoretical literature, see Kelsey (1988) and Baumol and Benhabib (1989). 4There is also a burgeoning field in stochastic (random) modeling, which incorporates the assumption of random FEDERAL RESERVE BANK OF ST. LOUIS change over tim e and are related in a propor tionate m anner. Consider, fo r example, the following simple production function, w hich has only labor as an input: (1) Y = AN A > 0, w here Y is output, w hich is com pletely consum ed by w orkers (there is no saving or investment), N is labor employed and A is the productivity param eter. This equation states that output is positively related to the amount of labor em ployed. Given the value of A and the labor sup ply, the exact value of output can be determined. This type of model is highly restrictive; any change in labor changes output by a constant percentage. Hence, the production function ex hibits constant retu rn s to scale. Allowing the model to be nonlinear (that is, not necessarily proportionate) provides a m ore general model in w hich equation 1 is a special case. An example of a nonlinear production function is given by: (2) Y = ANa A > 0, a > 0. If a = 1, this model is identical to the one shown in equation 1. By not restricting a to equal one, how ever, this model can be used to examine the case in w hich output can vary disproportionately w ith respect to changes in labor. This is illustrated in figure 2, w hich shows the relationship betw een output and labor for different values of a (for simplicity, A = 1). Notice that if a is betw een zero and one, the production function exhibits decreasing retu rns to scale (that is, output increases less than p ro portionately w ith resp ect to a change in labor); if a is greater than one, production is ch aracter ized by increasing retu rn s to scale (output in creases m ore than proportionately with respect to an increase in labor). Em pirical tests of actual production relationships can be perform ed to determ ine if a is actually different from or equal to one. ness used in econometric models into the theoretical litera ture. Recent papers also look at the properties of chaos in the presence of a random component [see, for example, Kelsey (1988)]. For simplicity, this paper focuses only on purely deterministic systems. 39 Figure 2 Linear vs. Nonlinear Production Functions Y = ANa w h ere C and D are constants, and a dot over a variable m eans the change in the variable with respect to a very small change in time. This type of equation is called a differential equation. Equation 3 states that the percentage rate of change of the labor fo rce [N(t)/N(t)l, w h ere time is continuously changing, equals the difference b etw een the rate of birth, C, and the rate of death, given by DN(t)/Y(t), w h ere N(t)/Y(t) is the num ber of individuals who have to subsist on each good at tim e t. Using the linear production function given in equation 1 and substituting it into equation 3 provides a linear specification of the percentage change in the population: (4) N(t)/N(t) = C -D /A . Notice that w hen the production function is linear the rate of death, D/A, is constant. Solving equation 4 yields the following solu tion fo r the population: (5) N(t) = Ke(c‘D )t< /A N Linear Dynam ic M odels One disadvantage of these static models is that they can be used to describe the relationship b etw een output and em ploym ent only if the labor force or population rem ains constant over tim e .5 Suppose instead that we w ant to examine the behavior of output over time as it is related to a continuously changing labor force. A stan dard equation borrow ed from Haavelmo (1954), used to describe the grow th of the labor force w here that grow th is dependent on the level of output, is given by: (3) N(t)/N(t) = C — DN(t)/Y(t) C > 0, D > 0, 5For expositional ease, the terms “ population” and “ labor force” are used interchangeably. 6Equation 4 is solved by the variable separable method of solving differential equations found in most calculus books. K is the constant of integration, which can be determined by choosing an initial condition. 7ln fact, stability is an important issue which is frequently ignored or abstracted from in economics. Stability is impor tant because, for example, an unstable equilibrium is not a sustainable equilibrium. Stability is also important in the choice between linear and nonlinear models. Linear models w h ere K is an arb itrary constant .6 This solution has the property that, unless the rate of birth (C) is exactly equal to the rate of death (D/A)—in w hich case the population will equal K—the population will either rise exponen tially or fall to zero. This result is highly restric tive, how ever, because the likelihood of either the two rates being identical or the population increasing infinitely is, in reality, very small. In other words if C # D/A, the system is u nstable .7 Unfortunately, in models of oth er types o f eco nomic variables, results that greatly restrict the possible values of the param eters of the models are not uncommon. In addition, because of the complexity of many econom ic models, the im plications of restricting the value of the param e ters to determ ine the solution or to ensu re a have three possible cases: stable converging dynamics (such as when C = D/A in the model above), unstable dynamics (when C * D/A) and cyclical dynamics, which is the least common of the three. In nonlinear models, how ever, cyclical dynamics are far more common, and ex ploding dynamics may not occur. Thus, it is also important to consider the desirable and realistic stability properties when choosing a model. Obviously the nonlinear case is more general and the most realistic for variables that ex hibit cyclical variation. For the purpose of this paper, however, the issue of stability is ignored. MARCH/APRIL 1990 40 stable solution are not always as obvious as in the population grow th model. Because linear dif ferential equations are fa r sim pler to solve than nonlinear differential equations, and because th eir solutions are m ore often stable and easier to interpret, how ever, they are used in econom ic models m ore often than may be appropriate. dynamic models are called d iffe r e n c e equations; they m easure time in distinct intervals rath er than the d ifferen tia l equations used above, w hich m easure tim e continuously. Equation 6 can be transform ed into a difference equation by let ting the rate of change of N (previously given by N) equal the difference betw een the value of N at tim e t and t + 1. Thus, equation 6 becom es Nonlinear Dynam ic M odels (7) (Nt+ 1 - N t)/N, = C - D(Nt/Yt), Combining the nonlinear production function given by equation 2 w ith the description of population grow th given in equation 3 provides a less restrictive model of population growth: w h ere Yt = AN,". Combining these equations and simplifying the result yields: (8 ) Nt+I = N, [(1 + C) - DN,‘ -“/A] , (6 ) N(t)/N(t) = C - DN(t)‘ "7A . Unlike equation 4, equation 6 allows the labor fo rce to vary m ore or less as the cu rren t labor fo rce changes. Unfortunately, the price of the generality provided by such nonlinear d ifferen tial equations is that most either cannot be solved or have solutions so com plex the results cannot be interpreted. Not surprisingly, econom ists often avoid these types of models. The model used here, however, was chosen for its tractability and can be solved for the value of labor at any time t .8 All that is necessary for a stable solution is that the production function ex hibits decreasing returns to scale (0 < a < 1 ). Regardless of the value of the oth er param eters, if a is betw een zero and one, the population will reach a stable equilibrium level. Hence, in contrast to the dynamic linear model discussed previously, the results of this model are m ore realistic and provide a m ore general description of population and output growth. Discrete M odels One problem w ith using continuous time models in econom ics is that data are available only in distinct intervals (daily, weekly, monthly, etc.). One approach typically taken by econom ists, th erefore, is to convert these continuous tim e models into discrete tim e models. D iscrete 8For the solution and discussion of this model, see Haavelmo (1954), pp. 24-29. 9 Allowing N, = [A(1 +C)/D],/(I“,) X, only changes the scale of the population and has no effect on the general charac teristics of the solution. For further discussion of this pro cedure and the solution, see Stutzer (1980). 10Although the model given by equation 6 is stable in con tinuous time, it is not necessarily stable in discrete time RESERVE BANK OF ST. LOUIS FEDERAL w hich, following Stutzer (1980), can b e rew ritten using a change of variables as (9) Xt+1 = k Xt( l - X / " ) , w h ere k = 1 + C .9 The models shown in equations 8 and 9 de scribe the m ost general specification of popula tion and output grow th given the assumptions made above. Behavior is not restricted to being linear, n or is population or output restricted to rem aining constant over time. On the other hand, as noted earlier, these m ore general models often cannot be solved or have solutions w ithout any econom ic interpretation. Neverthe less, unless th ere are theoretical reasons for assuming relationships are static or linear, dynamic nonlinear models, w hich provide the most general specification of behavior, should at least be considered in econom ic analysis. Al though generality fo r its own sake is not a de sirable goal, using a m ore general model would be appropriate w hen sim pler models have solu tions that are highly unrealistic or w hen param eters have to be restricted beyond reason (as in the model presented here). In addition, if eco nomists w ant to test th eir models and results empirically, then these variables should be modeled in the form in w hich they are esti m a te d -d iscre te form .10 As it turn s out, these types of nonlinear dynamic models can exhibit chaos. since in discrete time this model can generate chaotic dynamics for certain parameter values. Differential equa tions can also exhibit chaos, although only in more com plicated models. This is discussed in greater detail later. 41 A N IN T R O D U C T IO N T O THE T H E O R Y OF CHAOS The possibility that chaos exists in econom ic variables has strong implications fo r the w ay in which econom ics is modeled. For example, some variables that appear to be random processes, like one of the variables shown in figure 1 , might in fact not be random at all; instead it might be com pletely explained using the ap propriate determ inistic model. This section dem onstrates the properties of chaos, using a simple model. Figure 3 The Logistic Growth Curve For Various Values of k Xt+1 = kX t ( 1 - X t ) In the m ost general sense, the term chaos is used to describe the behavior of a variable over tim e that appears random but, in fact, is d eter ministic; m ore precisely, given the initial value of the variable, all future values of the variable can be calculated w ith exact precision .11 In con trast, the value of a random variable can never be predicted w ith certainty. M ore formally, a function is chaotic if, for certain param eter values, the following two con ditions hold: First, the function never reaches the same point tw ice under any defined interval o f time. In this case, the function is said to ex hibit ap erio d ic behavior. Second, the time path is sensitive to changes in the initial condition, so that a small change in the value of the initial condition will greatly alter the tim e path of the function .12 Chaos only arises in certain types of nonlinear dynamic systems, although not all nonlinear dy namic equations are chaotic. M oreover, equa tions that can be characterized as chaotic need not exhibit chaos for all param eter values. Rath er, functions that can exhibit chaos will do so only for certain param eter values. This is ex plained by example below. And N o w f o r Something Completely Different . . . The properties of chaos can be dem onstrated using a simple m athem atical equation, called the logistic grow th equation. W hile this model has no particular econom ic interpretation, it is the 1 For simplicity, only single-variable equations are discussed. 1 Although chaos exists in multivariate economic systems, tests for chaos in these systems are just beginning to be developed, and the mathematics of such systems are ex tremely complex. 12There are many different characterizations of deterministic chaos, but they all include the one used here. For more rigorous definitions and discussion of the different defini- x xt simplest model that exhibits chaos and provides reasonably interpretable graphical results. This equation is given by: (10) X t+1 = k Xt( l - X t), 0 < X, < 1, 0 < k < 4. Equation 10 describes the tim e path of a vari able, X (which fo r expositional purposes is called a population), that is a function of its previous value and a param eter k. To dem onstrate chaotic behavior in this simple fram ew ork, the value of X can only take on values betw een zero and one. The value of k, the only param eter in the equa tion, is called the “tuning” param eter; it deter m ines the steepness o f the function. Figure 3 shows the function given in equation 1 0 for various values of k. Increases in the popula tion below X increase futu re values of X more than proportionately. Past this point, the popula tion begins to d ecrease .13 For larger values of k, tions, see Li and Yorke (1975), Brock and Dechert (1988) and Melese and Transue (1986). For a good mathematical description of chaos and the mathematical tools used in the theory of chaos, see Devaney (1989). 13This behavior is similar to that of a total product curve where, once the marginal product becomes negative, fur ther increases in an input decreases output. MARCH/APRIL 1989 42 Figure 4 A Stable Tim e Path for a Logistic Growth Curve Xt+1 = 3 X t ( 1 - X t ) X0 = .20 t = 1 to 500 xt+ i xt the absolute value of the rate of change of X is larger. For certain values of the tuning param eter (k < 3), the system is stable; this m eans the population will reach some sustainable steadystate value w hich differs from X. Figure 4 illustrates how the tim e path fo r X t+1 is solved graphically. The parabola represents equation 1 0 w hen k is equal to three; all values of X, and Xt+1 m ust lie on this curve. The 45degree line depicts the points w here X t+1 = X„ 14Notice that X1+l, which must always lie on the parabola, can be either above the 45-degree line (as in point D) or below it (as in point F). For precision, the equation is solved numerically and then graphed. RESERVE BANK OF ST. LOUIS FEDERAL w hich is required fo r a steady-state equilibrium. In this example the initial value (when t = 0) is .20. To determ ine the value of X,, draw a line betw een the initial value (X0) and the parabola (line segm ent X 0B). To find the value of X2, set X, = Xj by drawing a line from point B to the 45-degree line (point C). T h en draw a straight line from point C to the parabola. This is the value of X 2 (point D). This process, called itera tion, can be used to determ ine as many subse quent values of X as is desired, once the initial value is determ ined .14 As we can see in figure 4, 43 the population appears to be converging to a steady-state equilibrium value at 2/3 (X*). If the value of k increases past three, how ever, the equilibrium point becom es unstable and the time path exhibits a two-period cycle, w here the variable alternates betw een two values. Fu rth er increases in k produce a fourperiod cycle (that is, the tim e path repeats the same sequence of num bers every fifth iteration), then an eight-period cycle, and so on, w ith the periodicity increasing by 2° (n = 1,2,3, . . .). If k increases past a certain point called the “point o f accum ulation” (for this function, it occurs at k = 3.5700), the tim e path en ters into a region in w hich the function can exhibit chaos .15 In the chaotic region (3.57 < k < 4 for this function), th ere can be both an infinite num ber of periodic cycles and an infinite num ber of initial condi tions that produce an aperiodic tim e path .16 Us ing this simple example, we can dem onstrate some of the properties of chaos in graphical form . Figure 5 A Logistic Growth Curve Exhibiting Chaos Xt+1 = 3.82840 Xt ( 1 - Xt ) X0 = . 0 1 0 1 t = 1 to 500 xt.i Properties o f Chaos An example of aperiodic behavior is seen in figure 5. The first 500 iterations are shown in this figure (that is, t = 1, 2, . . . 500), and no single poin t is ev er re a c h ed tw ice . 17 In fact, no m atter how m any tim es this equation is iterated, X, never has the same value tw ice .18 If the data are plotted as a tim e series, it would look similar to figure la , the chaotic series in figure 1 , and one might conclude that the data are generated by a random process, such as figure lb , because they follow no obvious pattern. This is not the case here; the data in figure l a and figure 5 w ere generated from models w ithout a random com ponent and th erefo re are completely determ inistic. The oth er ch aracteristic of a chaotic function is that its tim e path is sensitive to the choice of initial values. An example o f how changing the 15This process of increasingly complex periodicity is called bifurcation and is discussed in most papers on chaos. For a nontechnical discussion of bifurcation, see Gleick (1987) and Stewart (1989). For a more analytical treatment of bifurcation, see May (1976) and Baumol and Benhabib (1989). A more rigorous discussion of the relationship be tween bifurcation and chaos is given in Li and Yorke (1975). 16Notice that not every initial condition gives rise to an aperiodic time path. 0.0 0.5 1.0 X, initial condition can affect the time path is shown in figure 6 . In this figure, the values of X, are plotted against time, as in figure 1. This diagram dem onstrates how changing the initial value, X0, at the fou rth decimal place (from .0 1 0 1 to .0 1 0 0 ) causes the time paths generated by equation 1 0 to deviate substantially from each o th er .19 Although not all sections of the time path differ as dram atically as the one shown here, figure 5 graphically dem onstrates that the choice o f an initial condition or, for forecasting purposes, the choice of a time in ter val (that is, determ ining w h ere to start the sam ple), can greatly alter the results. In fact, despite icity is required for chaos, tests for chaos in actual data take a different approach, thus avoiding the problem. 1 8The time path can avoid having repeat values because the number of possible points between zero and one is infinite. 19Although it sometimes looks like the function is periodic, this appearance is a result of the lack of precision of the printer and the scale of the graph. In fact, there are no periodic points in this function. 1 This property is unlikely to be found in actual data, how 7 ever, because of rounding. Although theoretically, aperiod- MARCH/APRIL 1989 44 Figure 6 A Segm ent of the Time Trend Showing Sensitivity to Initial Conditions Xt+ = 3.82840 Xt ( 1 - X t ) 1 Xo=.0101 Xo=.0100 seemingly trivial differences in the initial condi tions, the time path produced by one initial value will not necessarily be similar to the time path generated by a marginally different initial value. In general, the two time paths that arise from the different initial values will have periods during which they are arbitrarily close together and periods during which they deviate substantially. Chaotic functions also exhibit sensitivity to very small changes in the param eter values. A third- or fourth-order change in the value of a param eter can alter the tim e path from stable to chaotic or vice versa. 20Recall that when a function is in a chaotic region (that is, when the parameters are such that the function can ex hibit chaos), there can be both periodic and aperiodic time paths. FEDERAL RESERVE BANK OF ST. LOUIS Sensitivity to changes in the param eter values is illustrated in figure 7. Here, a fifth-order change in the value of the tuning param eter (from 3.82840 to 3.82844) produces not only a substantially different tim e path from the one in figure 5, b u t also one that exhibits periodic rath er than chaotic behavior .20 A re Attractors Strange? Another feature often found in chaos, although neith er necessary or sufficient for chaos, is a strange a ttra cto r .21 The properties of attractors and strange attractors are best illus21The only examples of chaos without the presence of a strange attractor are found in certain types of dissipative systems. 45 Figure 7 A Logistic Growth Curve With Periodic Points Xt+1 = 3.82844 Xt ( 1 - X t ) X0 = .0101 t = 1 to 500 Figure 8 The Strange Attractor for a Chaotic Function Xt+1 = 3.82840 Xt ( 1 - Xt ) X0 = .0101 t = 1 to 1500 x,., 0.0 trated by example. In a stable system, the time path converges to an equilibrium point (for ex ample, X* in figure 4). T he equilibrium point is also called the attractor, because the tim e path is "attracted ” to the equilibrium point. A nother possibility is that the tim e path has two attract ors, and the system oscillates betw een them, never rem aining at one equilibrium point. This is found in predator/prey population models, w h ere the population grow s until it is so large it begins to die o ff and then shrinks to a level so small it begins to grow again. A "strange a ttracto r” is the nam e given to the case w here th ere is a region, rath er than a finite set of points, that attracts the tim e path o f the variable. That is, after some num ber of iterations, w hich varies depending on the fu n c tion, the time path of the variable is completely contained in this region (the strange attractor). Thus, even though the path is aperiodic and 22Another definition of a strange attractor is an attractor with fractal dimension. In fact, if a strange attractor exists, the variable has fractal dimension. Random variables have in finite dimension, however. As a result, tests for dimension are one of the main ways data are tested to determine if 0.5 1.0 th erefo re never reach es an equilibrium in the standard sense, it also never leaves the strange attractor and th erefo re is not unstable (for ex ample, never goes to positive or negative infini ty). An example of this is shown in figure 8 , w hich takes the same num erical example as in figure 5, but iterates it 1500 rath er than 500 times. In this picture, the values of X are still contained in the same area as in figure 5, but the distribution of points is becom ing denser. T he bounded region (shown by the dotted line in figure 8 ) is the strange attractor fo r this function. If the function is iterated fu rth er, the area w ithin the bounded region would appear to be a solid block, although the function would never have the same value tw ice. In fact, the existence of a strange attractor is an im portant way to distinguish betw een a random and chaotic tim e path .22 they are chaotic. For the purpose of this paper, however, the issue of fractals and fractal dimension will be ignored For a discussion of these topics, see Mandelbrot (1983) and Gleick (1987). MARCH/APRIL 1989 46 LESSONS FROM CHAOS Although econom ists are beginning to in cor porate chaos into their econom ic and econo m etric models, th ere has been little discussion of the ways in w hich chaotic dynamics are use ful or realistic fo r econom ic models. Clearly, chaos holds considerable appeal fo r econom ists who are looking fo r a determ inistic explanation of the apparent random ness in econom ic vari ables. Econom ists frequently assum e random ness w hen they are unable to explain the behavior of an econom ic variable empirically. T he presence of an alternative explanation, chaos, will require them to consider m ore carefully the rationale behind their assumptions. One problem with incorporating chaos into econom ics is that, while econom ists can either postulate an equation and test it for the presence of chaos or, alternatively, see if the data them selves are chaotic, it is especially difficult to identify the co rre ct functional form that ch arac terizes the data. T he choice of a functional form is always a problem in econom ics, but, as p re viously discussed, it is particularly difficult to model nonlinear dynamics. This problem is ex acerbated because, as a result of the m athem at ics required, the study o f nonlinear dynamics has, until recently, been relatively limited in general and largely ignored in econom ics .23 Even w hen it is possible to estim ate nonlinear dynamic equations, the models them selves often cannot be solved analytically. W ithout explicit solutions to these models, their usefulness is ex trem ely limited. Obviously, the difficulty of de term ining the "tru e” underlying model from a data series is a problem w hether or not chaos exists. The "discovery” of chaos, how ever, has focused m uch m ore attention on this problem, especially if the data are nonlinear. E conom ic M odeling and Chaos The study of chaos em phasizes the im por tance of rigorously modeling the dynamics of a system ra th er than m erely taking a static model (like equations 1 and 2 ) and adding time sub 23For recent work in nonlinear dynamics, see Grandmont (1987). 24The “ order” of an equation refers, for a differential equa tion, to the highest power attained by the derivative and, for a difference equation, the highest degree of differen cing. For a more complete discussion, see Chiang (1984). FEDERAL RESERVE BANK OF ST. LOUIS scripts and an erro r term . Although these sim pler models may be m ore likely to have solutions with explicit results that can be tested em pirical ly, the dynamics that arise may not capture the behavior of the variable of interest. T he richness of a model may be found in explaining the b e havior of a variable over time as m uch as in the direct, tim e-independent (or time-constant) rela tionship betw een the variables. In addition, the study of determ inistic chaos il lustrates some of the pitfalls of first d ifferen cing a dynamic model to convert it to discrete time, as was done in the model of population grow th presented above. This practice is com mon in econom ics because data are only available in discrete intervals. As is shown in Stutzer (1980) and dem on strated here, th ere are first-order differential equation models (such as equation 6 ) w hich con verge to a steady-state equilibrium th at are cha otic w hen expressed in discrete tim e (equation 9). Thus, the dynamic properties of the discrete analog o f a differential equation cannot be as sumed to be the same. In fact, it has been shown that, although chaos can arise in firstorder d iffe r e n c e equations, it can only arise in third-order or higher d ifferen tia l equations .24 As a result, an econom ist m ust be carefu l about either converting a continuous tim e dynamic model into a discrete model (such as converting equation 6 into equation 7), or taking a static model and simply adding a tim e subscript, rath er than postulating a model that is dynamic (in eith er discrete or continuous time) and esti mating or solving it in that form . The choice of the appropriate type of model should depend on the econom ic variables being described rath er than analytical convenience. This issue is partic ularly im portant if a continuous-tim e dynamic model is estim ated in discrete time using the steady-state equilibrium properties of the continuous-tim e solution. The discrete-tim e equa tion that is being estim ated m ay not reach a steady state at all, or the solution could differ qualitatively from th at found in the continuous time version of the model. 47 Empirical Applications of Chaos in Economics T h ere are generally tw o approaches used in the em pirical literature to test for the pres en ce of determ inistic chaos in econom ic and financial data. The first approach tests for the presence of nonlinearities in the data .1 Since chaos only arises in nonlinear systems, finding nonlinearities in the data suggests that testing directly fo r the presence of chaos is appropriate. In addition, the presence of nonlinearities in the data provides inform a tion to theorists modeling these types of econom ic systems. Because testing fo r nonlinearities in the data is m uch sim pler (and less controversial) than testing fo r chaos, these tests are often perform ed first. Many m acroeconom ic tim e series have been found to behave in a nonlinear m anner. Brock and Sayers (1988) find such evidence in data for quarterly em ploym ent (1950-83), quarterly unemploym ent (1949-82), monthly post-war industrial production and pig-iron production (1877-1937). Nonlinearities have also been found in the Divisia M l m onetary aggregates .2 O ther studies have found nonlinearities in financial data as well. For example, Hinich and Patterson (1985a, 1985b) find strong evi dence o f nonlinearity in daily stock returns. The second approach is to test directly for the presence of chaos .3 T h ere are many pro blems with testing directly fo r chaos using econom ic data, how ever. The most obvious, and perhaps the m ost im portant, is the sen sitivity of chaotic systems to small changes in the param eter values and initial conditions. For these tests to be accurate, the data need to be especially exact. This degree of preci sion presents a particular problem fo r eco nom ics, w h ere controlled experim ents are essentially impossible, especially on the m acro level. Data collection is far from perfect, and the quality of the data declines as the degree of aggregation increases, introducing m easure m ent e rro r in the data. In addition, because o f rounding, the data are not as precise as they should be. For this reason, tests fo r chaos are not simply tests for aperiodicity. The quantity o f high-quality data is also ex trem ely im portant. Even if the results show aperiodicity fo r a sample of 1 0 0 observations, the system need not b e aperiodic. The existing em pirical tests fo r the presence of chaos re quire an extrem ely large num ber of highly accurate data. Rarely are both of these avail able to econom etricians. As a result, any evi dence from tests for chaos should be viewed w ith caution. Given these caveats, some statistical tests, originating in the physical sciences, do look fo r the presence o f chaos in econom ic data. These tests are run on variables that have long tim e series available, are not aggregate variables, and are thus m ore likely to provide accurate results. T ests have found evidence consistent w ith chaos in exchange rates (Ellis, 1990), daily gold and silver prices on the Lon don m arket (Frank and Stengos, 1988) and in the Divisia m onetary aggregates (Barnett and Chen, 1988).4 'For a discussion of the tests used, see Brock and Sayers (1988), as well as the other papers cited above. drawbacks, see Barnett and Hinich (forthcoming), Brock (1986) and Ramsey (1989). 2For a definition and discussion of the Divisia monetary aggregates, see Barnett and Spindt (1982). 4This is by no means a comprehensive survey of the empirical literature applying chaos to economic and financial data. For a more comprehensive discussion of the empirical work on chaos, see Barnett and Hinich (forthcoming) and Ramsey (1989). 3A description of the actual empirical techniques used to test for chaos is beyond the scope of this paper. For a description of the tests available for chaos and their Econom etrics and Chaos The study of determ inistic chaos also offers several lessons for econom etricians. If forecast ing is a goal of econom ic modeling, inappropriate modeling techniques in the presence of chaos becom e m ore costly. If the data are chaotic, forecasting is close to impossible since a small erro r in the value o f the initial condition can lead to highly inaccurate predictions (see, for example, figure 6 ). Similarly, an erro r in any param eter value can also produce in correct fore casts (see figures 5 and 7). Thus, it is im portant MARCH/APRIL 1989 48 to realize the limitations of econom ic forecasts in the presence of chaotic variables. Chaos does not have to be present in the data to find the sort of fluctuating behavior (although w ithout any clearly defined periodicity) that is often found in econom ic data. Nonlinear nonchaotic models often can generate time paths that appear random , and testing for nonlineari ties is the likely next step fo r future research in this area. In fact, em pirical econom ists are b e ginning to test fo r both nonlinearities and chaos in econom ic data (see insert on page 47). As a result, m ore w ork needs to be done in und er standing nonlinear estim ation so that econom ic models can describe a g reater variety of b e havior and be m ore accurate as well. In addi tion, the existence of chaos suggests that econo mists might w ant to try nonlinear specifications of a variable b efore resorting to modeling it as a random variable. This in tu rn will help to im prove the quality o f econom ic forecasts in the presence of nonlinear variables. In addition, the use of a random com ponent in estim ation does not necessarily imply that the variable itself is random, but rath er that other relevant variables might be excluded from the regression. Although each of these oth er vari ables could have a small influence on the system by itself, the total effect of these excluded vari ables could be substantial. Given both the diffi culty in detecting w hat these missing variables might be and data limitations, such a complex system might best be approxim ated by a random variable, even if th ere is no tru e random ness in the variable being estimated. In fact, some argue (see, for example, Kelsey, 1988) that, since eco nom ic models do not include such (chaotic) phe nom ena as w eather and other biological factors w hich can influence econom ic variables, it "seem s inevitable that we will have random term s in our equations .”25 CONCLUSION The study of determ inistic chaos and its subse quent application to econom ics has opened a new realm of possibilities for econom ists trying to explain cyclical or erratic behavior in eco nomic variables. As discussed above, chaos has implications for both theoretical modeling and em pirical applications in econom ics. By illustrat 25Kelsey (1988), p. 12. FEDERAL RESERVE BANK OF ST. LOUIS ing explicitly how restrictive the assumption of linearity can be, the study of chaos emphasizes the im portance of allowing fo r the possibility of nonlinear behavior. T he use of chaos in econom ics also has offered new explanations fo r behav ior that, until recently, has been able to be ex plained only by random forces. The techniques that have arisen from the study of chaos in the physical and biological sciences are in th eir infancy. As these tech ni ques becom e m ore refined, and econom ists b e com e b etter trained in working with these types of models, their ability to explain the behavior of variables such as exchange rates, business cycles and stock prices is likely to improve. That possibility alone is sufficient reason for econom ists to take a closer look at determ inistic chaos in particular and nonlinear dynamics in general. REFERENCES Barnett, William A., and Ping Chen. “ The AggregationTheoretic Monetary Aggregates are Chaotic and have Strange Attractors: An Econometric Application of Mathematical Chaos,” in William A. Barnett, Ernst R. Berndt, and Halbert White, eds., Dynamic Econometric Modeling, Proceedings of the Third International Sym posium in Economic Theory and Econometrics (Cambridge University Press, 1988), pp. 199-245. Barnett, William A., and Melvin J. Hinich. "Has Chaos Been Discovered with Economic Data?” in Ping Chen and Richard Day, eds., Evolutionary Dynamics and Nonlinear Economics (Oxford University Press, forthcoming). Barnett, William A., and Paul A. Spindt. “ Divisia Monetary Aggregates: Compilation, Data, and Historical Behavior,” Board of Governors of the Federal Reserve System, Staff Studies 116 (May 1982). Baumol, William J., and Jess Benhabib. “ Chaos: Significance, Mechanism, and Economic Applications,” Journal of Economic Perspectives (Winter 1989), pp.77-105. Benhabib, Jess, and Richard H. Day. “ Rational Choice and Erratic Behaviour,” Review of Economic Studies, (July 1981), pp. 459-71. Brock, W. A. “ Distinguishing Random and Deterministic Systems: Abridged Version,” Journal of Economic Theory, (October 1986), pp. 186-95. Brock, W. A., and W. D. Dechert. “ Theorems on Distinguish ing Deterministic from Random Systems,” in William A. Barnett, Ernst R. Berndt, and Halbert White, eds., Dynamic Econometric Modeling, Proceedings of the Third Interna tional Symposium in Economic Theory and Econometrics (Cambridge University Press, 1988), pp. 247-65. Brock, W. A., and Chera L. Sayers. “ Is the Business Cycle Characterized by Deterministic Chaos?” Journal of Monetary Economics, (July 1988), pp. 71-90. Chiang, Alpha C. Fundamental Methods of Mathematical Economics, 3rd ed. (McGraw-Hill Book Company, 1984). Day, R. H., and Wayne Shafer. “ Keynesian Chaos,” Journal of Macroeconomics (Summer 1985), pp. 277-95. De Grauwe, Paul, and Kris Vansanten. “ Deterministic Chaos in the Foreign Exchange Market,” Discussion Paper Series No. 370, Centre for Economic Policy Research, 1990. Deneckere, Raymond, and Steve Pelikan. "Competitive Chaos,” Journal of Economic Theory (October 1986), pp. 13-25. Devaney, Robert L. An Introduction to Chaotic Dynamical Systems (Addison-Wesley Publishing Company, Inc., 1989). Ellis, Jennifer M. “Are Exchange Rates Chaotic?” mimeo, University of Oregon, 1990. Frank, Murray, and Thanasis Stengos. “ Chaotic Dynamics in Economic Time-Series,” Journal of Economic Surveys (Vol. 2, 1988), pp. 103-33. Gleick, James. Chaos: Making a New Science (Penguin Books, 1987). Grandmont, Jean-Michel, ed. Nonlinear Economic Dynamics (Academic Press, 1987). Grandmont, Jean-Michel. “ On Endogenous Competitive Business Cycles,” Econometrica (September 1985), pp. 995-1045. Haavelmo, T. A Study in the Theory of Economic Evolution (Amsterdam: North-Holland Publishing Company, 1954). Hinich, Melvin J., and Douglas M. Patterson. “ Identification of the Coefficients in a Non-Linear Time Series of the Quadratic Type,” Journal of Econometrics (November 1985a), pp. 269-88. ________ "Evidence of Nonlinearity in Daily Stock Returns,” Journal of Business and Economic Statistics (January 1985b), pp. 69-77. Kelsey, David. “ The Economics of Chaos or the Chaos of Economics,” Oxford Economic Papers (March 1988), pp. 1-31. Li, Tien-Yien, and James A. Yorke. “ Period Three Implies Chaos,” American Mathematical Monthly (December 1975), pp. 985-92. Mandelbrot, Benoit B. The Fractal Geometry of Nature (W. H. Freeman and Company, 1983). May, Robert M. “ Simple Mathematical Models With Very Complicated Dynamics," Nature (June 10, 1976), pp. 459-67. Melese, Francois, and William Transue. “ Unscrambling Chaos Through Thick and Thin,” Quarterly Journal of Economics (May 1986), pp. 419-23. Poincare, Henri. Science and Method, translated by Francis Maitland (Dover Publications, Inc., 1952), pp. 67-68. Ramsey, James B. "Economic and Financial Data as Non linear Processes,” in Gerald Dwyer and R. W. Hafer, eds., The Stock Market: Bubbles, Volatility, and Chaos (Kluwer Academic Publishers, 1989). Stewart, Ian. Does God Play Dice?: The Mathematics of Chaos (Basil Blackwell, Inc., 1989). Stutzer, Michael J. “ Chaotic Dynamics and Bifurcation in a Macro Model,” Journal of Economic Dynamics and Control (November 1980), pp. 353-76. 49 David Laidler David Laidler is a professor of economics at the University of Western Ontario and is affiliated with the C.D. Howe Institute. This paper, the fourth annual Homer Jones Memorial Lecture, was delivered at the Federal Reserve Bank of St. Louis on April 11, 1990. The L egacy Of The M on etarist C o n tro v ersy IN T R O D U C T IO N , It is not quite true, as one (hostile) com m entary , , „ . , , , , has asserted, that M onetarism was developed by » i. • t, • . i t-« j it, t> . Milton Friedm an at the Federal Reserve Board (sic) o f St. Louis,” but it is nevertheless the case that the intellectual environm ent created at this Bank by Homer Jones ensured that the doctrine took root and flourished h ere w hen it was very m uch a m inority taste elsew h ere .1 And indeed, at least two early and seminal contributions to the M onetarist controversy, Andersen and Jordan (1968) and of course Bru nner (1968), w hich gave the controversy its label, first appeared in the Bank’s R eview . The M onetarist controversy, therefore, is surely a suitable topic for this lecture. Now M onetarism has been m uch defined and debated over the years, to the point at w hich one may find authority fo r applying the term to almost any econom ic and/or political doctrine one likes, or m ore probably dislikes. However, it is not so m uch my purpose h ere to define that doctrine in detail y et again, as it is to discuss the consequences fo r the developm ent of m onetary econom ics, both in theory and practice, of the debates to w hich it gave ° rise during the 1960s and 1970s. ° In this lecture, I shall first of all describe the issues that w ere at stake at the outset of those debates. I shall show that although the Monetarist policy agenda was very different from . x £ . . , x that o f what we might call Keynesian ortho, , . . .... . doxy, the positive differences in econom ic analyJ S1S whlch underlay the P °hcy debate w ere at first em pirical in nature, raising no fundam ental questions of econom ic theory. I shall also show, how ever, that, w h eth er it was logically necessary or no*' theoretical considerations o f profound im portance did get introduced into the Monetarist controversy as it progressed. Although these at first seemed to strengthen the Monetarist position, I shall go on to argue that these very considerations in the longer run undermined it, leaving M onetarism w ithout distinct theoretical foundations, and incapable of coping with the em pirical difficulties w hich it began to encounter from the m id-1970s onward. Furtherm ore, I shall suggest that most, though not all, Qf those em pirical problem s stemmed from an attempt by "Keynesian” orthodoxy to adapt Monetarist ideas to its own use. Finally, I shall argue that though the M onetarist controversy was to all intents and purposes over by the early 1980s, ,, , . , .. , , . the problem s w hich it bequeathed to m onetary . .. , . „ . .. , econom ics continued to affect theoretical, empirical- and policy-oriented aspects of the sub-discipline into the 1980s, with results which 'See Gould, Mills and Stewart (1981), p. 26. MARCH/APRIL 1990 50 I cannot help but view w ith considerable discom fort. THE M O N E TAR IST C O N TR O VERSY IN THE 1960s The M onetarist controversy concerned the role played by m onetary variables in general, and the quantity of money in particular, in the m acroeconom y. D ifferent exponents of M one tarism stressed d ifferent propositions, but it would be fair to say that, from the point of view of the non-academ ic observer whose main con cern was the conduct of econom ic policy, M onetarism involved first a theory of inflation, second a theory of the cycle, and third, as a corollary of these, a recom m endation for the conduct of m onetary policy. Specifically, infla tion was said to b e explicable in term s o f the rate of grow th of the money supply, and the cycle, or m ore precisely its turning points, in term s of changes in that rate o f grow th. From these propositions, it immediately seem ed to follow that both inflation and cycles could be avoided by choosing an appropriate rate of grow th for the m oney supply, and binding the m onetary authorities to deliver it y ear in and year out by imposing upon them some quasi constitutional rule of conduct. T he cen tral item on the M onetarist policy agenda was thus to eliminate inflation and stabilize the real econom y by taking discretionary pow er away from the cen tral b an k .2 W hatever position may b e taken about its validity, this M onetarist agenda is nowadays treated as having been (and perhaps as still b e ing) w orthy of serious discussion. It was not so 25 or 30 years ago. Then, though the Phillips curve, about w hich I shall have m ore to say below , was coming onto the scene, the predom i nant view treated inflation as largely a m atter of "cost-push” forces. The cycle was regarded as having its roots in investm ent fluctuations, and some m ixture of wage-price guidelines and fiscal m easures was thought best able to cope with 2ln an earlier essay, Laidler (1982), Ch. 1., I analyzed the essential characteristics of Monetarism from an academic perspective, concentrating there on the role of a stable de mand for money function and the expectations-augmented Phillips curve in defining the doctrine. As the reader will see, these relationships play a large part in the following discussion, and I regard this essay as supplementing rather than in any way contradicting this earlier piece. 3Thus Laidler (1969) used this framework to motivate its discussion of the significance of the demand for money RESERVE BANK OF ST. LOUIS FEDERAL the policy problem s the tw o phenom ena presented. M onetary tools had at best a m inor role to play in the conduct of m acro policy, and the relevant variables w ere, in any event, thought to b e not the quantity of money, but the level and stru ctu re of nominal interest rates. In 1960, say, that as yet un-named body o f doctrine w hich we now know as M onetarism was hardly debated for the simple reason that it was regarded as quite outlandish. This was strange indeed, because at that time no funda m ental questions of econom ic theory separated proponents of the conventional m acroeconom ic wisdom of the early 1960s from their M onetarist critics. Each side in the debate that was to follow could, and did, derive their views from specific quantitative hypotheses about relation ships embodied in what was, nevertheless, quali tatively speaking, essentially the same m acroeco nomic model, that staple o f the contem porary textbooks, the Hicks-Hansen IS-LM fram ew ork .3 The first stage of the M onetarist controversy was about the em pirical nature and stability of the demand fo r m oney function, or, as it was then thought equivalently, the relationship deter mining m oney’s velocity of circulation. As early as 1956, Friedm an had advanced the hypothesis that the demand fo r m oney was a stable fun c tion of a few argum ents, and by 1959 had pro duced em pirical evidence w hich seem ed to show that, as far as real m oney balances w ere concerned, "few " m eant "one”: namely, perm a nent real incom e. The implications of this result w ere startling, because, once fed into the IS-LM fram ew ork, they suggested that if the quantity of m oney was held to an appropriate constant grow th rate, th ere would b e essentially no scope for shocks originating on the real side of the econom y, fo r example, in the investm ent com ponent of aggregate demand, to bring about any significant fluctuations in nominal income. Nor would th ere be any role fo r fiscal m easures to play in influencing that variable. Fu rth er m ore, a slightly later, but essentially com plem en tary, study by Friedm an and M eiselman (1963) function, while Friedman (1971) also used it as an ex pository device. Note, however, that Brunner and Meltzer (e.g., 1976), because of their insistence on the importance of credit market effects in the generation of the money supply, were led to extend this framework to a point at which it became sufficiently different in its characteristics to make it misleading to treat it as simply one more variant on IS-LM. 51 seemed to confirm directly the irrelevance of variations in autonomous expenditure, while at the same time attributing a considerable in fluence on m oney incom e’s behavior to the grow th rate of the money supply. The appealing simplicity of these results did not long survive fu rth er em pirical investigation. A num ber of studies soon found a role for in ter est rates to play in influencing velocity, hence re-opening the theoretical possibility of variables other than m oney affecting the time path of m oney incom e, and the Friedman-M eiselman re sults w ere not robust in the face of small changes in the way in w hich the Keynesian con cepts of autonomous and induced expenditure w ere m easured .4 Nevertheless, from a practical point of view, the m odifications to the basic M onetarist position required by these results w ere rath er minor. The quantity of m oney did seem to be an econom ic variable o f potentially strategic im portance; and real shocks originating in the private sector, not to m ention impulses coming from fiscal policy, did seem to play a potentially less im portant role in determ ining money incom e’s tim e path than the conventional wisdom of about 1960 would have had it. By 1967 results such as these w ere sufficiently well established and widely accepted that it was pos sible for the present w riter to begin w ork on a supplem entary textbook (Laidler, 1969) which summarized the em pirical evidence on the de mand fo r m oney function, and explicitly in ter preted it in term s of the above-m entioned IS-LM model along ju st such lines as these. Now the read er will b e w ell aw are that our earlier confidence in the em pirical stability of the demand fo r m oney function has not been entirely justified by subsequent experience. I will take up this m atter below. For the m om ent it is m ore im portant to con cen trate upon another anomaly in this aspect of the M onetarist con tro versy, namely that em pirical evidence on the stability of the demand for money function, though relevant to M onetarist propositions about the causative role of m oney vis a vis both inflation and the cycle, and hence also to pro posals to tie down m onetary policy by w ay of a grow th rate rule, stops fa r short of establishing them . Thus, if m onetary elem ents in the gen era 4Among early papers finding a significant interest elasticity of demand for money were Meltzer (1963) and Laidler (1966). Both Ando and Modigliani (1965) and De Prano and Mayer (1965) showed that fairly small changes in the definition of “ autonomous" expenditure led to rather large tion of inflation on the one hand, and the cycle on the other, are to b e discussed coherently, one requires m ore than a link betw een the time path of money and money income. One also needs a theory of how fluctuations in money incom e are divided up betw een its price level and real in come com ponents. M oreover, a stable demand fo r money function is quite compatible with the existence of cost-push inflation and a cycle w hose origins lie in the private sector of the econom y, provided only that institutional a r rangem ents are such as to ren d er the supply of m oney as an essentially passive variable. Empirical evidence about the stability o f the demand for money function, that is to say, got M onetarist analysis taken seriously enough for it to becom e controversial, but it could not in and of itself guarantee its victory in a debate w hich rath er centered on the two issues raised above. The first issue becam e known as the problem of Friedm an’s "missing equation,” and debate about it overlapped heavily w ith discus sions o f the "Phillips curve,” otherw ise known as the "inflation-unem ploym ent trade off." As to the second, it was addressed by such w orkers as Philip Cagan (1965), and Bru nn er and M eltzer (e.g., 1964, 1976) who opposed th eir findings to the then "new view” of money propounded by Jam es Tobin (eg., 1969) and his associates. It will be helpful to discuss these tw o aspects of what we might term the second stage of the M onetarist controversy in turn. The problem of decomposing variations in money incom e into th eir real and price level com ponents was a longstanding one in m acroeco nomics, dating b ack at least to Hicks' original (1937) form ulation of the IS-LM fram ew ork as a model of the determ ination of money income. Though subsequent developments of this system reinterpreted it as dealing with real income, that still left prices unexplained. Here, the usual solution was to have them determ ined by the behavior of the m oney wage, and as I have al ready noted, to treat the latter variable as being driven by exogenous "cost-push” factors. How ever, this is not the w hole story, fo r "demandpull” explanations of inflation also had th eir ad herents, and the Phillips (1958) curve relating the rate of m oney wage inflation inversely to effects on the assessment of its influence on aggregate demand. Strangely enough, at this stage of the Monetarist controversy, questions about vagueness in the definition of money were not raised. MARCH/APRIL 1990 52 the level of unem ploym ent soon cam e to be seen as a device fo r synthesizing these two points of view. W age inflation, according to the Phillips curve w as proxim ately caused by an ex cess demand fo r labor, for w hich variable the unem ploym ent rate was a proxy. Such excess demand could eith er result from "pull” forces shifting the labor demand curve, or "push” in fluences shifting the supply curve. In either case, how ever, price inflation would b e d eter mined by the d ifference betw een wage inflation and labor productivity grow th. Hence there seemed to exist a stru ctu ral inverse trade-off b etw een inflation and unemploym ent, or equiva lently a positive relationship betw een the level of real output and inflation .5 From the point o f view of m icroeconom ics, the above analysis was fatally flawed, being based on a theory w hich had the supply and de mand fo r labor determ ine the m oney, ra th er than the real, wage, and it was not long b efore Phelps (1967) and Friedm an (1968) indepen dently pointed this out. The latter, m oreover, brought his critique into the cen ter of the Mon etarist controversy, by noting, first that the o r thodox Phillips curve predicted that an inflation ary m onetary policy could generate perm anent gains in real incom e, and second, that w hen its underpinnings w ere corrected to take account of the m oney wage-real wage distinction, the inflation-unem ploym ent trade-off was reduced to a tem porary phenom enon at most. Hence, a key ingredient of the theory w hich yielded the characteristic M onetarist propositions about money grow th affecting only prices in the long ru n but quantities too, in the short run, was created a ft e r the em pirical observations upon w hich those propositions w ere based had been published, and turned out to be a theory not of money, but of labor m arket behavior. It seem ed to be, m oreover, in its original form , a m odifica tion of a device borrow ed from one b ran ch of the orthodoxy to w hich M onetarism was oppos ed. As a practical m atter, rath er than replace it with some alternative equation, the PhelpsFriedm an critique o f the Phillips curve simply added the expected inflation rate to the relation ship’s right-hand side, w ith a coefficient of unity. 5Phillips’ original analysis was much elaborated by Lipsey (1960), but responsibility for explicitly treating the Phillips curve as a structural relationship constraining policy choice probably rests with Samuelson and Solow (1960). 6 The relevant empirical literature was voluminous, but is surveyed by Laidler and Parkin (1975). FEDERAL RESERVE BANK OF ST. LOUIS As w ith M onetarist propositions about the de mand fo r m oney function, then, questions raised by the M onetarist "co rrectio n ” to the Phillips curve seemed to be inherently em pirical. Either the labor fo rce was immune to m oney illusion in the long run, so that th ere was no perm anent inflation-unem ploym ent trade-off, or it was not, in w hich case such a perm anent trade-off ex isted. This question generated m uch em pirical w ork in the late 1960s and early 1970s, and it is a fair generalization that as m ore and m ore evi dence was added by the passage of time, the m ore the w ork in question cam e to support the M onetarist position. Evidence from the 1950s and early 1960s was com patible w ith the ex istence of an inflation-unem ploym ent trade-off, but the experience of higher inflation rates from the m id-1960s onw ard made this hypothe sis harder and harder to support .6 Once again, though, th ere was no reason why this result could not b e incorporated into the fram ew ork of a suitably extended IS-LM model. However, tw o factors militated against so simple and h ar monious an outcom e to this stage of the Mone tarist controversy. To begin with, the original Phelps-Friedman critique of the Phillips curve had been advanced on theoretical grounds, and b efo re em pirical evidence seem ed to requ ire econom ists to change their notions about labor m arket behavior. W hen confronted with a choice betw een what em pirical evidence seemed to show, and what elem entary econom ic theory required to b e the case, Phelps and Friedm an had chosen the lat ter. E^-post, they w ere vindicated by em pirical evidence, and this vindication served notice on m acroeconom ists that they would b e wise to pay m ore attention to the m icroeconom ic foundations of their em pirical generalizations than had typi cally been the case up until then. Second, Fried m an’s version of the critique had been accom panied by a b rief account of labor m arket b e havior, part of w hich was quite incom patible w ith the then conventional interpretation of real incom e and em ploym ent fluctuations as m anifestations of variations in excess demand in the economy. 53 Conventionally enough by the standards of the 1960s Friedm an described the early stages o f the econom y’s response to a m onetary expan sion in the following term s .7 ...much or most of the rise in income will take the form of an increase in output and employment rather than in prices. People have been expecting prices to be stable, and prices and wages have been set f o r some tim e in the fu tu r e on that basis. It takes time for people to adjust to a new state of demand. Producers will tend to react to the intitial expansion in aggregate demand by increasing output, employees by working longer hours, and the unemployed by taking jobs now offered at former nominal wages. This much is pretty standard doctrine, (p. 103, my italics) However, Friedm an immediately w ent on to elaborate this account of the m echanism s that brought about short-term fluctuations in incom e and employment: Because the selling prices of products typically res pond to an unanticipated increase in nominal de mand faster than prices of factors of production, real wages received have gone down—though real wages anticipated by employees went up, since employees implicitly evaluated the wages offered at the earlier price level. Indeed, the simultaneous fall ejc-post in real wages to employers and rise e^c-ante in real wages to employees is what enabled employment to increase, (pp. 103-04) Unemployment fluctuations in the conventional view of the Phillips curve had been treated as m anifestations of variations in the pressure ex erted by excess demand in goods and labor m arkets characterized by less-than-perfect price flexibility, and h ence likely to b e operating out of equilibrium in the w ake of any shock to ag gregate demand. The first passage quoted above is quite compatible w ith this view. In the second passage quoted, how ever, em ploym ent fluctua tions are explicitly pictured as arising from vol untary decisions taken in response to price changes and, in the case of the suppliers of labor, on the basis of faulty expectations w hich would in due course be corrected. Thus Friedm an’s critique of the Phillips curve potentially involved m uch m ore than the addi tion of an extra variable to the right-hand side of an equation. It also pointed tow ard a funda m ental reinterpretation of the labor m arket behavior underlying it. Viewed with hindsight, it provides as an unm istakable sketch of the m icrofoundations of the short-run aggregate 7Friedman’s (1969) Optimum Quantity of Money reprints many of his seminal contributions, and the page supply curve w hich w as to becom e the central analytic device of New-classical econom ics. As we shall see below, this device would in due course, and quite paradoxically, not strengthen but thoroughly underm ine the very M onetarist explanation of the business cycle to w hich Fried m an’s analysis of inflation-unemployment in ter action was particularly addressed. The analysis of inflation-unemployment in ter action was by no m eans the only area in w hich, during the 1950s and 1960s, m acroeconom ists w ere seeking to strengthen the m icroeconom ic foundations of their analysis. A whole set of questions concerning the determ ination of the money supply, and the m echanism s w hereby m onetary changes might in teract w ith aggregate demand in the economy, w ere also addressed in such term s, both by adherents of the conven tional Keynesian m acroeconom ic wisdom of the time, and by M onetarists. Here as elsew here, the contentious issues w ere m ore em pirical than theoretical. As H arry Johnson noted as early as 1962, th ere was no debate in principle betw een, say, Bru nner and M eltzer on the one hand and Jam es Tobin on the other about the basic nature of the linkages betw een the m onetary and real sectors of the economy. Both saw these as in volving disturbances to the stru ctu re of the port folios of the banking system and the non-bank public generating changes in the relative rates of retu rn on various assets, financial and real, w hich would in tu rn provoke attem pts on the part of agents to resto re equilibrium by way of sales and purchases of various assets. Both sides also agreed that financial institutions and the non-bank public alike should be analyzed as maximizing agents. W hat defined the M onetarist position h ere was not its general approach, but rath er a set of specific hypotheses about the quantitative nature of the responses in question. First, as far as the non-bank public w as concerned, M one tarists took a broad view of the array of assets whose rates of retu rn w ere relevant to what we might term the transm ission m echanism of m on etary policy. Specifically, they argued th at m one tary policy would have effects not just on rates of retu rn born e by financial assets, but also on the implicit rates of retu rn yielded by producer and consum er durables. Hence they saw its ef fects as being both m ore pervasive, and quantireferences here and elsewhere to quotations from his work are to this source unless otherwise noted. MARCH/APRIL 1990 54 tatively m ore significant too, than did proponents o f the Tobinesque "new view.” Fu rtherm ore, w hile not denying that disturbances originating in the private sector of the econom y would im pinge upon the behavior of financial institu tions, so that the quantity of m oney was un doubtedly in this sense an endogenously d eter m ined variable, they nevertheless strongly re sisted the idea that the endogeneity in question also involved that variable being passively demand-determ ined, even w hen the m onetary authorities used short-term in terest rates as th eir policy instrum ent. Such an outcom e was logically possible, to be sure. If in terest rate changes disturbed only the m argin betw een financial assets (let us call them bonds) and money, then a reduction, say, in in terest rates would lead to the public simply of fering bonds to the banks in exchange fo r m on ey with no fu rth er consequences. However, con sistent w ith their broad view of the range of assets relevant to the transm ission m echanism , B ru nn er and M eltzer argued that the principal m argin likely to be disturbed by a change in in terest rates was that betw een bonds and physical capital, including consum er durables. T he public would sell bonds to the banks, of course, but as p art of a process of replacing those bonds with physical capital. Furtherm ore, and crucially, this would be only a first-round effect, fo r the sale of bonds to the banks would b e in exchange fo r newly created money which would have to be accom m odated in the port folios of the non-bank public. Since the banks had presum ably changed interest rates fo r a reason in the first instance, the possibility of their simply acquiescing in the destruction of newly created cash by the public discharging debts to them could be discounted. Hence, fu r th er substitution effects, changes in aggregate demand, and ultim ately in the price level, would be set in motion. M onetarists’ theoretical position vis-a-vis the behavior of the quantity of money in circu la tion, then, may be described succinctly as fol lows. Rather than being a passive dem and-deter m ined variable, m oney also had a separate sup ply function w hich was derivable from analysis of the interaction o f banks and the public in the m arket for bank credit. In turn, the quantity of 8Perhaps the best known of early econometric studies of the money supply function is Brunner and Meltzer (1964). The standard historical study is that of Cagan (1965). Brunner’s (1968) St. Louis Review piece contains perhaps FEDERAL RESERVE BANK OF ST. LOUIS credit w hich banks w ere willing to grant, as well as the term s on w hich they would m ake it available, w ere both subject to a strong (though not unique) influence flowing from the quantity of reserves w hich the central bank made avail able to them . The quantity of m oney was an en dogenous variable, certainly, arising from a complex set of interacting portfolio choices, but it was nevertheless controllable by the m onetary authorities. This theoretical position, m oreover, was supported by a good deal of em pirical evidence, some yielded by form al econom etric studies, and some by less-form al historical w ork to the effect that the behavior of bank reserves, or m ore precisely of the quan tity of high-powered money, was the principal determ inant of the quantity of money in circula tion and that variations in the ratio of highpow ered money to the m oney supply proper could be modeled as the outcom e of systematic maximizing portfolio choices .8 THE SUBVER SIO N OF MONE TAR ISM B Y NEW -CLASSICAL ECONOMICS In the previous section of this essay, I have described the main elem ents of M onetarist doc trine, and have tried to show that they w ere all reasonably well-developed in the academic literature by the end of the 1960s. However, it took the inflationary experience of the early 1970s, particularly in the United States and Britain, to draw popular attention to th at doc trin e by providing, during its early stages, some thing as close to a controlled experim ent as one ever gets in econom ics. In both countries, m one tary expansion preceded an inflation w hich the authorities attributed to cost-push forces and attem pted to control w ith wage-price control program s, and in both countries those program s failed. Though the OPEC-led energy price in creases of 1973 contam inated the later stages of the experim ent by introducing an extraneous cost-side impetus to the upward progress of prices, the early 1970s was nevertheless the last time that wage-price controls w ere deployed as an alternative to m onetary restrain t in an anti inflation policy. M oreover, the fact that the high his single best exposition of the Monetarist position on the generation of the money supply, and of the characteristics distinguishing it from the Tobinesque “ new view.” 55 inflation of the 1970s was com bined with an obvious deterioration in real econom ic p erfo r m ance, rath er than an im provem ent, ensured that M onetarist ideas about the absence of a long-run inflation-unem ploym ent trade-off quick ly becam e conventional wisdom. At the very tim e at w hich M onetarist ideas w ere gaining popular acceptance, how ever, their academ ic foundations began to shift dangerous ly as New-classical econom ics was developed. I have suggested elsew here that the w ork of R obert E. Lucas, Thom as Sargent, Neil W allace and Robert Barro is m ore usefully treated as separate and distinct from M onetarism, and for w hat I still regard as good reason s .9 However, this was and rem ains, something of a minority viewpoint; and the fact rem ains that the two central characteristics of New-classical econom ics, namely the interpretation of the Phillips curve as a reflection of an aggregate supply rela tionship, and the rational expectations hypothe sis, w ere quickly adopted by leading Monetarists. This was unfortunate, because as the M onetarist controversy moved into the 1970s and 1980s, it increasingly becam e, as a m atter of fact, a con troversy about New-classical econom ics. T he m a jority o f econom ists failed to distinguish betw een M onetarism and New-classical econom ics, and accepted Jam es Tobin’s characterization of Newclassical econom ics as M onetarism M ark II .10 W hen New-classical econom ics was academically discredited, so too was M onetarism in general, leading to what, as I shall argue in the final sec tion of this paper, is a dangerous gap in the stru ctu re of contem porary m onetary econom ics. I have already noted that Friedm an’s 1968 analysis of the inflation-unem ploym ent trade-off relied on tw o incom patible theories of labor m arket behavior. In the passages quoted earlier, we had quantities moving in stead o f wages and prices, w hich had already been set and hence w ere unable immediately to change; and we also had quantities moving in resp o n se to asym m etrically perceived m oney-price changes. W e had, in short, both an inform al account of the effects of wage and price stickiness w hich was, as Friedm an said, "pretty standard d octrine” in the m acroeconom ics o f the 1960s; but we also 9This argument is developed in some detail in Laidler (1982), Ch. 1. 10This characterization of Tobin’s is developed and defend ed by him in Tobin (1981), in a paper prepared for the same conference as the above mentioned Laidler (1982), Ch. 1. had an unm istakable sketch of an aggregate sup ply curve interpretation of the short-run Phillips curve in w hich money-wage and price flexibility was of the essence. T h ere is little point in speculating about the extent to w hich Friedm an was aw are of the ten sions inherent in his analysis in 1968; but it is interesting to note that, at that time, he ch arac terized Phillips’ w ork as "...containing a basic defect—the failure to distinguish betw een n om i nal wages and rea l w ages...” (p. 102, Friedm an’s italics) w hereas in 1975 he referred not only to this point, but also attributed to Phillips an e r ro r in having "...taken the level of em p loy m en t [instead of the rate of change of prices] as the independent variable...” in the relationship. By 1975 Friedm an clearly was aw are that th ere was a choice to be made concerning the m icroeco nomic underpinnings of the Phillips curve and had explicitly rejected that w hich hinged on in terpreting the unemploym ent rate as a proxy variable fo r some excess demand fo r labor con cept. Bru nner and M eltzer made a similar choice concerning the modeling of the linkages betw een output and price-level variations in designing the basic fram ew ork w hich they and th eir asso ciates used to analyze the inflationary process in a num ber of countries in the mid- 1970s. Like Friedman too, they chose to model expecta tions as being form ed not adaptively but, follow ing Lucas and Sargent and W allace, rationally, as the inflation forecast of the "tru e model” of the econom y under analysis .11 These developments seemed at the tim e to be in no sense revolutionary. I have already re m arked that the Phelps-Friedman critique of the Phillips curve initially involved using simple mi croeconom ic analysis to mount a theoretical at tack on what at the tim e seem ed like an hypothe sis well-supported by em pirical evidence; and in the event, m icroeconom ic principles proved a sounder guide than what turned out to be some misleadingly special observations. To show that price-output interaction could b e derived from a supply and demand apparatus w ithout reso rt to purely em pirical generalizations about price stickiness, and to show that the analysis in ques11The relevant work here is contained in Brunner and Meltzer, eds. (1978). Note that the details of their formula tion of the aggregate supply curve, with output’s rate of change rather than level playing a major role, set it apart from the standard Lucas (e.g., 1973) formulation. These matters, discussed by McCallum (1978), are not central to the matter under discussion here. MARCH/APRIL 1990 56 tion was com patible with agents making use of "all available" inform ation in a utility m aximiz ing fashion, simply seem ed to be making even m ore secu re the m icroeconom ic foundations of a particular piece of m acroeconom ics, and hence to be rendering it less prone to excessive dependence on theoretically unsupported em pirical observations of a type that had proved so misleading in the recen t past. Also, and crucially, New-classical m acroeconom ics still seem ed to yield M onetarist implications, nam ely that inflation in the long ru n was a m on etary phenom enon, and that, in the short run, so was the cycle. To be sure, it broadened the m enu of m onetary rules that would enable the cycle to be avoided to any that the general pub lic could understand and th erefo re use as a basis for expectations form ation, but a constant grow th rate rule w as still a particularly simple, and th erefo re viable, item on the m enu in ques tion, and it was of course a particularly appro priate choice if price stability in the long run was added to the elimination of the cycle as a proper goal fo r m onetary policy. Even in the m id-1970s, it should have been apparent that the attem pt to underpin the M one tarist position with New-classical foundations was dangerous. As early as 1958, Friedm an had suggested that m onetary policy affected the econom y w ith a "long and variable” time lag. T he evidence w ith w hich he supported this sug gestion identified a change in m onetary policy as a change in the rate of grow th o f the quanti ty of money in the econom y; and it showed that dow nturns in that rate of grow th occu rred on average 16 m onths b efore the corresponding upper turning point o f the cycle, while upturns in money grow th led cyclical troughs by about twelve months. Fu rtherm ore, as Friedman (1987) was later to note m ore explicitly, the e f fect of changed money grow th on nominal incom e 12 ...typically shows up first in output and hardly at all in prices. If the rate of monetary growth increases or decreases, the rate of growth of nominal income and also of physical output tends to increase or de crease about six to nine months later, but the rate of price rise is affected very little...the effect on prices comes some 12 to 18 months later...(p. 17) 12The reader’s attention is drawn to the recent (1987) vin tage of this statement. I have not been able to find so clearcut and concise an exposition of the point in Fried man’s earlier work, though I believe that the basic message contained in the passage quoted here can be distilled from the evidence presented in Friedman (1969), Chs. 10-12. RESERVE BANK OF ST. LOUIS FEDERAL In 1963, Friedm an and Schw artz had presented a m ore elaborate analysis of m oney’s role in generating the cycle, in the course of w hich they explained the length of the tim e lags in volved in term s that amounted to an inform al sketch of a dynamic version of the M onetarist version of the transm ission m echanism w hich Bru nner and M eltzer w ere also expounding at the time. The central element in the transmission mecha nism...is the concept of cyclical fluctuations as the outcome of balance sheet adjustments, as the effects on flows of adjustments between desired and actual stocks. It is this interconnection of stocks and flows that stretches the effects of shocks out in time, pro duces a diffusion over different economic categories, and gives rise to cyclical reaction mechanisms. The stocks serve as buffers or shock absorbers of initial changes in rates of flow, by expanding or contracting from their "normal” or "natural” or "desired” state, and then slowly alter other flows as holders try to regain that state, (p.234) The em pirical evidence referred to here, and particularly that part of it dealing with the tim ing of output responses relative to those in prices, and an explanation of that evidence in term s of a transm ission m echanism involving portfolio disequilibria w orking them selves slow ly out over real historical time are quite incom patible with New-classical analysis. Since both the evidence in question and the above explana tion of it w ere available and well established b efore the development of New-classical ideas, the incompatibility in question ought to have prevented those ideas from being adopted as a basis for M onetarist propositions, but it did not. W h ether this was because the problem was not fully appreciated at the time, or because that shift in methodological priorities away from em pirical evidence and tow ard "sound" theoretical foundations upon w hich I have already com m ented caused those who w ere aw are of it to opt for the latter is hard to say. I suspect that a strong elem ent of the latter consideration must have been at work, though, since the inconsisten cy in question is hardly subtle .13 To begin with, the price flexibility postulate of New-classical econom ics creates problem s fo r the 13Though I do not claim to have understood this matter fully from the outset, I did discuss it in some detail as early as 1978 in an essay reprinted as Chapter 4 of Laidler (1982). 57 M onetarist account of the transm ission m e chanism . Slow adjustm ent of portfolios in the wake of a m onetary disturbance is of the very essence here, and it is usual to explain the slow ness in question in term s of transactions costs. But if the price level moves freely and instanta neously to keep m arkets cleared it also, in the process, eliminates portfolio disequilibria quite costlessly fo r agents. According to Friedm an and Schw artz, excessive m oney holdings develop b e cause, w hen m oney grow th increases, the rate of inflation does not respond immediately. But according to New-classical econom ics it does, and so increased m oney grow th cannot cause a tem porary rise in buffer-stocks of m oney as a prelim inary to increased expenditure flows. The conflict h ere betw een traditional M onetarism and New-classical analysis concerns rival th eo ret ical constructions. Much m ore serious is the conflict betw een theory and evidence which arises w hen the New-classical aggregate supply curve is confronted w ith the em pirical evidence concerning the interaction of money, output and prices over the course of the cycle, evidence upon w hich traditional M onetarism laid con siderable stress. should have led m onetarists to reject that theory from the outset, did eventually underm ine it. So long as em pirical w ork was confined to testing the proposition that output and employment fluctuations could be modeled as responses to "unanticipated” money, all seemed well. How ever, Robert J. B arro (1978) noted that the theo ry in question made specific predictions about the relationship betw een m onetary shocks and price level changes as well. W hen he cam e to test the latter predictions, he found that, in order to reconcile them w ith his data, the price level’s response to unanticipated m onetary shocks had to be characterized by a rath er slowmoving distributed lag process, in an econom y in w hich, how ever, prices could respond with no lag to anticipated money. Fu rtherm ore, his results also seemed to require that the aggregate demand fo r m oney display a greater sensitivity to transitory than to perm anent changes in in come, the very opposite result to that implied by a wide variety of oth er studies. New-classical analysis could be forced to fit the data generated by the U. S. econom y, that is to say, only by way of some extrem ely implausible subsidiary assumptions. As is well known, the New-classical aggregate supply curve explanation of the decomposition o f nom inal incom e changes into th eir real and price-level com ponents hinges upon the distinc tion betw een demand side shocks whose price level effects can be anticipated by agents, and those that cannot. The w ord “anticipated” nor mally m eans "expected an d a cted u pon ,” but since the New-classical model is one in w hich flexible prices always move costlessly and in stantly to equate supply and demand in all m ar kets, the second phrase is redundant in its con text. If a price level change is anticipated by agents, th ere will be no quantity changes associ ated w ith it; but if it is not, then the specific money-price changes in particular m arkets that are associated w ith it will be m isinterpreted as reflecting relative price changes and voluntary responses in quantities of goods and services supplied will occur. Cyclical fluctuations in real variables may th erefo re be interpreted as the consequence of unanticipated price level changes. The trouble h ere is that it is hard to see how output and employment fluctuations can be re sponses to price level fluctuations i f th ey p r e c e d e th o se p r ice level flu ctu ation s; but the em pirical evidence tells us that they do just that. Nor did the results of subsequent em pirical w ork enhance New-classical econom ics’ claim to be taken seriously. Mishkin (1982) repeated Barro ’s tests of the irrelevance of anticipated m onetary shocks fo r output using m ore sophisti cated econom etric techniques, and found that this resulted in a reversal of the initial results— apparently anticipated m onetary changes did have real effects. Boschen and Grossman (1982) noted that m oney supply estim ates are published weekly, and that only the erro rs in these esti mates properly can be regarded as constituting the unanticipated com ponent of the m oney sup ply. They fu rth er noted that the latter w ere im plausibly small to form the basis o f a m onetary explanation of the cycle, and that output changes w ere in fact correlated w ith m onetary changes about w hich inform ation had previously been published. This inconsistency of the timing o f data with the basic stru ctu re o f New-classical theory, which In addition to these problem s raised by aca demic w ork, of course, th ere was the experience of the early 1980s recession, w hich played the same role in publicly discrediting New-classical econom ics, as did the inflation of the 1970s in undermining "Keynesian" theory. New-classical econom ics discounted the im portance of real world wage and price rigidities, and placed con siderable faith, th erefore, in the public’s will ingness to react immediately to a well-publicized MARCH/APRIL 1990 58 anti-inflation policy based on m onetary con trac tion in such a w ay as to reduce inflation without serious real incom e and employment conse quences. T h e policies in question certainly did reduce inflation, but the recession w hich accom panied that reduction was, in some respects, the w orst since the 1930s. M oreover, the United Kingdom and Canada carried out similar experi m ents at about the same time with similar results. By the m id-1980s, the New-classical de velopm ent o f the M onetarist account of the role of m oney generating the business cycle was thus widely recognized to have failed in its en cou nter w ith em pirical evidence. This could have led to a revival of interest in m ore tradi tional M onetarist analysis; but it did not, because the very postulate that had established the re spectability of that analysis in the first place, namely a stable aggregate demand fo r money function, had also ru n into difficulties. THE KEYNESIAN A D O P T IO N OF THE STABLE D EM AN D FOR M ONEY FUNCTION The early studies of the aggregate demand for money function, w hich established M onetarism ’s respectability, w ere carried out using long runs of U.S. data, some stretching back into the 19th century. M oreover, the data them selves w ere highly tim e aggregated. Thus Friedm an’s (1959) seminal study covered the years 1869-1956, and used cycle averages of variables in estim ating its basic equation. One business cycle, th at is to say, lasting on average about fou r years, provid ed one observation. Later studies, such as those of M eltzer (1963) or Laidler (1966) dealt with essentially the same time period, but used an nual observations. It was an obvious enough ex tension of such w ork to test the hypotheses at stake in it against data drawn from other coun tries and also against m ore tim e disaggregated data too, and such extension proceeded apace in the 1960s mainly at the hands of people far m ore interested in exploiting new data and com puting techniques than furthering any par ticular policy agenda. portion to the general price level—displayed rem arkable robustness; so m uch so that, by the early 1970s, the demand fo r money function was a prim e candidate to becom e the c en ter piece of stabilization policy, m uch as had the Keynesian consumption function or the Phillips curve at earlier times. For the demand fo r m on ey function’s full potential fo r such a use to be exploited by policy m akers, detailed knowledge of the function’s contem porary form was ob viously needed, and in the early 1970s a rem ark ably simple version of the equation seemed to be able to deal with quarterly U. S. data with a high degree of precision. This relationship, nowadays know n as the "Goldfeld equation” after its m ost carefu l exponent (Goldfeld, 1973), had the long-run average value of m oney holdings determ ined by real incom e and in terest rates, but involved the hypothesis that w hen some disturbance took money holdings away from this long-run average, they would move back tow ard it slowly over time, w ith the speed of adjustm ent in question being proportional to the size of the gap to be closed .14 The Goldfeld equation fitted U. S. data well, appeared to be stable over time, and crucial for policy pur poses, dealt w ith data at a relatively low degree of time aggregation. It did indeed appear to be so policy relevant that, in his Presidential ad dress to the Am erican Econom ic Association, Franco Modigliani (1977) argued that it could, and should b e used as the basis of an activist m onetary stabilization policy in the Keynesian mold. At first the hypotheses in question—that the demand fo r m oney varied w ith some real in com e or w ealth m easure, some m easure of the opportunity cost of holding money, and in pro Modigliani’s address in fact appeared after the first signs of trouble with the relationship had appeared. Before dealing w ith that, how ever, it is w orth reiterating that the progress of the de mand fo r m oney function so briefly dealt with above was by no m eans synonymous w ith the progress of M onetarism, but rath er it involved a com ponent of M onetarism being taken over by the Keynesian opposition. Ju st as the association of M onetarist ideas about the cycle w ith those of New-classical econom ics was to prove de structive, so too was this association, and once m ore this could have been, but was not, dis cerned at the tim e in the light of evidence then available. That the association in question was indeed being form ed was, o f course, obvious enough once Modigliani made the existence of a 14Though the relationship in question is now irrevocably known as the “ Goldfeld Equation,” it was in fact used before him in studies of the demand for money by, among others, Teigen (1964) and Chow (1966). Goldfeld, be it ex plicitly noted, did not claim originality for the relationship in question. RESERVE BANK OF ST. LOUIS FEDERAL 59 stable demand for money function an im portant part of the basis o f his case for m onetary finetuning, a policy stance to w hich Monetarism was root and b ran ch opposed; and I am not here claiming otherw ise. However, I am also claiming that the equation fitted by Goldfeld, and the interpretation he put upon it, w ere quite antithetical to earlier M onetarist ideas about the demand fo r m oney function in par ticular, and the role of money in the econom y in general. To begin with, the demand for money fun c tion fo r w hich Friedm an had initially claimed stability was the "long ru n ” relationship. In 1959 he had tested it against cycle average data, among other reasons, in ord er to abstract from the complex interactions among money, output, prices and interest rates that characterized the cycle, and which would tend to obscure the underlying stability of the relationship in ques tion. Nor was such a procedure a quirk of one particular paper: the em pirical w ork of Fried man and Schw artz’s M onetary T ren ds..., though not published until 1982, was largely completed in the late 1960s and was based upon cycle phase average data. The apparent stability of demand for money functions such as Goldfeld’s, w hich used quarterly data and hence w ere dominated by within-cycle interactions among variables, should have been a source of puzzle m ent to Monetarists, therefore, not of satisfaction. M onetarists should also have seen that Goldfeld’s interpretation of his equation ran quite counter to th eir ideas about the transm is sion m echanism .15 He used the m oney supply as the dependent variable of his relationship, and treated it as responding passively, albeit w ith a distributed lag, to variations in the argum ents of his demand for money function. It is of the very essence of the M onetarist view that there exists a supply function of money that is in dependent of the demand function, and that the interaction over time of the m oney supply, in come, interest rates and the price level involves 15The first sign of discomfort about this matter that I can find in my own writings appears on pp. 143-44 of the se cond (1977) edition of my Demand for Money, where I refer to there being a fallacy of composition involved in pro ceeding from the individual to the market experiment when analyzing adjustment processes in the demand for money function. Chapter 2 of Laidler (1982) developed my doubts about all this in much more detail. 16And this decline in velocity was associated with rapid money growth failing to produce renewed inflation in the causation running predominantly, though not uniquely, from m oney to the oth er variables. Goldfeld’s w ork on the demand fo r money, and many oth er studies in the same vein, w ere thus based implicitly on the "new view" of the money supply process discussed above, of w hich Brun n er (1968) had been so critical. And indeed, in the 1970s, as central banks becam e interested in controlling the tim e path of m onetary aggre gates, their procedures involved m easuring or forecasting the values of all the right-hand-side variables of a Goldfeld-style equation except the interest rate, and then setting the latter in order to achieve a value of the quantity of money demanded equal to the money supply target. Clearly such a procedure left no room for tak ing account o f the subtle interactions among m arkets for money, credit and equity that lay at the h eart of the M onetarist view of the m atter. Be that as it may, the stability of the Goldfeld short-run demand fo r money function that ought to have puzzled M onetarists did not last long. By 1976 he was inviting his readers to solve "The Case of the Missing Money,” while by the early 1980s, a num ber of com m entators w ere contem plating an unexplained decline in m oney’s velocity of circulation.16 Nor w ere pro blems with the demand fo r money function a uniquely Am erican phenom enon; Canada and the United Kingdom too had problem s with bad ly behaved demand for money functions in the same years; so did Australia and New Zealand a little later; and this list is far from exhaustive. T h ere is little point h ere in attempting a survey of the voluminous literature generated by these events. Suffice it to say that a wide variety of a d h o c explanations, often relying upon par ticular institutional changes w ere proposed, and often (not always) proved fragile. The upshot was that in the eyes of the m ajority of com m en tators the postulate of a stable aggregate de mand fo r m oney function was discredited, and along with it the very foundation of Monetarism. recovery from the 1982 recession. Monetarists should not have predicted renewed inflation then, because actual and expected inflation fell dramatically in the preceding downswing, and hence should have led to an increase in the demand for real balances which could be met only by a falling price level or a growing nominal money supply. What ought to have been done, and what was done, however, are different matters, and careless Monetarist predictions at this time did help to discredit the doctrine. MARCH/APRIL 1990 60 Of course this upshot was preposterous. The stability that M onetarism had from the outset attributed to the demand for m oney function was long ru n in nature, and the relationships that collapsed w ere short-run form ulations, espoused by Keynesian econom ists intent on establishing a basis fo r a policy of m onetary fine-tuning. M oreover, those relationships w ere based on a view of the money supply process w hich M onetarists had vigorously opposed from the earliest stages o f the controversy. An alter native in feren ce to b e made from the collapse of the em pirical stability of short-run demand fo r money functions from the m id-1970s on w ard was that this was the result of their failure to model the dynamic relations among the quantity o f money, interest rates, real in com e and prices as the outcom e of the in terac tion of an independent supply o f money fun c tion with the argum ents of the demand fun c tion; that the problem stem m ed not from in stability of the latter relationship at all, but from the inability of simple single equation distributed lag techniques to com e to grips with the dynamic com plexities involved in the transm ission m echanism of m onetary policy. Those of us w ho advanced the latter explana tion, how ever, did not find m uch of a sym pathetic audience, even among M onetarists. No doubt this was partly because w e did not make our case as clearly as w e might have done. But it was mainly because the case in question had as a key com ponent the notion that m arkets failed to clear instantaneously in the m anner demanded by New-classical econom ics. Rather, it was argued that the transm ission m echanism w orked along the portfolio disequilibrium lines sketched by Friedm an and Schw artz (1963) in the passage quoted on page 56. In the w ake of the success of the Phelps-Friedman critique of the Phillips curve, M onetarism had com e to at tach great im portance to adhering to "sound” m icroeconom ic foundations, so m uch so that it had, as we have seen, becom e intertw ined with New-classical econom ics; and New-classical eco nom ics was unable to tolerate such “disequilib rium ” analysis.17 As has already been noted, the failure of New-classical business cycle theory in the early 1980s was widely regarded as a failure 17Some of the problem here stemmed from semantics. If “ disequilibrium” behavior is read as synonymous with “ unplanned” behavior, then it is understandable that an economist would not wish to rely on it in constructing an economic model. If it means merely behavior incompatible with the existence of continuous competitive equilibrium in FEDERAL RESERVE BANK OF ST. LOUIS of the M onetarist view of the phenom enon; and w hat I am now arguing is that this same associa tion with New-classical ideas prevented M one tarism from deploying its own earlier analysis o f the transm ission m echanism as a defense against an attack on another of its key com po nents. Em pirical evidence about the instability o f the short-run demand for money function ought not to have been interpreted as u nd er mining M onetarist propositions about the stabili ty of the long-run relationship, but it was. By the early 1980s, the M onetarist controversy was over, w ith M onetarism discredited in the eyes of most observers. THE LEGACY OF THE CO NTR O VER SY The M onetarist controversy was concerned with policy, but the issues involved in it did, at the time, pose questions w hich defined a fro n tier of academ ic research in m onetary theory. Fu rtherm ore, and again at the time, the latest in econom etric techniques w ere applied to the in vestigation of the em pirical questions w hich the controversy raised. T h ere was nothing special about all this. T he Keynesian revolution too had been simultaneously about theory, em pirical evidence and policy, and so had virtually every previous debate in the history of m onetary eco nomics. The end of the M onetarist controversy, how ever, ushered in a period during w hich m onetary econom ics began to disintegrate into relatively self-contained bodies of theoretical w ork on the one hand and em pirical policy-related w ork on the other. W h eth er this disinte gration is a tem porary or perm anent phenom e non, only tim e will tell. New-classical econom ics was underpinned by a strong methodological p referen ce on the part of its exponents for grounding m acroeconom ic theorizing on explicit m icroeconom ic founda tions. Such foundations, how ever, are capable of yielding a w ider variety of m acro models than the m onetary explanations o f the business cycle that lay at the h eart of the w ork of Lucas and Sargent and W allace. Thus the em pirical failure of those models did not lead to the abanall markets, then it is surely more acceptable. The fact re mains that those of us who used it in the latter sense were read as using it in the former, and were not sufficiently careful to explain ourselves. The result was a considerable amount of unconstructive debate and confusion among Monetarists. 61 donm ent of the methodological agenda which had produced them , but m erely to an attem pt to replace them w ith an alternative explanation of the cycle w ith equally well, o r even better, defined m icro prem ises. I re fe r h ere to that body of research know n as "real business cycle theory” w hich is based on a stochastic version of the New-classical grow th model of Meade, Swan and Solow, and attributes cyclical fluctua tions to exogenous shocks to the aggregate pro duction function, in m uch the same way as, over a century ago, Jevons attributed the cycle to fluctuations in agricultural productivity associated with sunspot activity. T he exponents of real-business cycle theory, though hostile to econom etric testing, n everthe less do not ignore em pirical evidence, and have begun to address the question why, if it is not a causative factor in cyclical fluctuations, th ere are nevertheless system atic relations among the quantity of m oney and oth er variables. The very m anner in w hich the question is posed vir tually dictates the answ er offered, nam ely that the relations in question are the result of reverse causation running from the cycle to money, ra th er than vice versa. Thus, in what is surely one of the g reater ironies in the recen t history of econom ics, a research agenda which is widely regarded as a direct descendant of the M onetarism of the 1960s has ended up adopting a view of the role of m oney in the econom y di rectly opposed to that of its intellectual antece dent, and virtually identical to that of the most extrem e form of "post-Keynesian” econom ics.18 This is no accident, fo r a view of the world w hich has m arkets functioning perfectly and w ithout friction leaves no m ore room fo r m oney to play an im portant role than does a view in w hich m arkets do not function at all. Though one im portant group among the aca demic heirs of M onetarism has thus system atical ly adopted hypotheses that downgrade the im portance of m onetary phenom ena, and has also abandoned traditional econom etric m ethods as a basis for em pirical research , this does not mean that econom etric w ork on m onetary econom ics ceased in the early 1980s. On the contrary, a diverse body of contributors, including unrepentantly old-fashioned M onetarists, econom etri cians in search of an area in w hich to try out new techniques, not to m ention econom ists as 18l refer here to King and Plosser (1984) which uses a form of the old “ money-income-causes-money” hypothesis to sociated with central banks in various countries w hich do, after all, still have to carry out m one tary policy regardless of the state of academic opinion concerning its im portance, have gen er ated an extensive em pirical literature on the de mand for m oney function during the 1980s. The literature in question has been lively and, as I shall now argue, productive in tw o lines of inquiry. First, questions arising from the fact of in stitutional change in the m onetary sector have b een examined using both historical and con tem porary data. In both cases, this phenom enon has been found to be sometim es im portant. Thus Bordo and Jonung (1987), using data going back to the 19th century fo r five countries, have shown that the slow decline in velocity which occu rred largely b efore the first world w ar seems to have b een associated w ith the increas ing degree of m onetization of the econom ies in volved, and its later rise w ith the increasing ef ficiency of m onetary exchange. Closer to our own time, the sharp increase in the velocity of M l balances in Canada that occu rred at the tu rn of the 1980s has been shown, beyond reasonable doubt, to have been associated with the simultaneous spread of daily in terest ch eck ing accounts. As to the United States, the w ork of Barnett (e.g., 1990) using Divisia aggregates, provides evidence that various instances of financial deregulation have produced shifts in the demand fo r m ore conventionally m easured aggregates; w hile the very fact that it proved necessary not so long ago to redefine those ag gregates is itself testim ony to the im portance of institutional developments. The other line of inquiry to w hich I refer above has involved the application of far m ore sophisticated techniques than w ere previously available, both to the explicit modeling of the so-called short-run demand fo r money function, and closely related, to the task of extracting in form ation about the underlying long-run rela tionship from data w ith a high degree of time disaggregation. Here the results have been quite startling. Short-run dynamics dominate q u arter ly data and have to be modeled w ith a great deal m ore care than exponents of, say, the Goldfeld equation brought to b ear or, given the state of econom etric technique, could have been expected to bring to bear, on the task. Never- reconcile monetary phenomena with a productivity-shock theory of fluctuations in real variables. MARCH/APRIL 1990 62 theless, once this is done, stable long-run rela tionships, very m uch like those w hich w ere first estim ated by Friedm an (1959), M eltzer (1963) or Laidler (1966), are after all to be found buried in those data. This same result em erges with pow erful simplicity from a recen t "low -tech” study by Robert E. Lucas (1988), w ho shows that data fo r the last 25 years or so of United States history are still scattered (albeit widely, and w ith com plex serial correlation) around a velocity function directly derived from M eltzer’s (1963) estim ates of the long-run demand for money. Nor are results of this sort confined to the United States. Similar conclusions arise w hen British, Canadian or Japanese data are analyzed. As yet, this em pirical evidence has not at tracted the academ ic attention it deserves, large ly, I suspect, because the treatm ent of the short-run dynamics in the studies which generate it has been based m ore on econom etric technique than econom ic theory. Exponents of the so-called "bu ffer-stock” approach to model ing the demand for money, w ho have of course followed up the analysis of the transm ission m echanism of Bru nner and M eltzer as well as Friedm an and Schw artz, have tried to bridge this gap. They have tended to ground their ex plicit analysis on rath er simple special cases, how ever, w hich have not proved empirically robust; and this in tu rn has led to an identifica tion of the general approach with those special cases and a tendency to dismiss prem aturely the broad insight w hich it yields: namely that the dynamics of w hat we have learned to call the short-run d em a n d f o r m o n ey fu n ctio n are not the property of a structural relationship at all, b ut rath er reflect that complex interaction of the quantity of money with other variables to w hich the M onetarists of the 1950s and 1960s used to re fe r as a transm ission m ech an ism s u b je c t to long a n d variable lags.19 The M onetarist controversy was about theory and em pirical evidence, to be sure, but it was also about m onetary policy; and h ere it has left its strongest m ark. To begin with, the idea that 19For an informal but more complete exposition of the arguments involved here, see Laidler (1987). One of the better known pioneering special case empirical formula tions of the approach is that of Carr and Darby (1981), who refer to it as a “ shock-absorber” approach. 20Let it be clear, however, that I do not regard “ real business cycle theory” as a total waste of time, and that I do think it merits policymakers’ attention. Though I find it hard to believe that shocks to technology will turn out to FEDERAL RESERVE BANK OF ST. LOUIS inflation is fundam entally a m onetary phenom e non, so outlandish in the 1950s, has by now becom e something close to conventional wisdom. W e nowadays h ear very little about cost-push forces and the need to control them w ith wage and price guidelines, controls and so on. Closely related, cen tral banks routinely pay attention to the behavior of m onetary aggregates in design ing their anti-inflation policies. Though even such hard-core M onetarists as Brunner and M eltzer (1987) no longer argue that the domi nant impulse driving the business cycle is always the quantity of money, neither they, nor the practitioners of m onetary policy, have shown the slightest sign of taking the productivity shock hypothesis of the real business cycle theorists seriously.20 Rather, the exclusively monetary inter pretation of the cycle has been replaced by an eclectic approach in w hich m onetary factors have an always potentially im portant, and som e times an actually im portant, role to play. That is why the pursuit of price stability by m onetary means is tem pered by caution concerning the short-term costs that could b e in cu rred if the pursuit in question was to becom e too vigorous. And eclecticism about the causes o f cyclical in stability has not been accompanied by a revival of interest in the use of fiscal policy fo r stabili zation purposes—though this probably has as m uch to do with the fiscal deficits that are the legacy of supply-side econom ics as with any les sons learned during the M onetarist controversy itself. A superficial reading o f the above record might suggest that, w hatever its academ ic stand ing, M onetarism is alive and well in policy cir cles. That it has left a lasting m ark in those cir cles is beyond doubt, but its success on the poli cy fron t has b een fa r from com plete. Indeed, on one interpretation of the evidence, M onetarism ’s partial success h ere may have b een w orse than failure. The M onetarist agenda, after all, involved establishing the im portance of the quantity of money as a policy variable a s a prelim in ary step to rem ovin g fr o m the m on etary au thorities their d iscretion ary con trol o v er it; and the record be the sole or even main source of real world cyclical fluc tuations, it is nevertheless the case that, from a theoretical point of view, such shocks as do originate in such phenomena are likely to be welfare improving. The valuable message of real business cycle theory, then, is that we should not take it for granted that stabilizing the cycle is always and everywhere desirable. It might not be in particular instances. That reminder is surely worth having. 63 shows that only this prelim inary step was com pleted. M onetary policy plays a fa r m ore im por tant role in m acroeconom ic affairs now than it did 30 years ago, but those in control of it have, as a result, m uch m ore pow er than previously, not less, as the M onetarist agenda intended. This cannot really be helped, because part of the academ ic legacy of the M onetarist con troversy has been a g reater appreciation of the role of institutional change in influencing veloci ty’s long-run behavior; but once that role is ad mitted, the case fo r tying down m onetary policy w ith rules becom es fa r m ore difficult, perhaps impossible, to make. All this, though, poses a challenge. If m onetary policy cannot be tied down by rules, then the next-best solution is to subject it to the discipline of constant public scrutiny and criticism . Aca demic econom ics has a large role to play here, in providing the intellectual basis fo r such ac tivity, but at the m om ent it is not in good condi tion to do so. W e do have, as I have noted above, a large and growing body of em pirical w ork w hich points to the long-run stability (not constancy, note) of m oney’s velocity, institution al change notwithstanding. That same body of w ork, how ever, tells us that the short-run dy nam ics of that function are at least as complex as anyone thought they w ere 30 years ago; and we are no fu rth er forw ard now than we w ere then in understanding just why this is the case. At the same time, the gap betw een theoretical w ork on the role of m oney in the economy, and our em pirical knowledge in the area, w hich was opened up w hen the M onetarist controversy becam e a debate about New-classical econom ics and began to pay m ore attention to m icrofoun dations than to data, still rem ains. Intellectual vacuum s of this sort rarely rem ain unfilled for long, but until this one attracts m ore attention than it has to date, the M onetarist controversy’s legacy must be judged to be a distinctly uncom fortable one. The controversy weakened the links betw een academ ic research and contem porary policy w hich in the past made such r e search a natural source of constraining criticism of the conduct of policy, while simultaneously enhancing the discretionary pow ers of policy m akers. That is hardly what its participants intended. Ando, A., and F. Modigliani. “ The Relative Stability of Mone tary Velocity and the Investment Multiplier,” American Economic Review (September 1965), pp. 693-728. Barnett, W. A. “ Developments in Monetary Aggregation Theory,” Journal of Policy Modeling (forthcoming). Barro, R. J. “ Unanticipated Money, Output, and the Price Level in the United States,” Journal of Political Economy (August 1978), pp. 549-80. Bordo, M. D., and L. Jonung. The Long-Run Behavior of the Velocity of Circulation: The International Evidence (Cam bridge University Press, 1987). Boschen, J. F., and H. I. Grossman. “ Tests of Equilibrium Macroeconomics Using Contemporaneous Monetary Data,” Journal of Monetary Economics (November 1982), pp. 309-33. Brunner, K. “ The Role of Money and Monetary Policy,” this Review (July 1968), reprinted in this Review (September/Oc tober 1989), pp. 4-22. _______ and A. H. Meltzer. “ Some Further Investigations of Demand and Supply Functions for Money,” Journal of Finance (May 1964), pp. 240-83. _______ a n d ________. “An Aggregative Theory for a Closed Economy,” in J. Stein, ed. Monetarism (Amsterdam: North Holland, 1976), pp. 69-103. _______ a n d _______ . “ Money and the Economy: Issues in Monetary Analysis,” The 1987 Raffaele Mattioli Lectures, Carnegie Mellon University, mimeo. _______ a n d ________ “ The Problem of Inflation,” in K. Brunner and A. H. Meltzer, eds., The Problem of Inflation, Carnegie-Rochester Conference Series on Public Policy (Vol. 29, 1988), pp. 137-68 Cagan, P Determinants and Effects of Changes in the Stock . of Money, 1875-1960 (Columbia University Press, 1965). Carr, J., and M. R. Darby. “ The Role of Money Supply Shocks in the Short-Run Demand for Money,” Journal of Monetary Economics (September 1981), pp. 183-99. Chow, G. “ On the Long-Run and Short-Run Demand for Money,” Journal of Political Economy (April 1966), pp. 111-31. DePrano, M., and T. Mayer. “ Tests of the Relative Importance of Autonomous Expenditures and Money,” American Economic Review (September 1965), pp. 729-52. Friedman, M. “ The Quantity Theory of Money — A Restate ment,” in Studies in the Quantity Theory of Money (Universi ty of Chicago Press, 1956), pp. 3-21. _______ “ The Demand for Money: Some Theoretical and Empirical Results,” Journal of Political Economy (August 1959), reprinted in Milton Friedman, The Optimum Quantity of Money and Other Essays (Aldine Publishing Co., 1969), pp. 111-39. _______ “ The Role of Monetary Policy,” American Economic Review (March 1968), reprinted in Milton Friedman, The Optimum Quantity of Money and Other Essays (Aldine Publishing Co., 1969), pp. 95-110. _______ The Optimum Quantity of Money and Other Essays (Aldine Publishing Co., 1969). _______ “A Monetary Theory of Nominal Income,” Journal of Political Economy (March/April 1971), pp. 323-37. REFERENCES _______ Unemployment versus Inflation? An Evaluation of the Phillips Curve (London: Institute of Economic Affairs, 1975). Andersen, L. C., and J. L. Jordan. “ Monetary and Fiscal Actions: A Test of their Relative Importance in Economic Stabilization,” this Review (November 1968), pp. 11-24. _______ . “ The Quantity Theory of Money,” in J. Eatwell, M. Milgate and P Newman, eds. The New Palgrave—A Dic . tionary of Economics (London: Macmillan, 1987). MARCH/APRIL 1990 64 _______ and D. Meiselman. “ The Relative Stability of Mon etary Velocity and the Investment Multiplier in the United States, 1897-1958,” in Commission on Money and Credit Stabilization Policies (Prentice-Hall, 1963). _______ and A. Schwartz. “ Money and Business Cycles,” Review of Economics and Statistics (February 1963), reprinted in Milton Friedman, The Optimal Quantity of Money and Other Essays (Aldine Publishing Co., 1969), pp. 189-235. _______ a n d _______ Monetary Trends in the United States and the United Kingdom (University of Chicago Press, 1982). Goldfeld, S. M. “ The Demand for Money Revisited,” Brookings Papers on Economic Activity (1973:3), pp. 577-638. _______ “ The Case of the Missing Money,” Brookings Papers on Economic Activity (1976:3), pp. 683-730. Gould, B., J. Mills and S. Stewart. Monetarism or Prosperity? (London: Macmillan, 1981). Hicks, J. R. “ Mr. Keynes and the ‘Classics’: A Suggested In terpretation,” Econometrica (April 1937), reprinted in J. R. Hicks, Critical Essays in Monetary Theory (Oxford: Claren don Press, 1967), pp. 126-42. United Kingdom, 1862-1957: A Further Analysis,” Economica (February 1960), pp. 1-31. Lucas, R. E. Jr. “ Some International Evidence on OutputInflation Tradeoffs,” American Economic Review (June 1973), pp. 326-34. _______ . “ Money Demand in the United States: A Quantitative Review,” in K. Brunner and B. T. McCallum, eds., Money, Cycles and Exchange Rates: Essays in Honor of Allan H. Meltzer, Carnegie-Rochester Conference Series on Public Policy (Vol. 29, 1988), pp. 137-68. McCallum, B. T. "Inflation and Output Fluctuations: A Comment on the Dutton and Neumann Papers,” in K. Brunner and A. H. Meltzer, eds., The Problem of Inflation, Carnegie-Rochester Conference Series on Public Policy, (Vol. 8, 1978), pp. 277-88. Meltzer, A. H. "The Demand for Money: The Evidence from the Time Series,” Journal of Political Economy (June 1963), pp. 219-46. Mishkin, F. “ Does Anticipated Monetary Policy Matter? An Econometric Investigation,” Journal of Political Economy (February 1982), pp. 22-51. Johnson, H. G. “ Monetary Theory and Policy,” American Economic Review (June 1962), pp. 335-84. Modigliani, F. “ The Monetarist Controversy or, Should We Forsake Stabilization Policy?” American Economic Review (March 1977), pp. 1-19. King, R. G., and C. I. Plosser. “ Money, Credit, and Prices in a Real Business Cycle,” American Economic Review (June 1984), pp. 363-80. Phelps, E. S. “ Phillips Curves, Expectatons of Inflation and Optimal Unemployment over Time,” Economica (August 1967), pp. 254-81. Laidler, D. “ The Rate of Interest and the Demand for Money — Some Empirical Evidence,” Journal of Political Economy (December 1966), pp. 545-55. Phillips, A. W. “ The Relation Between Unemployment and the Rate of Change of Money Wage Rates in the United Kingdom, 1861-1957,” Economica (November 1958), pp. 283-99. _______ The Demand for Money: Theories and Evidence (International Textbook Publishing Co., 1969). _______ Monetarist Perspectives (Harvard University Press, 1982). _______ “ 'Buffer-Stock' Money and the Transmission Mechanism,” Federal Reserve Bank of Atlanta Economic Review (March/April 1987), pp. 11-23. _______ and J. M. Parkin. “ Inflation: A Survey,” Economic Journal (December 1975), pp. 741-809. Lipsey, R. G. “ The Relation Between Unemployment and the Rate of Change of Money Wage Rates in the FEDERAL RESERVE BANK OF ST. LOUIS Samuelson, P.A., and R. M. Solow. “Analytical Aspects of Anti-Inflation Policy,” American Economic Review (May 1960), pp. 177-94. Teigen, R. “ Demand and Supply Functions for Money in the United States: Some Structural Estimates,” Econometrica (October 1964), pp. 476-509. Tobin, J. “A General Equilibrium Approach to Monetary Theory,” Journal of Money, Credit and Banking (February 1969), pp. 15-29. _______ . “ The Monetarist Counter-Revolution Today— An Appraisal,” Economic Journal (March 1981), pp. 29-42. Federal Reserve Bank of St. Louis Post Office Box 442 St. Louis, Missouri 63166 The Review is published six times per year b y the Research and Public Information Department o f the Federal R eserve Rank o f St. Louis. Single-copy subscriptions are available to the public f r e e o f charge. Mail requests f o r subscriptions, back issues, o r address changes to: Research and Public Information Department, Federal R eserve Rank o f St. Louis, P.O. Rox 442, St. Louis, Missouri 63166. The views expressed are those o f the individual authors and do not necessarily reflect official positions o f the Federal Reserve Rank o f St. Louis o r the Federal R eserve System. Articles herein may b e reprinted provided the source is credited. Please provide the Rank’s Research and Public Information Department with a copy o f reprinted material.